Lecture Notes in Electrical Engineering 875

Pradeep Kumar Singh

Maheshkumar H. Kolekar
Sudeep Tanwar
Sławomir T. Wierzchoń
Raj K. Bhatnagar Editors

Emerging
Technologies
for Computing,
Communication
and Smart Cities
Proceedings of ETCCS 2021
Lecture Notes in Electrical Engineering

Volume 875

Series Editors

Leopoldo Angrisani, Department of Electrical and Information Technologies Engineering, University of Napoli
Federico II, Naples, Italy
Marco Arteaga, Departament de Control y Robótica, Universidad Nacional Autónoma de México, Coyoacán,
Mexico
Bijaya Ketan Panigrahi, Electrical Engineering, Indian Institute of Technology Delhi, New Delhi, Delhi, India
Samarjit Chakraborty, Fakultät für Elektrotechnik und Informationstechnik, TU München, Munich, Germany
Jiming Chen, Zhejiang University, Hangzhou, Zhejiang, China
Shanben Chen, Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai, China
Tan Kay Chen, Department of Electrical and Computer Engineering, National University of Singapore,
Singapore, Singapore
Rüdiger Dillmann, Humanoids and Intelligent Systems Laboratory, Karlsruhe Institute for Technology,
Karlsruhe, Germany
Haibin Duan, Beijing University of Aeronautics and Astronautics, Beijing, China
Gianluigi Ferrari, Università di Parma, Parma, Italy
Manuel Ferre, Centre for Automation and Robotics CAR (UPM-CSIC), Universidad Politécnica de Madrid,
Madrid, Spain
Sandra Hirche, Department of Electrical Engineering and Information Science, Technische Universität
München, Munich, Germany
Faryar Jabbari, Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA,
USA
Limin Jia, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China
Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland
Alaa Khamis, German University in Egypt El Tagamoa El Khames, New Cairo City, Egypt
Torsten Kroeger, Stanford University, Stanford, CA, USA
Yong Li, Hunan University, Changsha, Hunan, China
Qilian Liang, Department of Electrical Engineering, University of Texas at Arlington, Arlington, TX, USA
Ferran Martín, Departament d’Enginyeria Electrònica, Universitat Autònoma de Barcelona, Bellaterra,
Barcelona, Spain
Tan Cher Ming, College of Engineering, Nanyang Technological University, Singapore, Singapore
Wolfgang Minker, Institute of Information Technology, University of Ulm, Ulm, Germany
Pradeep Misra, Department of Electrical Engineering, Wright State University, Dayton, OH, USA
Sebastian Möller, Quality and Usability Laboratory, TU Berlin, Berlin, Germany
Subhas Mukhopadhyay, School of Engineering & Advanced Technology, Massey University,
Palmerston North, Manawatu-Wanganui, New Zealand
Cun-Zheng Ning, Electrical Engineering, Arizona State University, Tempe, AZ, USA
Toyoaki Nishida, Graduate School of Informatics, Kyoto University, Kyoto, Japan
Federica Pascucci, Dipartimento di Ingegneria, Università degli Studi “Roma Tre”, Rome, Italy
Yong Qin, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China
Gan Woon Seng, School of Electrical & Electronic Engineering, Nanyang Technological University,
Singapore, Singapore
Joachim Speidel, Institute of Telecommunications, Universität Stuttgart, Stuttgart, Germany
Germano Veiga, Campus da FEUP, INESC Porto, Porto, Portugal
Haitao Wu, Academy of Opto-electronics, Chinese Academy of Sciences, Beijing, China
Walter Zamboni, DIEM - Università degli studi di Salerno, Fisciano, Salerno, Italy
Junjie James Zhang, Charlotte, NC, USA
The book series Lecture Notes in Electrical Engineering (LNEE) publishes the
latest developments in Electrical Engineering - quickly, informally and in high
quality. While original research reported in proceedings and monographs has
traditionally formed the core of LNEE, we also encourage authors to submit books
devoted to supporting student education and professional training in the various
fields and applications areas of electrical engineering. The series cover classical and
emerging topics concerning:
• Communication Engineering, Information Theory and Networks
• Electronics Engineering and Microelectronics
• Signal, Image and Speech Processing
• Wireless and Mobile Communication
• Circuits and Systems
• Energy Systems, Power Electronics and Electrical Machines
• Electro-optical Engineering
• Instrumentation Engineering
• Avionics Engineering
• Control Systems
• Internet-of-Things and Cybersecurity
• Biomedical Devices, MEMS and NEMS
For general information about this book series, comments or suggestions, please
contact leontina.dicecco@springer.com.
To submit a proposal or request further information, please contact the Publishing
Editor in your country:
China
Jasmine Dou, Editor (jasmine.dou@springer.com)
India, Japan, Rest of Asia
Swati Meherishi, Editorial Director (Swati.Meherishi@springer.com)
Southeast Asia, Australia, New Zealand
Ramesh Nath Premnath, Editor (ramesh.premnath@springernature.com)
USA, Canada:
Michael Luby, Senior Editor (michael.luby@springer.com)
All other Countries:
Leontina Di Cecco, Senior Editor (leontina.dicecco@springer.com)
** This series is indexed by EI Compendex and Scopus databases. **

More information about this series at https://link.springer.com/bookseries/7818

Pradeep Kumar Singh · Maheshkumar H. Kolekar ·
Sudeep Tanwar · Sławomir T. Wierzchoń ·
Raj K. Bhatnagar
Editors

Emerging Technologies
for Computing,
Communication and
Smart Cities
Proceedings of ETCCS 2021
Editors
Pradeep Kumar Singh Maheshkumar H. Kolekar
Department of Computer Science Department of Electrical Engineering
KIET Group of Institutions IIT Patna
Ghaziabad, Uttar Pradesh, India Patna, Bihar, India

Sudeep Tanwar Sławomir T. Wierzchoń

Department of Computer Engineering Institute of Computer Science
Nirma University Polish Academy of Sciences
Ahmedabad, Gujarat, India Warsaw, Poland

Raj K. Bhatnagar
Department of Electrical Engineering
and Computer Science
University of Cincinnati
Cincinnati, OH, USA

ISSN 1876-1100 ISSN 1876-1119 (electronic)

Lecture Notes in Electrical Engineering
ISBN 978-981-19-0283-3 ISBN 978-981-19-0284-0 (eBook)
https://doi.org/10.1007/978-981-19-0284-0

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2022, corrected publication 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

This volume constitutes the referred proceedings of the second International Confer-
ence on Emerging Technologies for Computing, Communications, and Smart-
World (ETCCS-2021), held in Baba Farid College of Engineering and Technology,
Bathinda, Punjab, India, in August 21–22, 2021.
The conference was organized jointly with the technical support from Maharaja
Ranjit Singh Punjab Technical University, India, University of Arad, Romania, ISTE
India, IAC Education, and with the academic partnership of Southern Federal Univer-
sity Russia, SD College of Engineering and Technology, India, and UJW Poland, as
well. The conference was supported by the All India Council for Technical Educa-
tion (AICTE), Council of Scientific & Industrial Research (CSIR), and Ministry of
Electronics and Information Technology (MeitY). Technical support was provided
by the Digital India, National e-Governance Division.
Thematically, the works submitted to the conference set out five main paths. These
are as follows: (i) emerging computing technologies, (ii) network and computing
technologies, (iii) wireless networks and Internet of Everything (IoE), (iv) commu-
nication technologies, security, and privacy, and (v) next-generation computing tech-
nologies. The attractiveness of this topic has attracted many researchers and scientists,
and its potential is reflected in the large number of submitted papers. We hope that
the reports collected in this volume will be met with wide interest.
The inaugural speech was delivered by Professor Anil D. Sahasrabudhe,
Chairman, AICTE, New Delhi, India. Welcome address was given by Dr. Pardeep
Kaura, Dean Academics, followed by the introduction of conference speakers by
Dr. Nimisha Singh, Dean Trainings, BFCET. Dr. Jayoti Bansal, Principal, BFCET,
briefed about the conference information and contributions. The first keynote was
delivered by Dr. Sachin Kumar Mangla, University of Plymouth, UK, and vote of
thanks by Dr. Tejinderpal Singh Sarao, Dean R&D, BFCET. The keynote talks were
delivered by Dr. Anand Nayyar, Duy Tan University, Da Nang, Vietnam; Prof. Pao
Ann Hsuing, National Chung Cheng University, Taiwan; Dr. Sanjay Misra, Convent
University, Nigeria; Dr. Pastor Jr. Arguelles, Dean, College of Computer Studies;
Calabarzon, Philippines; and Dr. Zdzisław Pólkowski, from the UJW, Poland. These
keynote speakers also took part in technical discussion organized for the authors.

v
vi Preface

We are highly thankful to our valuable authors for their contribution and our
technical program committee for their immense support and motivation toward
making the ETCC-2021 a grand success. We are thankful to various session chairs
for chairing the session and giving their valuable suggestions; few of them include
Dr. Chaman Verma, Eötvös Loránd University (ELTE), Hungary, Dr. Vivek Sehgal,
Dr.YugalKumar, Dr. R.K. Saini, Dr. Rohit Tanwar, Dr.Anupan Singh, Dr. Sudhanshu
Tyagi, Dr. Ashwani Kumar, and many more co-session chairs.
The volume editors are thankful to Patron Dr. Gurmeet Singh Dhaliwal, Chairman,
Baba Farid Group of Institutions, and Co-Patron, Dr. Manish Goyal, Principal,
BFCET, Bathinda, for extending their support during the conference. BFCET is
thankful to the All India Council for Technical Education (AICTE), Council of
Scientific & Industrial Research (CSIR), and Ministry of Electronics and Information
Technology (MeitY) for providing grant to BFCET to support the conference.
Lastly, we express our sincere gratitude to our publication partner, LNEE Series,
Springer, for believing in us.

Editors
Ghaziabad, India Pradeep Kumar Singh
Patna, India Maheshkumar H. Kolekar
Ahmedabad, India Sudeep Tanwar
Warsaw, Poland Sławomir T. Wierzchoń
Cincinnati, USA Raj K. Bhatnagar
August 2021
Contents

Emerging Computing Technologies

Healthcare Transformation Traditional to Telemedicine:
A Portrayal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Babita Yadav and Sachin Gupta
Comparative Analysis of Semantic Similarity Word Embedding
Techniques for Paraphrase Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Shrutika Chawla, Preeti Aggarwal, and Ravreet Kaur
Machine Learning Approach for Identifying Survival of Bone
Marrow Transplant Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Varun Sapra, Luxmi Sapra, Yashi Bansal, Gunjan Chhabra,
and Rohit Tanwar
Human Disease Prognosis and Diagnosis Using Machine Learning . . . . . 41
Sunil Kumar, Harish Kumar, Rashi Agarwal, and V. K. Pathak
Incorporating Financial News Sentiments and MLP-Regressor
with Feed-Forward for Stock Market Prediction . . . . . . . . . . . . . . . . . . . . . . 55
Junaid Maqbool, Preeti Aggarwal, and Ravreet Kaur
Framework for Context-Based Intelligent Search Engine
for Structured and Unstructured Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Ritendra R. Sawale, Lokesh Nagar, and Vaibhav K. Khatavkar
Design and Development of Enhanced Secured Video
Communication System for Medical Video Sensor Network . . . . . . . . . . . 81
A. Sivasangari, D. Deepa, R. Vignesh, Suja C. Mana,
and B. Keerthi Samhitha
Detecting Twitter Hate Speech Using Sentiment Analysis . . . . . . . . . . . . . . 95
Arpita Tripathy, Anshika Goyal, Urvashi Tyagi, and Poonam Tanwar

vii
viii Contents

Comparative Evaluation of Semantic Similarity Upon Sentential

Text of Varied (Generic) Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Richa Dhagat, Arpana Rawal, and Sunita Soni
The Role of Artificial Intelligence and Data Science Against
COVID-19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Saumya Tripathi and Dilip Kumar Sharma
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based
Numerical Mapping Using Optimal Filtering . . . . . . . . . . . . . . . . . . . . . . . . . 137
Subhajit Kar, Madhabi Ganguly, and Aryabhatta Ganguly
A Novel Ensemble Model for Breast Cancer Diagnosis . . . . . . . . . . . . . . . . 151
Aalif Hamid
A Novel Smart Aging Approach for Monitor the Lifestyle
of Elderlies and Identifying Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Mohammed Shakil Malek, Pooja Gohil, Sharnil Pandya,
Anand Shivam, and Kuldeep Limbachiya
Spatial Analysis of Violent Crime Dataset Using Machine Learning . . . . 183
Falade Adesola, Ambrose Azeta, Sanjay Misra, Aderonke Oni,
Ravin Ahuja, and Ademola Omolola
Analysis of Violent Crime Dataset Using Support Vector Machine
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Falade Adesola, Ambrose Azeta, Sanjay Misra, Aderonke Oni,
Ravin Ahuja, and Ademola Omolola
Nigeria Human Population Management Using Genetic Algorithm
Double Optimized Fuzzy Analytics Engine Approach . . . . . . . . . . . . . . . . . 203
Abraham Ayegba Alfa, Sanjay Misra, Blessing Iganya Attah,
Kharimah Bimbola Ahmed, Jonathan Oluranti, Ravin Ahuja,
and Robertas Damasevicius
Flower Species Detection System Using Deep Convolutional Neural
Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Arun Solanki and Tarana Singh
Data Localization and Privacy-Preserving Healthcare for Big Data
Applications: Architecture and Future Directions . . . . . . . . . . . . . . . . . . . . 233
Ashwin Verma, Pronaya Bhattacharya, Yogesh Patel, Komal Shah,
Sudeep Tanwar, and Baseem Khan

Network and Computing Technologies

Design of Facial Recognition Based Touchless Biometric System . . . . . . . 247
Priya Mishra, Shashwat Singh Mann, Mohit Sharma,
and Madhulika Bhatia
Contents ix

Seeker Optimization for Linear Phase Fir Filter Design . . . . . . . . . . . . . . . 259

Harmandeep Kaur, Satish Saini, and Amit Sehgal
Cloud Supported Secure e-Card College Management System . . . . . . . . . 275
Garima Jain and Rashmi Mishra
An Optimization of VM Allocation and Migration in Cloud
Computing Systems for High Throughput and Low Tolerance . . . . . . . . . 291
Iram Warsi, Hiresh Gupta, and Monika
GraDex—Graph-Based Data Analytics for Extractive Text
Summarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Maher Thakkar, Siddhant Patel, and Jai Prakash Verma
Smart Traffic Monitoring with Fog and Cloud Computing . . . . . . . . . . . . . 317
Varsha Choudhary, Yashwant Singh, and Pooja Anand
A Review of Precision Agriculture Methodologies, Challenges,
and Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
Mohammedshakil Malek, Bachwani Dhiraj, Deep Upadhyaya,
and Dixit Patel
Power Quality Enhancement and Low Voltage Ride Through
Capability in Hybrid Grid Interconnected System by Using D-Fact
Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
T. Joshiram and S. V. R. Laskhmi Kumari
Optimal LFC Regulator for Frequency Regulation in Multi Area
Power System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
V. Santoshi Raju and P. Venkatesh
Optimal Power Flow Using Firefly Algorithm with Solar Power . . . . . . . . 375
Thangella Aravind and Balusu Srinivasa Rao
Designing of Learning Environment of an Intelligent Tutoring
System for Academic Learning Needs of Learning-Disabled
Learners Based on Survey Report of Region-Specific Target Group . . . . 389
Neelu Jyothi Ahuja, Monika Thapliyal, Anand Nayyar,
and Adarsh Kumar
Trusted 6G-Envisioned Dynamic Spectrum Allocation:
A Reference Model, Layered Architecture, and Use-Case . . . . . . . . . . . . . . 405
Pronaya Bhattacharya, Farnazbanu Patel, Vishaka Ralegankar,
Bhaumik Thakkar, Sudeep Tanwar, and Mohammad Abouhawwash
Experimental Evaluation of QFT Adders on IBM QX Hardware . . . . . . . 419
Simran Jakhodia, Divyanshu Singh, and Babita Jajodia
x Contents

Wireless Networks and Internet of Everything (IoE)

A Microcontroller Based Hardware Implementation to Detect
and Protect a Transformer Against Various Faults . . . . . . . . . . . . . . . . . . . . 439
Shafqat Nabi Mughal and Mustansir Nizam
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth
and Wi-Fi Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
Shubhangi Verma, O. P. Singh, Sachin Kumar, and Sumita Mishra
A Proposal of Iot Based Evapotranspiration Irrigation System
for Water Conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
Simrat Walia and Jyotsna Sengupta
A Comprehensive Study of “etcd”—An Open-Source Distributed
Key-Value Store with Relevant Distributed Databases . . . . . . . . . . . . . . . . . 481
Husen Saifibhai Nalawala, Jaymin Shah, Smita Agrawal, and Parita Oza
AgriBot: Smart Autonomous Agriculture Robot for Multipurpose
Farming Application Using IOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
Hari Mohan Rai, Manish Chauhan, Himanshu Sharma,
Netik Bhardwaj, and Lokesh Kumar
Mobile Application Voting System: A Means to Achieve a Seamless
Election Process in Developing Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Abidemi Emmanuel Adeniyi, Roseline Oluwaseun Ogundokun,
Sanjay Misra, Jonathan Oluranti, and Ravin Ahuja
Adoption of Blockchain for Data Privacy in 6G-Envisioned
Augmented Reality: Opportunities and Challenges . . . . . . . . . . . . . . . . . . . 519
Umesh Bodkhe, Ashwin Verma, Deepti Saraswat,
Pronaya Bhattacharya, and Sudeep Tanwar
Integration Blockchain for Data Sharing and Collaboration
in Mobile Healthcare Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
Inaam Zia, Pawan Singh, Anil Kr. Tiwari, and Amit Pandey

Communication Technologies, Security and Privacy

Reusability Estimation of Object-Oriented (O–O) Systems Using
SOM Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
Bharti Bisht and Parul Gandhi
The First Corpus for Detecting Fake News in Hausa Language . . . . . . . . 563
Sukairaj Hafiz Imam, Abubakar Ahmad Musa, and Ankur Choudhary
A Tour Towards the Security Issues of Mobile Cloud Computing:
A Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
Vaibhav Kalra, Shubhankar Rahi, Neerav pawar, Poonam Tanwar,
and Ms Shweta Sharma
Contents xi

Subpixel Image Registration Using FFT-Based Technique . . . . . . . . . . . . . 591

Swati Jindal, Bhaskar Dubey, and Jitali Patel
Impact of Green Communication and Technology System . . . . . . . . . . . . . 605
Sumit Dhariwal and Avinash Raipuria
Recent Advancement in Fine-Grained Access Control and Secure
Data Sharing Scheme for Distributed Environment . . . . . . . . . . . . . . . . . . . 617
Vidhi Pandya, Atufaali Saiyed, and Khushi Patel
Context-Based Multi-System Integration for Authorization Using
Dynamic Privilege Escalation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
Nidhi Sinha, Meenatchi Sundram, and Abhijit Sinha

Next Generation Computing Technologies

Machine Learning-Based Security Solutions for Healthcare:
An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
Pallavi Arora, Baljeet Kaur, and Marcio Andrey Teixeira
Fake News Detection Techniques: A Survey . . . . . . . . . . . . . . . . . . . . . . . . . . 661
Sahitya Diwakar, Divyansh Bansal, Raghav Jagya, Poonam Tanwar,
and Shweta Sharma
Impact of Image Classification in Dermatology: Automated
Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671
Anukriti Singh, Devam Kumar Sinha, and Namrata Dhanda
IPFS-Based Image Sharing Using Perceptual Hash Code
with Digital Right Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683
D. Geethanjali, R. Priya, and R. Bhavani
Mid-Price Prediction Using Online Kernel Adaptive Filtering . . . . . . . . . . 701
Shambhavi Mishra, Tanveer Ahmed, and Vipul Mishra
Blockchain-Based Model for Secure IoT Communication in Smart
Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715
Gawhar Hameed, Yashwant Singh, Shahidul Haq, and Bharti Rana
Internet eXchange Points: A Systematic Case Study in Indian
Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
Deepti Saraswat, Pronaya Bhattacharya, Samprat Bhavsar,
Hardik Dave, Karan Patel, Shivam Punjani, Ashwin Verma,
and Sudeep Tanwar
Review on Security of Internet of Things: Security Requirements,
Threats, and Proposed Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
Rayeesa Muzafar, Yashwant Singh, Pooja Anand,
and Zakir Ahmad Sheikh
xii Contents

Emerging Security Issues in IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757

Umaira Ahad, Yashwant Singh, and Pooja Anand
Robust and Imperceptible Multiple Watermarking Using
Transform Domain Algorithm for Digital Media . . . . . . . . . . . . . . . . . . . . . . 769
Namita Agarwal, Amit Kumar, and Pradeep Kumar Singh
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain
Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779
Pronaya Bhattacharya, Ashwin Verma, and Gulshan Sharma
Correction to: Impact of Green Communication and Technology
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C1
Sumit Dhariwal and Avinash Raipuria

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793

About the Editors

Dr. Pradeep Kumar Singh is currently working as a Department of Computer

Science, KIET Group of Institutions, Delhi-NCR Campus, Ghaziabad, India. Dr.
Singh is having life membership of Computer Society of India (CSI), Life Member
of IEI and promoted to Senior Member Grade from CSI and ACM. He is Associate
Editor of International Journal of Information Security and Cybercrime (IJISC) a
scientific peer reviewed journal from Romania. He has published more than 120
research papers in various International Journals and Conferences of repute. He has
received three sponsored research projects grant from Govt. of India and Govt. of
HP worth Rs 25 Lakhs. He has edited total 10 books from Springer and Elsevier and
also edited several special issues for SCI and SCIE Journals from Elsevier and IGI
Global. He has Google scholar citations 1451, H-index 20 and i-10 Index 49 in his
account. He is section editor of Discover Internet of Things, Springer Journal.

Dr. Maheshkumar H. Kolekar is working as Associate Professor in Department of

Electrical Engineering at Indian Institute of Technology Patna, India. He received
the Ph.D. degree in Electronics and Electrical Communication Engineering from
the Indian Institute of Technology Kharagpur in 2007. During 2008 to 2009, he
was a Post-Doctoral Research Fellow with the Department of Computer Science,
University of Missouri, Columbia, USA. During May to July 2017, he worked
as DAAD fellow in Technical University Berlin where he worked in the research
area EEG signal analysis using machine learning. His research interests are in the
area of Digital Signal, Image and Video Processing, Video surveillance, Biomed-
ical Signal Processing, Deep Learning. Recently, he has authored a book titled as
Intelligent Video Surveillance Systems: An Algorithmic Approach, CRC Press, Taylor
and Francis Group, 2018. He has successfully completed R&D Project Sponsored
by Principal Scientific Advisor to Government of India on abnormal human activity
recognition.

Dr. Sudeep Tanwar (Senior Member, IEEE) is currently working as a Professor with
the Computer Science and Engineering Department, Institute of Technology, Nirma
University, India. He is also a Visiting Professor with Jan Wyzykowski University,

xiii
xiv About the Editors

Polkowice, Poland, and the University of Pitesti in Pitesti, Romania. He received

B.Tech. in 2002 from Kurukshetra University, India, M.Tech. (Honor’s) in 2009
from Guru Gobind Singh Indraprastha University, Delhi, India and Ph.D. in 2016 with
specialization in Wireless Sensor Network. He has authored two books and edited 13
books, more than 250 technical articles, including top journals and top conferences,
such as IEEE Transactions on Network Science and Engineering, IEEE Transac-
tions on Vehicular Technology, IEEE Transactions on Industrial Informatics, IEEE
Wireless Communications, IEEE Networks, ICC, GLOBECOM, and INFOCOM.
He initiated the research field of blockchain technology adoption in various verti-
cals, in 2017. His H-index is 44. He actively serves his research communities in
various roles. His research interests include blockchain technology, wireless sensor
networks, fog computing, smart grid, and the IoT. He is a Final Voting Member of the
IEEE ComSoc Tactile Internet Committee, in 2020. He is a Senior Member of IEEE,
Member of CSI, IAENG, ISTE, and CSTA, and a member of the Technical Committee
on Tactile Internet of IEEE Communication Society. He has been awarded the Best
Research Paper Awards from IEEE IWCMC-2021, IEEE GLOBECOM 2018, IEEE
ICC 2019, and Springer ICRIC-2019. He has served many international confer-
ences as a member of the Organizing Committee, such as the Publication Chair for
FTNCT-2020, ICCIC 2020, and WiMob2019, a member of the Advisory Board for
ICACCT-2021 and ICACI 2020, a Workshop Co-Chair for CIS 2021, and a General
Chair for IC4S 2019, 2020, and ICCSDF 2020. He is also serving the editorial boards
of Computer Communications, International Journal of Communication System, and
Security and Privacy. He is also leading the ST Research Laboratory, where group
members are working on the latest cutting-edge technologies.

Prof. Sławomir T. Wierzchoń received M.Sc. and Ph.D. degrees in Computer

Science from Technical University of Warsaw, Poland. He holds Habilitation (D.Sc.)
in Uncertainty Management from Polish Academy of Sciences. In 2003 he received
the title of Professor from the President of Poland. Currently he is full professor
at the Institute of Computer Science of Polish Academy of Sciences. His research
interests include Computational Intelligence, Uncertainty Management, Informa-
tion Retrieval, Machine Learning and Data Mining. He is an author/co-author of
over 100 peer reviewed papers in international journals and international confer-
ences. He published, as author/co-author, 11 monographs from the field of Machine
Learning. In the period 2000–2013 he co organized 13 international conferences on
intelligent information systems. Co-authored proceedings from these conferences
were published by Springer. He co-edited two volumes of proceedings of the inter-
national conference on Computer Information Systems and Industrial Management
and he has served as a guest co-editor of three special issues of Information & Control
journal. Currently, he is member of the editorial board for some international journals,
as well as member of many program committees for international conferences. He
cooperated with medical centers in the area of statistical data analysis and knowledge
discovery in data bases.
About the Editors xv

Prof. (Dr.) Raj K. Bhatnagar is currently working as a Professor of Computer

Science, Department of Electrical Engineering and Computing Systems, University
of Cincinnati, USA. He completed his B.Tech. in IIT Delhi, India (1979) and Ph.D.
in University of Maryland, USA (1989). He worked as a Visiting Professor in many
institutions and he has taken many administrative responsibilities along with teaching
profession at University of Cincinnati such as Graduate Program Director of CS
Department and ECECS Department, Director of UG Computer Science Program,
Co-Director of Graduate Certificate in BioInformatics, Assoc. School Director and
Graduate Program Director School of Computing Sciences and Informatics, Interim
Department Head, Computer Science Department. His research interests includes
Data mining, AI and pattern recognition problems, knowledge discovery and big
data and cloud computing environments.
Emerging Computing Technologies
Healthcare Transformation Traditional
to Telemedicine: A Portrayal

Babita Yadav and Sachin Gupta

Abstract The biggest challenge nowadays is to cope with massive amount of data
generated digitally which is being originated from various healthcare entities and
sources. The use of this data in developing various healthcare models is to predict
diseases as well as creating services that can be benefitted to humanity. The business
of redesign through information technology practices which are broadening their
spectrum to made possible that each and every health seeker should be benefitted
through each and every ICT practices. As well as the data generated from various
sources should be analyzed properly so that it can be taken out from healthcare
information systems for prediction of future diseases and help medical practioners
to advise patient in more advance ways. This is only possible through various change
that took place in information technology. This merging of healthcare experts from
medical domain with big data and predicting value points from this data will create a
total difference in Medical industry. The prediction not only help health seekers but
also helps in reducing the cost that a patient spend on health advises. The innovation in
healthcare system storing information will not only improve service quality but also
reduce delivery cost. Big data in clinical practices will also help in better management
of data which will prove an effecting way of various data of patient as well as
doctors in their expert field. In this chapter, the transformation of healthcare sector
is amazingly changing because of taking perspective of big data into account.

Keywords Telemedicine · Health care · Big data · Clinical practices

B. Yadav (B) · S. Gupta

MVN University, Haryana, India
e-mail: Babita.yadav@mvn.edu.in
S. Gupta
e-mail: Sachin.gupta@mvn.edu.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 3
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_1
4 B. Yadav and S. Gupta

1 Introduction

Analytics in healthcare system including taking account analysis of health data for
patient. The patient records, cost of disease and diagnosis of data as well as in both the
states micro- and macro-level are being considered. There are various visualization
tool that help health managers by providing real-time insight about health data.
Systematic analysis of data helps various professional working in healthcare system
to get the findings and various opportunities that can improve health management
system. Various trends and patterns can be generated from health data along with
real-time data [1]. With these trends, various actionable insights from data can be
taken out to improve existing health systems and services. When there is system
of improved quality care, there will be improved diagnosis system and methods of
improved treatment which in turn provide better health services to health system.
The health analytics also play an important role in initiating awareness about
disease. It also helps public care units in predicting outbreak of certain disease [2].
Forecasting arrangements of healthcare units as well as arranging control of disease
spread in healthy population. Finding trends will also uncover the certain hidden
facts about disease that normal population should know to protect themselves from
getting infection.
The various steps adopted to obtain, review and management of data related to
health for entire population are carried out by public health organizations in an effort
to maintain health of population [3]. The data related to health of public may include
reports of mortality, demographic data, socioeconomic data and data of diagnosis of
patient, and patient claiming medical claims. Health analytics monitor various disease
using various patterns in population helping various public health organizations that
may utilize analytics to monitor trends of disease and deriving patterns in certain
category of populations which can be used to guide various programs that control
disease and priorities setting for assigning resources to populations for various health
need [4].
Predictive analysis plays an important role in healthcare system helping health
providers as well as health seekers, medical claim companies applying machine
learning to health care.
This type of analysis results in designing a model for prediction which can help
in taking decision in a better way which may help in investing planning in healthcare
system, helping health ministry to invest funds on health care and serving population
in a much better way.

1.1 Telemedicine

According to World Health Organization (WHO), this term is said to be delivery of

various services related to healthcare services.
Healthcare Transformation Traditional to Telemedicine … 5

The areas where distance is a crucial factor by which professional healthcare

providers utilize various kind of information and technologies to communicate for
exchanging various information related to patient diagnosis, treating them and how
they can prevent disease and injuries. An electronic use of telehealth information and
technologies to support how patient at a long distance can be cured without visiting
them physically. There is a system which is required to be an interface between two
remote locations and exchange information to avail consultancy regarding health of
patient [5].
There is rapid change in the field of technology which has made the existence of
telemedicine really a need. It’s really became a technology to be used in health care to
provide services related to health easily reach patients to locations far away. This tech-
nology has now transformed into an integrated service which is very much utilized
in hospitals, homes, private physician offices, and in providing various healthcare
facilities. The history of Telemedicine justified it use in the form of electromagnetic
signals. A video consultation between patient and doctor made possible using this
telemedicine technology [6].
In its original state Telemedicine was only used for treating patients which were
located in places of remote areas and in case of local health facilities are shorter
as well as less professional medical health caretaker [7]. Now the current state of
Telemedicine is used as a very convenient tool in medical care. Everyone in this life
is too much busy the health seekers really to save their time waiting in traffic or at
hospital queue. Existing Telemedicine framework can be adopted to be untilled and
implemented during pandemic times via existing Tele devices as shown in Fig. 1.

Fig. 1 Telemedicine adoption in pandemic

6 B. Yadav and S. Gupta

1.2 Current State: Telemedicine Today

Now a days the face of telemedicine has changed so fast because of technology
advancement at exponential levels. The low cost of mobile and availability of Internet
has given birth to live video telemedicine thorough face time, skype and many
telemedicine applications ready to use at time of extreme cases [8].

1.3 Major Contributions

• The current state of art related to telemedicine has been identified.

• The medical experts are overburdened and sometime unavailable and it’s expected
by various patients to have individual expectation [9]. This expectation is more
related to convenient care, individual exclusive attention. Sometime due to
absence of some expert hospital units transfer and outsource the cases to some
big extensive care units.
• In this age of mobile health various options given to patient in the form of
mobile health applications, portals to take appointment etc. [10]. The patients
have already started to use these technologies to monitor and using tracking their
health continuously.
• Evaluation of the user acceptance, diagnostic accuracy and educational value of
telemedicine. (Setting up hypothesis and research questionnaires to conduct a
survey to retrieve real picture of Telemedicine in India from the perspective of
medical experts) in future.
• After review study in this paper proposing a model of approach and techniques to
that how Telemedicine can be implemented in remote and urban areas addressing
various challenges based on above study including field study in healthcare sector.
• Analysis of telehealth data of patients generated by the proposed model to compute
its efficiency.

2 Background Study

The demand of patients is prioritizing convenience and less expensive care. The
author stated work which focus specifically described on various important use of
telemedicine in disasters and emergency situation in health of public in various
emergencies. Telemedicine has a huge challenge in creation and implemented but
US health systems that are already implemented innovation in telemedicine [11]. The
spread of virus globally rapidly become a global public health pandemic. The testing
of patients because of this disease is started rising. The symptoms are not known
completely. This has posed various challenges with managing health of such a vulner-
able population. This has pointed out various fields where this technology can help
a lot [12]. COVID-19 outbreak has impacted complete global healthcare systems.
Healthcare Transformation Traditional to Telemedicine … 7

The challenge imposed to many fields has also impacted clinical disciplines. The
outbreak is so severe that adoption of telecommuting is an effective working option,
others raise concerns about various potential downsides [13]. Persons with disabil-
ities in this pandemic unable to visit hospital and clinic. Traditional healthcare was
depending upon relationship between patient–physician and always been centered
around contacting face-to-face. The challenge has been addressed by applicability,
convenience and cost effectiveness of telemedicine [14]. The use of mobile phone
technologies for promoting health care and preventing disease came into existence
so rapidly. The purpose of this pilot evaluation study is to find out the efficiency
of telemedicine and its importance. The proposed systematic review will evaluate
the current evidence regarding finding out effectiveness of telemedicine and its role
in situation like pandemic [15]. There are various patients in areas which are under-
served and are at remote location. They are facing scarcity of specialists. The popu-
larity about has increased dramatically in the situation like COVID-19. The experts
related to medical are getting proper training regarding this technology including
demonstration of the telemedicine programs. The telemedicine practice is flourishing
with support from government agencies and various other sources to implement this
technology. The technology like telemedicine despite of its popularity is not univer-
sally adopted in many regions and many people. The objective of this research paper
is to show and review that how much technology has been in progress that this tech-
nology can prove a very emergency tool if being utilized properly in specialty care
services [16]. Developed country like USA accepted telemedicine so widely. The
programs related to telemedicine have been in existence since the 1960s, but only in
the last two or three years, have they begun to proliferate and can be taken as miracle
in situation like pandemic.

2.1 Gaps Identified

There is a lot of study and research work done by researchers where it has been found
that despite having strategies, approaches existing regarding telemedicine, there is
still a lack of adoption of proposed solutions. There are challenges and gaps still left
in adoption of telemedicine that are extracted from the research article from survey
of literature. Some of the gaps identified from literature are mentioned below.
1. There is a general lack of awareness among patients in most urban and rural
areas.
2. Professional training of medical experts is an area of concern.
3. The various methods of medical information exchange are still developing.
Standardization in information format to suit telemedicine is another gap area.
4. Technologically challenged practitioners and patients face resistance in adoption
of telehealth care.
5. The cost factors of establishing e-medicine and telehealth care are currently
high, due to less permeation in the market.
8 B. Yadav and S. Gupta

3 Current Trends in Telemedicine

Telemedicine can be considered to be the most effective diagnosing of treatment of

patients by means of technology related to telecommunication. If this can be added
and supported in government funding agencies, it can be further thereby providing
substantial healthcare to low income regions also. Earliest records of telemedicine are
in the first half of the twentieth century when ECG was transmitted over telephone
lines. Today to then, telemedicine has been a long way in terms of both delivery
of healthcare and technology. There are different types of telemedicine services
including store and forward, real-time and remote or self-monitoring services, which
provide different levels of educational, healthcare delivery and management options
for screening of diseases. These services can be scaled to spread all over the world
in a pandemic like situation for best benefits to humankind [10]. Even though
telemedicine cannot be a solution to all the problems, it can surely help decrease
the burden of the healthcare system to a large extent.
In various research studies, doctors utilizing telemedicine have acknowledged
that they have successfully treated patients virtually. Sometimes, many patients visit
doctor with minor medical conditions or questions then it does not require in-person
diagnosis or treatment. It has been seen that 75% of all regular doctor as well as visits
are either unnecessary and can be handled very effectively via phone or by video [5].
As well like patients with sinusitis, cold or flu will avoid coming out and will not
spread the virus further. According to research sources, 60% of telemedicine visits
ended with the doctor writing a prescription.
The latest trends in telemedicine today exploring benefits, the attitudes of patients
and doctors, and telemedicine’s potential cost-savings. The billion global market
percentage adopting telemedicine technologies is valued $17.8 billion in 2014 as
well as there will be about 800,000 online consultations in developing countries in
2015.

4 Reason of Adoption

The improvement in patient care as well healthcare delivery by remote methods is

the first reason for rapid adoption of this technology. Telemedicine improves care
delivery and patient communication. The consultation provided to health seekers by
continuing multispecialty consultation via various mode of telemedicine. There can
be several reasons of its adoption because of rise in digital revolution as well as
digital literacy as shown in Fig. 2.
Healthcare Transformation Traditional to Telemedicine … 9

Fig. 2 Reasons of adoption

4.1 Patient Per Doctor Unavailability (Patient:Doctor)

The availability of doctors in developing countries like India has been estimated as
per population according to registration comes to be low as compared to population
growth. This has been seen in literature that 4.8 practicing doctors per 10,000 popu-
lation available in India in 2014, in contrast to the belief of having seven doctors
per 10,000 people. It is also been estimated that the country would also be able to
achieve a ratio of about 6.9 practicing doctors per 10,000 people only by 2030 as
shown in Fig. 3.
The findings also been shown that the current availability of doctors per 10,000
people and their growth prospects over the next 15-year period will be same, and
also, it looks like an impossible task to achieve accepted doctor-population ratio of
1:1000 by 2030. Need of telemedicine is much needed to be implemented as widely
accepted method for patient care.
• As well as adding it to more our 5.5 million km road network transports 64.5%
or as well as 90% of India’s total passenger traffic uses this road network to
commute. The government itself has acknowledged that congestion on roads has
been increasing. Telemedicine will help people and more and more doctor will
be available for treatment which were later unavailable because of shortage of
medical practioners in country and around the world. Implementation and adop-
tion of telemedicine will help patient to get timely treatment and medical care.
10 B. Yadav and S. Gupta

Fig. 3 Patient to doctor ration 7:10,000, need of telemedicine

Doctors can be in contact with patients frequently. The outcome of telemedicine

adoption will be improved outcome of patients as well as better treatment. The gaps
will be filled between patient and doctors because of this adoption. The doctors
will available for more patient in one day. The health seekers will be saved from
no less shows as well as late cancelations. The patients can keep regular contact
with their doctors. The doctors can extend access to more specialists will save
time and money that patients have to spend on travel for appointment especially
will help rural patients. Discussing with their doctor from the comfort of their
own home will result in shorter wait time; the care will be more convenient and
on-demand. The involvement in decisions regarding treatment will be more.

4.2 Increase in Smart Phone Users

The number of smart phone users in India was estimated to reach over 760 million
in 2021, with the number of smart phone users worldwide forecasted to exceed to
3.8 billion users in 2021.
In the world of 5G, India is also moving very rapidly at the path of digital revolu-
tion which is powered by increase in broadband and Internet availability at a very low
cost; the data exponentially has been up taken [17]. The government’s initiatives also
focus on digitalization which has also increased trend of adopting technology across
industries. Internet statistic says that due to digital revolution, the use of smart phone
in India has much increased from 2015 to 2025. As well the estimation about smart
phone users globally increased and to be exceed to 3.8 billion users in 2020 as well
the number of smart phone users worldwide is almost projected to a figure nearly
2.7 billion by 2019 [18].
Healthcare Transformation Traditional to Telemedicine … 11

The rise in increase of smart phone users in country has been estimated to reach
over 760 million by 2021. The statistic has been shown in Fig. 4.
The popularity of telemedicine has been shown growth from 2010, 35% to 76%
in 2017 as shown in Fig. 5. Still, there is a pattern of visiting hospital then to consult
doctor on Internet or phone. There are reluctance that has been still observed adopting
Telemedicine [19].
As more people start using video chat apps and smart phones, telemedicine will
become even more intuitive and accessible to doctors and patients. Literature shows
that the number of video callers increased up to an estimated 380 million this year
as compare to 2010 which was up 63 million as shown in Fig. 6.
In developing countries like India, more than 85% of population has Internet
access. So partial infrastructure for adopting telemedicine is already existing, waiting
to be utilized. About 30% of patients already use computers or mobile devices to

Fig. 4 Users increased by the year from 2029 to 2021

Fig. 5 35% to 76% increase

from 2010 to 2017
12 B. Yadav and S. Gupta

Fig. 6 Rise of users

check for medical or diagnostic information. It can be said that the adoption of
telemedicine is been seen globally and population is equipped with basic require-
ments to adopt telemedicine completely as well as Nearly, half of healthcare profes-
sionals already planned to integrate mobile apps into their practices in coming
years.

4.3 Stringent Cyber-Security Measures

The adoption of telemedicine is relying totally on technology because of this cyber-

security has become a necessity. The contribution of cyber-security methods has
made that the industry of health care is no more prone to cyber-attacks but with
telemedicine, need of information protection has become priority [19]. The support
of cyber-security industry in recent years has shown a backbone to health care which
can be prove adoption of telemedicine with open arms.

4.4 Modeling Telemedicine Using Regression

To tackle the difficulties of telemedicine as well as defining relationship between

patient, doctors and other dependent and independent variables, linear regression
can be modeled to fit linear equation between variables [11]. The model supports
relationship between the variables of interest like, time, cost, on time support and
personal support, traffic, mobile users, digital literacy. The strength between patient
and doctor variable is shown with the help of a scatter plot [14]. A valuable numerical
Healthcare Transformation Traditional to Telemedicine … 13

measure of association between patient and doctor variables is the correlation coef-
ficient, which is a value between −1 and 1, indicating the strength of the association
of the observed data for the variables patient and doctor.
A linear regression line has an equation of the form Y = a + bX, where X is
the explanatory variable and Y is the dependent variable. The slope of the line is b,
and a is the intercept (the value of y when x = 0).

Hypothesis
In case of various variables assumed to be fit in regression equation, the multiple
regression equation will take form: y = b1x1 + b2 x2 + … + bnxn + c. Here, bi’s
(i = 1,2…n) are the regression coefficients, which represent the value at which the
criterion variable changes when the predictor variable changes.

y = b1x1 + b2x2 + . . . + bnxn + c (1)

time = b1, cost b2, on time support b3 and personal support b4, traffic b5, mobile
users b6, digital literacy b7 it comes out to be

y = b1x1 + b2x2 + b3x3 + b4x4 + b5x5 + b6x6 + b7x7 + c (2)

where x is the value of the independent variable (x)

5 Conclusion

There is no doubt that telemedicine will prove very effective in situations like
pandemic. The need for transforming to an era of telemedicine is actually employing
to its maximum utility which should be realized. The purpose of this research work is
to showcase actual scenario of telemedicine as well as portraying why telemedicine
needs to be implemented country like India where roads are full of traffic as well as
availability of doctors is terribly low as compare to population. The support of govern-
ment and healthcare organizations must produce required strategies to encourage
fourfold commitment: to support and provide funding for telemedicine.

References

1. Mishra SK, Kapoor L, Singh IP (2009) Telemedicine in India: current scenario and the future.
Telemed e-Health 15(6):568–575
2. Ganapathy K, Ravindra A (2009) Telemedicine in India: the Apollo story. Telemed e-Health
15(6):576–585
3. Chandwani RK, Dwivedi YK (2015) Telemedicine in India: current state, challenges and
opportunities. In: Transforming government: people, process and policy
4. Bagchi S (2006) Telemedicine in rural India. PLoS Med 3(3): e82
14 B. Yadav and S. Gupta

5. Pal A, Mbarika VWA, Cobb-Payton F, Datta P, McCoy S (2005) Telemedicine diffusion in a

developing country: the case of India (March 2004). IEEE Trans Inf Technol Biomed 9(1):59–65
6. Mishra SK, Gupta SD, Kaur J (2007) Telemedicine in India: initiatives and vision. In: 2007
9th international conference on e-health networking, application and services. IEEE, pp 81–83
7. Dasgupta A, Deb S (2008) Telemedicine: a new horizon in public health in India. Indian J
Community Med: Official Publ Indian Assoc Preventive Soc Med 33(1):3
8. Myers MB (2003) Telemedicine: an emerging health care technology. Health Prog 22(3):219–
223
9. Ebad R, Jazan KSA (2013) Telemedicine: current and future perspectives
10. Huston TL, Huston JL (2000) Is telemedicine a practical reality? Commun ACM 43(6):91–95
11. Hollander JE, Carr BG (2020) Virtually perfect? Telemedicine for Covid-19. New England J
Med
12. Korupolu R, Stampas A, Gibbons C, Jimenez IH, Skelton F, Verduzco-Gutierrez M (2020)
COVID-19: Screening and triage challenges in people with disability due to spinal cord injury.
Spinal Cord Ser Cases 6(1):1–4
13. Lincoln MH, Khan R, Cai J (2020) Telecommuting: a viable option for medical physicists amid
the COVID-19 outbreak and beyond. In: Medical physics
14. Ananthakrishnan AN, Singh S (2020) The doctor will call you now! telemedicine in the midst
of a pandemic. In: Clinical gastroenterology and hepatology
15. Burnham JP, Fritz SA, Yaeger LH, Colditz GA (2019) Telemedicine infectious diseases consul-
tations and clinical outcomes: a systematic review and meta-analysis protocol. Syst Rev
8(1):135
16. Wootton R (2001) Telemed BMJ 323(7312):557–560
17. Mishra S (2015) Drivers of mobile-commerce acceptance in India: an empirical study. In: XIV
international business and economy conference, Bangkok, Thailand
18. Bhave K, Jain V, Roy S (2013) Understanding the orientation of gen Y toward mobile
applications and in-app advertising in India. Int J Mobile Marketing 8(1)
19. Black PE, Scarfone K, Souppaya M (2008) Cyber security metrics and measures. In: Wiley
handbook of science and technology for homeland security, 1–15
Comparative Analysis of Semantic
Similarity Word Embedding Techniques
for Paraphrase Detection

Shrutika Chawla, Preeti Aggarwal, and Ravreet Kaur

Abstract Most of the state-of-the-art plagiarism detection tools focus on verbatim

reproduction of a document for plagiarism identification (PI), not keeping into
account its semantic properties. Recently, deep learning models have shown consid-
erable performance in identifying paraphrases using word embedding approach.
This paper gives an overview and comparison of the performances of five word-
embedding based deep learning models in the field of semantic similarity detection,
such as TF-IDF, Word2Vec, Doc2Vec, FastText, and BERT on two publicly avail-
able corpora: Quora Question Pairs (QQP) and Plagiarized Short Answers (PSA).
After extensive literature review and experiments, the most appropriate text prepro-
cessing approaches, distance measures, and the thresholds have been settled on for
detecting semantic similarity/paraphrasing. The paper concludes on FastText being
the most efficient model out of the five, both in terms of evaluation metrics, i.e.,
accuracy, precision, recall, F1-score, receiver operating characteristic (ROC) curve,
and resource consumption. It also compares all the models with each other based on
the above-mentioned metrics.

Keywords Paraphrasing · Plagiarism · Paraphrase detection · Plagiarism

detection · Paraphrase

1 Introduction

Paraphrasing is the procedure of rewriting a statement to change its form without

changing the meaning of the original text. Many deep learning models have shown
a promising performance in detecting semantic similarity between documents using
word embedding technique. Word embedding is a representation of document vocab-
ulary which is capable of capturing the semantic context of the document as a whole,
or the individual words in a document. The idea behind word embedding is assigning

S. Chawla (B) · P. Aggarwal · R. Kaur

UIET, CSE Department, Panjab University, Chandigarh, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 15
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_2
16 S. Chawla et al.

vector values to text, as the technique opens doors to various text evaluation and
analysis approaches in the field of linear algebra.
The main objective of this paper is to review and compare the performance of
various word-embedding based models (including some deep learning models) for
semantic similarity detection which leads to the following contributions: (a) System-
atic outline of corpus-based word embedding models, and (b) Elucidation of the best
models, preprocessing techniques and thresholds for plagiarism detection and the
kind of corpus they work well with.
The motivation behind this research is to in turn develop a framework which
uses machine learning and statistics to measure semantic similarity between given
documents and detect paraphrase plagiarism. The systematic study helps in cherry-
picking the mechanisms and hyperparameters best suited for a new framework.
This research compares five models such as Word2Vec [1], Doc2Vec [2], BERT
[3], TF-IDF [4], FastText [5] on two publicly available corpora, namely Quora Ques-
tion Pairs and Plagiarized Short Answers (PSA), and in turn, these models are evalu-
ated in terms of accuracy, precision, recall and F1-score. Before applying the models,
data preprocessing techniques, distance measures, and a threshold value for each
model were set after careful experimentation on each corpus.
The paper is organized as follows: Section 2 briefly discusses existing work in the
related field as well as specific work in the paraphrase detection area. Section 3
discusses approach followed in this paper, models implemented and the results
derived from the experiments. Section 4 draws conclusion and discusses future scope.

2 Related Work

The use of computers in the field of Natural Language Processing (NLP) is a challenge
due to the equivocacy of texts and passages. For example, the term ‘mango’ can be
referred to as a fruit as well as a clothing brand. Hence, the semantic features of a
text play a major role over and above its linguistic features.
A large number of established researchers have contributed to various plagia-
rism detection stages [6]. For handling plagiarism and paraphrasing cases of higher
complexities, [7] suggests that researchers should focus on linguistic, syntactic,
and most importantly semantic information of text rather than just on verbatim
reproduction.
Table 1 summarizes various works and approaches in the field of paraphrase
detection.
Inspired by the recent successes of neural networks (NNs) in the fields of informa-
tion retrieval and natural language processing, this paper experiments with various
deep learning models to test semantic similarity detection power of the same and find
out various factors affecting their performance. The succeeding section discusses the
approach adopted in this research paper.
Comparative Analysis of Semantic Similarity Word … 17

Table 1 Related work in paraphrase detection field

Research Technique/Algorithm Dataset Observations Performance
work
[8] Graph subsumption Recognizing Introduces a graph Accuracy: 65.90
and isomorphism Textual subsumption-based Precision: 64.70
Entailment approach for paraphrase Recall: 70.00
(RTE) identification. In this, input F1-score: 67.24
sentences are mapped to
corresponding graph
structures and paraphrasing
is evaluated based on graph
isomorphism
[9] Neural Networks Microsoft Lexical features in Accuracy: 75 and
(NN) Research association with syntactic, above
Paraphrase semantic, and composite
(MSRP) features are used to train a
back-propagation network,
wherein extracted features
are fed as input to the NN.
[10] Combination of deep MSRP; Detects paraphrasing based MSRP-
neural network with Sentences on lexical and semantic Accuracy: 80.2
keywords Involving similarity, measured by F1-score: 86.4
Compositional combining neural networks SICK-
Knowledge and keywords, and hence Accuracy: 87.1
(SICK) [16] identifies the relation
between the inputs in the
vector space.
[11] Textual Entailment RTE Paraphrase detection is Accuracy: 78.6
viewed as a bidirectional F1-score: 86.3
textual entailment task.
Word Sense
Disambiguation is
performed and using distant
supervised learning
technique, indirect facts are
mined from the predicates.
[12] Logistic Regression Quora Compares various machine Accuracies:
(LR); SVM; Neural Duplicate learning models after LR—59.9
Networks Question set preprocessing and encoding SVM—62.5
the input and concludes Siamese NN—63
recurrent neural network Two-layer
(RNN) to be the most NN—74
efficient algorithm for RNN—80
paraphrase identification
task.
(continued)
18 S. Chawla et al.

Table 1 (continued)
Research Technique/Algorithm Dataset Observations Performance
work
[13] Word Alignment Quora Question Introduces a novel Accuracy on
Information (WAI) Pairs (QQP) approach leveraging WAI various schemes:
to improve deep PI baseline Embedding—80.7
model RE2 [17]. Employs Multi-task—80.9
two major schemes to test Pre-train—82.1
the performance, proving Hybrid (Pre-train
‘pre-training’ the unlabeled +
in-domain data majorly Multi-task)—82.9
improving the performance
of baseline model.
[14] Curriculum Learning QQP; Estimates the effect of label Accuracy:
Large-scale noise on PI task and QQP—80.29
Chinese introduces an approach LCQMC—80.93
Question based on curriculum
Matching learning where: (a) a
Corpus loss-based noise metric is
(LCQMC) developed to compute noise
complexity of a sample,
and (b) similarity-based
noise metric classifies the
paraphrase.
[15] PIDG (Program 12 Introduces a novel Avg. Error Count
Interaction Undergraduate behavior-based Source at various %
Dependency Graph) programming Code Plagiarism Detection transformations:
assignments Tool, BPlag. Each 4.3
diversified with program’s behavior is
the help of represented with the help of
SPPlagiarise PIDG, and source code
[18] plagiarism is detected
based on the similarity
between the PID graphs.

3 Proposed Methodology

The goal of this paper is to measure and compare the efficiency of five word embed-
ding models (Word2Vec, Doc2Vec, BERT, TF-IDF, and FastText) in the task of
plagiarism/paraphrase detection, on two publicly available corpora: Quora Question
Pairs and Plagiarized Short Answers.
Appropriate preprocessing techniques are applied on both the corpora before
starting with model testing. The models are then evaluated based on the most appro-
priate threshold values and distance measure, and the results are produced based
on standard metrics—accuracy, precision, recall, and F1-score. For each model
tested against each corpus, ROC curve is generated for a deeper understanding of its
performance.
Comparative Analysis of Semantic Similarity Word … 19

3.1 Problem Definition

To formulate the task of paraphrase detection, the problem is defined as: Given
two text inputs T 1 and T 2 such that T 1 = {T 1 1 , T 1 2 , T 1 3 ,….,T 1 n } and T 2 = {T 2 1 ,
T 2 2 , T 2 3 ,….,T 2 n }. The task of paraphrase detection between the given inputs is
formalized as a binary task where L = {0,1} are the target labels such that if T 1 and
T 2 are duplicate L = 1 otherwise L = 0, i.e., if not duplicate.

3.2 Corpora

The paper aims at comparing the listed models based on their semantic similarity
detection. The following corpora, after taking into consideration their varied natures
are tested on the models:
Quora Question Pairs.
In 2017, Quora Question Pairs,1 an international competition by Quora, was released
to identify plagiarism and paraphrasing to be able to group similar/duplicate ques-
tions. The dataset contains genuine examples from the Web site with over 400,000
records.
Plagiarized Short Answers.
Plagiarized Short Answers2 (2009) is a corpus developed in a typical academic
setting for the task of plagiarism identification in which four levels of plagiarism
were committed. Since the input is classified based on a binary approach, four levels
of paraphrasing are encoded to only two, i.e., 0 (non-plagiarized) and 1 (plagiarized).
The dataset contained 95 documents plagiarized at different levels against 5 original
documents sourced from Wikipedia.
The two corpora differ a lot from each other in various aspects which is partially
the reason why they were chosen. (a) While PSA is a clean dataset containing only
textual information, QQP contains mathematical equations, abbreviations, slang,
typos, etc., all of which can be accounted for as noise. (b) PSA is a smaller dataset as
compared to QQP. (c) PSA also differs from QQP in the context that the constituent
documents at question are larger in length in PSA as compared to QQP.
The different natures of both the datasets helped in evaluating and reviewing the
word embedding models more precisely.
Before experimenting with the models, the datasets need to be transformed into
a standard form to maximize model performance. The following section discusses
the various text preprocessing steps applied to both the datasets to achieve the best
results.

1 https://www.kaggle.com/c/quora-question-pairs/.
2 https://ir.shef.ac.uk/cloughie/resources/plagiarism_corpus.html.
20 S. Chawla et al.

3.3 Text Preprocessing

Before feeding the input to the models, the text corpora need to be preprocessed to
eliminate noise and maximize the processing effect. The major preprocessing steps
are briefly discussed below and represented in the flowchart (Fig. 1) as well.
1. Each text input was transitioned from sensitive data to non-sensitive data by
translating it to lower case and tokenizing it.
2. Each token obtained was checked against a dictionary of stop-words obtained
from nltk library in python. If the word matched against the dictionary, it was
dropped to avoid unnecessary processing, as stop-words rarely have any role in
plagiarism.
3. Three alternatives of number manipulation were tested,

Fig. 1 Flowchart of text preprocessing steps applied to each corpus

Comparative Analysis of Semantic Similarity Word … 21

a. Leaving them
b. Replacing them with <NAN> and
c. Deleting them from the text.
Best results were achieved once they were removed altogether.
4. Stemming [19] was applied to original tokens to reduce them to their word stem
or word root, but the results proved to be better without this step, hence was not
carried forward.
5. Lemmatization [20] was applied by using WordNet Lemmatizer [21]. Lemma-
tization helps in retaining the context of text or achieving its base form. For
example, New and York are individual tokens which when lemmatized is treated
as a single entity, i.e., New York, and hence, the context of the text is retained.
The results proved to be slightly better when lemmatization was applied.
6. The tokens were then combined to form original text.
7. The original text was checked for punctuation marks and contractions. Punctu-
ation marks were straight away eliminated while contractions, like can’t, she’ll,
I’ve, etc., were expanded to their original form—here, cannot, she will, I have,
respectively.
8. The whole corpus is then searched for duplicate entries which are subsequently
removed to avoid overhead. The total number of duplicates were 353 which
accounted for only about 0.9% of the whole dataset and hence didn’t lead to
any unbalancing.
The above steps are applied to each dataset. No additional preprocessing is done
on any particular dataset to draw a fair result when tested with the models.
Following section discusses the various models which are applied on this
preprocessed data to test semantic similarity of the documents through word
embeddings.

3.4 Models Referred

Machine learning and almost all deep learning models are incapable of processing text
in its raw form. One of the first and most popular approach to measure the semantic
similarity between texts was vector space model [22], which introduced space density
computations on text such that each entity (such as characters in words, words in
sentences, sentences in documents and documents in dataset) could be represented
in an n-dimensional space as a vector. Hence, the proximity of two entities in the
space infers their semantic similarity.
Different models employ different strategies to utilize VSM for word embedding
of input. A brief description of the five models taken on for research is given below:
TF-IDF
The paper proceeds with the oldest word embedding model, i.e., Term Frequency-
Inverse Document Frequency which is a statistical frequency-based embedding
22 S. Chawla et al.

strategy developed by [23], and measures the significance of a term among the
corpus of documents. The words in a document are given different weights through
the following equation:
 
wi j = t f i j × log N /d f i (1)

where tf ij = number of occurrences of term i in document j, df i  = number of

documents containing the term i, and N = number of documents.
Word2Vec
Word2Vec is a deep learning model which is actually a combination of two strate-
gies—continuous bag of words (CBOW) and skip-grams. Both of these techniques
learn weights of each input word which are thus taken as the final word vector
representations.
The model creates vectors which are distributed numerical representations of
word features such as the context of individual words. With enough data, usage, and
contexts, Word2Vec can make accurate guesses about a word’s synonyms based on
its previous appearances (Fig. 2).
Figure 3 depicts the distribution of a small set of computed vectors in an n-
dimensional space.
BERT
Sent2Vec is a sentence to vector strategy where distributed representations of
sentences are achieved. Rather than extracting the semantic meaning of a word, the
whole sentence is taken into consideration. It can be thought of as an extension of
Word2Vec to sentences whereby, vectors are the average of source word embeddings.
BERTSimilarity library from pandas is used as sentence embedding technique in
this research which employs forward pass of BERT model to compute vectors for
sentences.

Fig. 2 Using Word2Vec to extract the most similar value of a given word based on nearest vector
values plotted in n-dimensional space
Comparative Analysis of Semantic Similarity Word … 23

Fig. 3 Distribution of vectors computed by Word2Vec model in an n-dimensional space

Doc2Vec
Using this strategy, the document can be represented as vectors by using paragraph
vector algorithm introduced by Mikolov et al. [2]. The model basically remembers
the context of words encountered, and hence, the whole document can be plotted as
a vector based on its semantic meaning.
FastText
FastText is a powerful word embedding model introduced by Facebook where in
a word is assumed to be formed by n-grams of characters. For example, rainy can
be represented as [rainy, rain, ainy], [rai, ain, iny]. It is particularly efficient over
traditional approaches in the sense that it accounts for rarer words and can give
vector representations to even the words absent from the dictionary.
All the above models have something in common: generating vector values for
textual data. Raw textual data is incapable of extensive experimentation as majority
of the existing algorithms work on numerical data. The advantage with extracted
vector values is that it allows putting the data through various concepts of linear
algebra, which opens doors to great level of experimentation and findings. The
following section discusses how the generated word embeddings can be used to
analyze semantic similarity between documents.
24 S. Chawla et al.

Table 2 Threshold values to

Quora question pairs Plagiarized short answers
classify plagiarized records in
each model corresponding to TF-IDF 0.52 0.54
the corpus Word2Vec 0.90 0.98
BERT 0.89 0.96
Doc2Vec 0.94 0.93
FastText 0.92 0.92

3.5 Distance Computation Between Word Embeddings

to Detect Plagiarism

After generating word embedding for the corpora based on various models mentioned
above, a distance measure was chosen to calculate the actual distance between vectors
in the n-dimensional space. This paper compares the vectors using the cosine distance
measure which can be obtained from cosine similarity. The cosine similarity between
two entities T1 and T2 can be calculated as:

similarity(T1 , T2 ) = cos(θ ) = (T1 · T2 )/(T1  × T2 ) (2)

The cosine distance can be calculated from similarity such that,

distance(T1 , T2 ) = 1 − similarity(T1 , T2 ) = 1 − cos(θ ) (3)

The distance between the two entities is checked against a threshold value. If the
score is greater than or equal to threshold, the entities are reported as plagiarized.
Since cosine distance is used as a metric, it is a given that the result of comparison
will always be in the range [0,1]. Thus, the threshold always lies in the range [0,1].
The threshold value chosen varies according to the model into consideration and
context of the given dataset. Hence, it is not a given that the same threshold value
will produce the most accurate result for a model over any corpus and should be
carefully chosen after experimentation.
The different threshold values which produced the most efficient results for chosen
datasets are presented in Table 2.

3.6 Evaluation and Results

Standard measures are used in this paper to calculate the performance of each model.
The measures include accuracy, precision, recall, and F1-score and can be calculated
with the help of sklearn library in python. The library gives the scores under two
major categories: macro-avg and weighted-avg. We have considered weighted-avg
in our evaluation to accommodate the size positive and negative samples.
Comparative Analysis of Semantic Similarity Word … 25

Table 3 Performance of models at their best thresholds

Accuracy Precision Recall F1-score
QQP PSA QQP PSA QQP PSA QQP PSA
TF-IDF 0.64 0.79 0.65 0.86 0.64 0.79 0.64 0.80
Word2Vec 0.64 0.73 0.63 0.74 0.64 0.74 0.64 0.73
BERT 0.66 0.77 0.66 0.83 0.66 0.77 0.66 0.77
Doc2Vec 0.65 0.74 0.66 0.72 0.65 0.79 0.65 0.753
FastText 0.69 0.93 0.70 0.92 0.69 0.92 0.69 0.92

Table 3 summarizes the values obtained of mentioned metrics against each model
and corpus.
Table 2 reveals that the threshold values for both the corpora against a partic-
ular model vary diversely. Table 3 gives the performance summary of all the tested
models. The behavior is dramatically different when the evaluation metrics for both
the corpora are compared against each other for all the models.
Both QQP and PSA datasets were made to undergo the same preprocessing steps
for a fair comparison but QQP is a much larger dataset compared to PSA. Hence, it is
lead to believe that the noise factor (like unaccounted abbreviations, orthographical
errors, etc.) plays a major role in the performance of a model for a particular corpus.
The derivation comes from the following two reasons:
• Deep learning models should ideally perform better when the size of dataset is
large to train their neural network. Therefore, QQP should have worked well with
all the models because of its size. But the performance of models, in contrast to
the expectation has greatly worsened in QQP dataset, accounting that noise has a
role to play.
• Since FastText is noise-tolerant to some extent as explained in Sect. 3.4, it should
be able to account for some faults in the dataset. The evaluation metrics prove
this point, as FastText performs significantly better on both corpora.
Another probable but indefinite reason for the poor performance of DL models on
QQP dataset can be the presence of label noise. Label noise is extensively noticed in
paraphrase identification datasets due to automatic labeling or non-expert labeling,
and it can severely impact the outcome of deep learning models as mentioned in [24].
QQP has undergone human labeling to classify a pair of questions as paraphrased or
non-paraphrased.
The presence of label noise, however, could not be confirmed in datasets due to
the shortage of time during implementation.
The ROC curves for various models on each dataset are depicted in Figs. 4 and 5
to help assess the performance of the models further. As mentioned in [25], a perfect
diagnostic test would constitute an ROC curve that is almost vertical from (0,0) to
(0,1) and then horizontal to (1,1).
It is evident that FastText exhibits the best performance among all the listed
models, with reference to Table 3 and Fig. 5.
26 S. Chawla et al.

Fig. 4 ROC Curves for various models on QQP dataset (left to right, top to bottom: Word2Vec,
TF-IDF, Doc2Vec, BERT, FastText)
Comparative Analysis of Semantic Similarity Word … 27

Fig. 5 ROC Curves for various models on PSA dataset (left to right, top to bottom: Word2Vec,
TF-IDF, Doc2Vec, BERT, FastText)
28 S. Chawla et al.

Even though BERT performs slightly better than Word2Vec, it has considerably
higher hardware and processing requirements, i.e., consumes more resources and
time.
Doc2Vec has better performance over BERT as well as Word2Vec, in terms of
both evaluation metrics and resource consumption. Contrary to expectations, TF-IDF
performs at par with Doc2Vec, Word2Vec, and BERT.

4 Conclusion and Future Scope

This paper proposes comparison of various corpus-based deep learning as well

as statistical word embedding models, specifically TF-IDF, Word2Vec, Doc2Vec,
BERT, and FastText. The models are tested on two publicly available corpora
namely—Quora Question Pairs and Plagiarized Short Answers. This research work
helps in understanding the performance of various word embedding techniques on
the mentioned corpora and effect of variations of various parameters like threshold,
hyperparameters, distance measures, preprocessing steps, on the evaluation metrics.
It is concluded that FastText produces the most accurate and efficient results, and
BERT is the most resource-demanding model. Based on the experimental results, it
can also be deduced that standalone deep learning models are not quite sufficient for
plagiarism detection and can be improved using additional mechanisms, for instance,
negative sampling, dissimilarity measures, label noise elimination, etc. The paper also
proves that TF-IDF has shown unexpected performance and hence probably holds
the capability to produce even better results, either in combination with another
well-suited DL model or with modifications to account for variations, like negative
sampling.
In the future, this research can be extended on other state-of-the-art models like
GloVe, ELMO, USE, etc. LSI can be considered as well for comparison of traditional
models with state-of-the-art models. Another gap in this research is lack of testing on
a corpus which is midway between PSA and QQP, i.e., a dataset that is significantly
larger as compared to PSA but has significantly lower noise when compared to QQP.
Another possible variation of the dataset can consider the length of constituent
documents to measure the trade-off between evaluation metrics and length. In addi-
tion to this, the results show great strength in TF-IDF model and the said can be
improved.

References

1. Mikolov T et al (2013) Distributed representations of words and phrases and their composi-
tionality. Adv Neural Inf Process Syst 1–9
2. Le Q, Mikolov T (2014) Distributed representations of sentences and documents. In: 31st Int
Conf Mach Learn ICML 2014, vol 4, pp 2931–2939
Comparative Analysis of Semantic Similarity Word … 29

3. Devlin J et al (2019) BERT: pre-training of deep bidirectional transformers for language

understanding. In: NAACL HLT 2019, vol 1, pp 4171–4186
4. Ramos J (2003) Using TF-IDF to determine word relevance in document queries. Proc First
Instr Conf Mach Learn 242(1):29–48
5. Joulin A, Grave E et al (2017) Bag of tricks for efficient text classification. In: 15th Conference
Eur Chapter Association Computer Linguist EACL 2017—Proceedings Conference, vol 2, pp
427–431
6. Salim AN et al (2012) Understanding plagiarism linguistic patterns, textual features, and
detection methods. IEEE Trans Syst Man Cybern 42(2):133–149
7. VMK and KK (2016) A survey on similarity measures in text mining. Mach Learn Appl Int J
3(1):19–28
8. Rus V, Mccarthy PM, Lintean MC, Mcnamara DS, Graesser AC (2008) Paraphrase identifica-
tion with lexico-syntactic graph subsumption, pp 201–206
9. Rajkumar (2010) Paraphrase recognition using neural network classification 1(29):43–48
10. Wang X, Xu B (2018) Paraphrase recognition via combination of neural classifier and keywords
11. Seethamol S, Manju K (2018) Paraphrase identification using textual entailment recognition.
ICICICT 2017, vol 2018-Janua, pp 1071–1074
12. Hunt E, Janamsetty R et al (2019) Machine learning models for paraphrase identification and its
applications on plagiarism detection. In: 2019 IEEE international conference big knowledge,
pp 97–104
13. Wang B, Wang L (2021) Enhancing deep paraphrase identification via leveraging word
alignment information. CNCERT/CC, pp 7843–7847
14. Li B, Liu T, Wang B et al (2020) Label noise robust curriculum for deep paraphrase
identification. In: 2020 international joint conference on neural networks (IJCNN), pp 1–8
15. Cheers H, Lin Y, Smith SP (2021) Academic source code plagiarism detection by measuring
program behavioral similarity. IEEE Access 9:50391–50412
16. Marelli M et al (2014) A SICK cure for the evaluation of compositional distributional semantic
models. In: Proceedings of 9th international conference language resource evaluation learning
2014, no May, pp 216–223
17. Yang R et al (2019) Simple and effective text matching with richer alignment features. In:
Proceedings of the 57th annual meeting of the ACL, pp 4699–4709
18. Cheers H, Lin Y, Smith SP (2019) Spplagiarise: a tool for generating simulated semantics-
preserving plagiarism of java source code. In: 2019 IEEE 10th ICSESS, pp 617–622
19. Porter MF (1980) An algorithm for suffix stripping. Program 14(3):130–137
20. Callison-burch C (2008) Syntactic constraints on paraphrases extracted from parallel corpora,
no October, pp 196–205
21. Fellbaum C (2010) WordNet. In: Theory and applications of ontology: computer applications.
Springer, Dordrecht, pp 231–243
22. Salton G, Wong A, Yang CS (1975) A vector space model for automatic indexing. Commun
ACM 18(11):613–620
23. Salton G, Buckley C (1988) Term-weighting approaches in automatic text retrieval. Inf Process
Manage 24(5):513–523
24. Li B, Liu T et al (2020) Label noise robust curriculum for deep paraphrase identification. In:
Proceedings of the International Joint Conference on Neural Networks (1)
25. NCSS Statistical Software (1997) Comparing two ROC curves—paired design, pp 1–14
Machine Learning Approach
for Identifying Survival of Bone Marrow
Transplant Patients

Varun Sapra, Luxmi Sapra, Yashi Bansal, Gunjan Chhabra,

and Rohit Tanwar

Abstract Bone marrow transplant is also known as hematopoietic stem cell trans-
plant which replaces unhealthy cells or marrow with healthy cells. This is useful for
curing many diseases especially few types of cancer. The success rate of bone marrow
transplant depends on many factors like genetic matching with donors and also on
some of the features of recipients. In this paper, we have implemented different
machine learning approaches like logistic regression, multilayer perceptron, random
forest and Bayes network to find out the survival of bone marrow transplant patients.
The same techniques have been applied on full dataset as well as on reduced dataset.
The dimensionality of the dataset is reduced using correlation feature subset method.
The results show that reduced dataset provides better accuracy with all the imple-
mented methods. Multilayer perceptron and Bayes network outshined other methods
with 94.1% accuracy.

Keywords Machine learning · Bone marrow transplant · Random forest ·

Hematopoietic cell transplantation · Multilayer perceptron

1 Introduction

Bone marrow is a spongy material inside the center of the bones, such as pelvis and
thigh bones; stem cells in it are responsible for producing platelets, white blood cells
(WBC) and red blood cells (RBC) [1]. Platelets prevent blood clotting, WBCs fight
against any infection and RBCs carry oxygen though out body. Bone marrow also
contains supporting fibrous tissues. It is of two types: red and yellow; one cannot
survive without a bone marrow because it is a factory that stores all the blood cells
and stem cells which are later transformed into RBC, WBC and platelets when
required [2, 3]. In some diseases, the bone marrow does not produce enough healthy

V. Sapra · G. Chhabra · R. Tanwar (B)

University of Petroleum and Energy Studies, Dehradun, India
L. Sapra · Y. Bansal
Dev Bhoomi Institute of Technology, Dehradun, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 31
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_3
32 V. Sapra et al.

blood cells, and in this case, patients can be aplastic anemiatic which can cause
leukemia, myelodysplastic syndrome, myeloproliferative disorders and so forth. In
these cases, transplantation is required. Some symptoms of bone marrow diseases
are tiredness, easy bleeding in injuries, extreme thirst, dehydration, loss of appetite,
frequent urination and many more [4]. It is also known as hematopoetic stem cell
transplantation. It is a medical procedure that is performed to replace unhealthy bone
marrow with the healthy one. The process does not involve any major surgery but
a process of blood transfusion. The procedure includes in which blood cells are
transplanted so that they can produce new blood cells and advances the development
of new marrow [5]. In some of the cases, chemotherapy or radiation is done at initial
stage to kill the diseased cells, and new or healthy cells are infused into the body
through an intravenous, i.e., IV catheter or tube. It is the same process as a patient
is given blood or medicine with the help of IV. There is no specific age for bone
marrow transplantation. The patient should fulfill certain criterion as approved by
medical practitioners. Yet young age donors are preferred for better and long-term
survival chances [6].
Machine learning is a field of computer science with the amalgamation of different
branches such as statistics, image processing feature engineering and artificial intel-
ligence [7–10]. It is the study of algorithm that learns automatically with examples
and generate predictive and descriptive models. Due to advancement in machine
learning methods and availability of huge amount of healthcare data in the form
of clinical reports, symptoms, imaging use of machine learning have tremendously
benefited the society. The paper is organized as follows. Section 2 contains the dataset
description. Section 3 discusses the proposed framework to predict the survival of
bone marrow transplant. Section 4 presents the results of the study. Finally, Sect. 5
presents the conclusion of the work.
Rosati et al. compared four different classifiers for the detection of hematopoi-
etically active BM. They have used CT images with 36 radiomic features for the
identification of hematopoietically active BM. For experiment purpose, they have
used a data of 25 patients which were affected with cell carcinoma of the anal canal.
Initially, the authors reduced the feature subset and parameters used for classification
by using genetic algorithms. The results obtained by applying the classifiers on the
CT images were compared with the literature. The performance of the classifiers was
found satisfactory for iliac bone marrow and lumbosacral bone marrow. It has also
been observed that the performance was poor in case of lower pelvis bone marrow
[11].
The success rate of hematopoietic cell transplantation especially in case of unre-
lated donors depends on strict genetic matching. To improve the choice of donor,
Buturovic et al. proposed a multivariate classifier model that can assign a label of
preferred or not preferred based on the survival rate of the recipients. They evaluate
the survival likelihood to five years based on clinical parameters of the donor. For
experimental purpose, the data has been collected for 1255 patients who underwent
unrelated donor HCT from Center for International Blood and Marrow Transplant
Research Center. The data was collected for a period of 2000–2010. Their attempt
to predict HCT outcomes was not successful [12].
Machine Learning Approach for Identifying Survival … 33

Sharma et al. present a framework for improving the performance of nuclear

segmentation technique for histopathological datasets using an AdaBoost classifier.
For experimental purpose, they have used a dataset of slides with varying degrees of
H&E stains. The results show that the research carried out by the authors is valuable
for extracting visual data from heterogeneous datasets [13].
Kimura et al. in their research work proposed a convolutional neural network
support system for identification of blood cell image. They implemented a deep
learning system with gradient boosting for better decision making. They applied their
model on the dataset comprised of 695,030 images. The proposed model was able
to achieve the sensitivity of 93.5% and specificity of 96.0% for leucocyte detection,
and for aplstic anemia, the model achieved the accuracy of 96.2% [14].

2 Dataset Description

Bone marrow transplant dataset is explored for the experimental purpose. The bench-
mark dataset is collected from UCI machine repository [15] having 37 features and
187 instances. The dataset consists of features such as donor age, ABO, age of
recipient, gender of recipient, body mass of recipient, types of disease for which
transplantation required, disease group, stem cell source, serological compatibility,
compatibility of antigens, relapse of the disease, neutrophils recovery time, etc. Table
1 shows the description of the bone marrow transplant dataset.

3 Proposed Framework

In this paper, we proposed machine learning framework in order to predict the survival
chances of bone marrow transplant in children’s. The framework consists of rele-
vant data selection, data preprocessing and feature engineering to reduce the dimen-
sionality of the dataset in order to choose only those features that contribute more
to outcome. Dimensionality reduction algorithm speedup the training and testing
time of the algorithm by eliminating the redundant and irrelevant feature from the
data [16, 17]. Initially, the dataset contained 37 features, and after implementing
feature subset selection method, the dimensionality of the dataset has been reduced
to six features. The most influential features are selected by correlation-based feature
subset algorithms which are disease, PLT recovery, extensive chronic GvHD, relapse
and survival time. Machine learning models are trained and tested using tenfold cross-
validation method. The algorithms used to train and test models are logistic regres-
sion, multilayer perceptron, random forest and Bayes network. The same methods
have been applied on full dataset as well as reduced dataset to evaluate the perfor-
mance of these methods. Figure 1 shows the proposed framework for bone marrow
transplantation survival prediction.
34 V. Sapra et al.

Table 1 Bone marrow transplant dataset description

Features Description Min Max Mean stddev
Donor age Numeric 18.6 55.55 33.47 8.272
Done age < 35 Yes/no Yes-104, No-83
Donor ABO A, B, AB, O A-71, B-28, AB-15, 0-73
CMV_Donor Preset/absent Present-72,Absent-113
Age recipient numeric 0.6 20.2 9.9 5.3
Age < 10 recipient Yes/no Yes-99,
no-88
Recipient age Numeric 5 to 10–51, 0 to 5–47,
(count) 10 to 20–89
Gender recipient Male/female Male-112, Female-75
Body mass Numeric 6 103.4 35.8 19.6
(recipient)
ABO (Recipient) Categorical A-75, B-50, AB-13,
(A, B, AB, O) O-48
Rh (Recipient) Categorical Plus -158,
(plus, minus) Minus -27
CMV (Recipient) Categorical Present-100,
(present, Absebt-73
absent)
Disease Categorical ALL-68, AML-33,
(ALL, AML, Chronic-45,
chronic, Nonmalignant-32,
nonmalignant, Lymphoma-9)
lymphoma)
Disease group Categorical Malignant-155,
(malignant, Nonmalignant-32
nonmalignant)
Gender match Categorical Others-155, female to
Others, female male-32
to male
ABO_Match Categorical Matched-52,
matched, Mismatched-134
mismatched
CMV Status Numeric 0 3 1.509 1.12 9
HLA Mismatch Categorical Matched-159,
(matched, Mismatched-28
mismatched)
Antigen Numeric 0 3 0.92 1
Allel Numeric 0 4 0.75 0.89
HLA Group-1 Categorical
(continued)
Machine Learning Approach for Identifying Survival … 35

Table 1 (continued)
Features Description Min Max Mean stddev
Risk group Categorical High-69, low-118
High, low
Stem Cell Source Categorical Peripheral blood-145
peripheral Bone marrow-42
blood/bone
marrow
Tx post relapse Binary Yes-164, no-23
(Yes/no)
CD34_x1e6 per kg Numeric 0.79 57.7 11.8 9.9
CD3_x1e8per kg Numeric 0.04 20.02 4.7 3.8
CD3_to_CD34_ratio Numeric 0.204 99.5 5.3 9.59
ANC_Recovery Numeric 9 1,000,000 26,752.8 161,747.2
PLT_Recovery Numeric 9 1,000,000 90,937.9 288,242.4
acute_GvHD_II_III_IV Binary Yes-112, No-75
(Yes/no)
acute_GvHD_III_IV Binary Yes-40, no-147
(Yes/no)
time_to_acute_GvHD_III_IV Numeric 10 1,000,000 775,408.04 418,425.0
extensive_chronic_GvHD Binary Yes- 128, no-28
(Yes/no)
Relapse Binary Yes-159,no-28
(Yes/no)
survival_time Numeric 6 3364 938.7 849.5
Survival status Binary Yes/no Yes-102, no-85

4 Results

The performance of the machine learning algorithms such as logistic regression,

multilayer perceptron, random forest and Bayes network is evaluated on the base
of performance measures such as accuracy, error rate, kappa stats, mean absolute
error, room mean square error, sensitivity, specificity, precision, F-measure and ROC.
Tables 2 and 3 display the performance of machine learning algorithm by considering
all the features.
In the next step, we applied feature engineering process where dimensionality of
the dataset is reduced by using CFS-based feature selection method. Dimensionality
of the feature space is reduced to five features. Tables 4 and 5 show the result of
reduced dataset.
36 V. Sapra et al.

Fig. 1 Machine learning framework for bone marrow transplantation

Table 2 Accuracy/error/kappa/MAE and RMSE of models using all the features

Model Accuracy (%) Error rate (%) Kappa MAE RMSE
Logistic regression 86.631 13.3 0.7301 0.134 0.3621
Multilayer perception 89.8396 10.1604 0.7953 0.1039 0.277
Random forest 93.5829 6.4171 0.8696 0.2495 0.2884
Bayes network 90.3743 9.6257 0.8051 0.0987 0.2354

Table 3 TPR/FPR/precision/recall/F-measure and ROC of models using all the features

Model Class TPR FPR Precision Recall F-measure ROC
Logistic regression Yes 0.882 0.153 0.86 0.882 0.73 0.936
No 0.847 0.118 0.857 0.847 0.73 0.935
Multilayer Perception Yes 0.902 0.106 0.911 0.902 0.906 0.968
No 0.894 0.098 0.884 0.894 0.889 0.968
Random Forest Yes 0.980 0.118 0.909 0.980 0.943 0.973
No 0.882 0.020 0.974 0.882 0.926 0.973
Bayes network Yes 0.931 0.129 0.896 0.931 0.806 0.977
No 0.871 0.069 0.914 0.871 0.806 0.977
Machine Learning Approach for Identifying Survival … 37

Table 4 Accuracy/error/kappa/MAE and RMSE of models after feature engineering

Model Accuracy (%) Error rate (%) Kappa MAE RMSE
Logistic regression 93.04 6.9 0.8602 0.0885 0.2385
Multilayer perception 94.1 5.8 0.8815 0.0805 0.2437
Random forest 93.04 6.9 0.8594 0.084 0.2237
Bayes network 94.1 5.8 0.881 0.0869 0.2115

Table 5 TPR/FPR/precision/recall/F-measure and ROC of models after feature engineering

Model Class TPR FPR Precision Recall F-measure ROC
Logistic regression Yes 0.922 0.059 0.949 0.922 0.935 0.973
No 0.941 0.078 0.909 0.941 0.925 0.973
Multilayer perception Yes 0.941 0.059 0.950 0.941 0.946 0.957
No 0.941 0.059 0.930 0.941 0.936 0.957
Random forest Yes 0.951 0.094 0.924 0.951 0.937 0.986
No 0.906 0.049 0.939 0.906 0.922 0.986
Bayes network Yes 0.961 0.082 0.933 0.961 0.947 0.979
No 0.918 0.039 0.951 0.918 0.934 0.979

5 Conclusion and Discussion

In this paper, we implemented machine learning framework to predict the survival

of bone marrow transplant patients. Figures 2, 3, 4, 5 and 6 show the performance
of different machine learning algorithm using parameters such as accuracy, error
rate, true positive rate, false positive rate, F-measures and ROC. In case of full
dataset, logistic regression achieves the accuracy of 86.6%, multilayer perceptron
achieves the accuracy of 89.8%, random forest achieves the accuracy of 93.5% and
in case of Bayes network, accuracy of 90.3% is reported. By reducing the dataset
by using feature reduction algorithm, logistic regression achieves the accuracy of
93%, multilayer perceptron achieves the accuracy of 94.1%, random forest achieves

100
80
60
40 Accuracy (%)

20 Error rate (%)

0
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 2 Accuracy and error rate of models using all the features
38 V. Sapra et al.

Chart Title
1

0.5 Kappa
MAE
0 RMSE
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 3 Value of kappa/mean absolute and root mean square by considering all the features

1.2
1
TPR
0.8
0.6 FPR

0.4 Precision
0.2 Recall
0 F-Measure
yes no yes no yes no yes no ROC
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 4 Value of TPR/FPR/precision/recall/F-measure/ROC by considering all the features

100
80
60
40 Accuracy (%)

20 Error rate (%)

0
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 5 Accuracy and error rate of models using feature engineering

1
0.8
0.6 Kappa
0.4
MAE
0.2
0 RMSE
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 6 Value of kappa/mean absolute and root mean square by using feature engineering
Machine Learning Approach for Identifying Survival … 39

1.2
1 TPR
0.8
0.6 FPR
0.4 Precision
0.2
0 Recall
yes no yes no yes no yes no
F-Measure
Logisc Mul Layer Random Bayes
ROC
Regression Percepon Forest Network

Fig. 7 Value of TPR/FPR/precision/recall/F-measure/ROC with feature engineering

96
94
92
90
88 Accuracy (AL)
86 Acuracy (FE)
84
82
Logisc Mul Layer Random Bayes
Regression Percepon Forest Network

Fig. 8 Comparison of accuracy using complete features (AL) and feature engineering (FE)

the accuracy of 93% and Bayes network achieves the accuracy of 94.1%. The bone
marrow transplant dataset consists of 37 features. We applied feature engineering
for dimensionality reduction, and the improvement of accuracy has been reported
in most of the cases. In case of logistic regression, the 6.44% of improvement is
reported; in case of MLP, 4.27% improvement of accuracy has been reported as well
as in case of Bayes network, 3.73% of improvement of accuracy is reported. But
in case of random forest, 0.54% decrement of accuracy is reported. For this type
of data, the performance of multilayer perceptron and Bayes network outperformed
with accuracy of 94.1%. The experiment demonstrates that in most of the cases,
improvement of accuracy is reported but in some case, accuracy can be decreased
by using feature engineering (Figs. 7 and 8).

References

1. Henze K, Herten M, Haversath M, Busch A, Brandau S, Hackel A, Jäger M (2019) Surgical

vacuum filter-derived stromal cells are superior in proliferation to human bone marrow aspirate.
Stem Cell Res Ther 10(1):1–14
2. Larsson A, Smekal D, Lipcsey M (2019) Rapid testing of red blood cells, white blood cells and
platelets in intensive care patients using the HemoScreen™ point-of-care analyzer. Platelets
30(8):1013–1016
40 V. Sapra et al.

3. Kawabata H, Isobe K, Watanabe T, Okudera T, Nakamura M, Suzuki M, Kawase T (2017)

Quality assessment of platelet-rich fibrin-like matrix prepared from whole blood samples after
extended storage. Biomedicines 5(3):57
4. Pranav Dheenathayalan S (2019) A descriptive study of Naadi Thervu and its clinical features
based on the text “Sadhaga Naadi” (Doctoral dissertation, Government Siddha Medical College,
Palayamkottai)
5. King W, Toler K, Woodell-May J (2018) Role of white blood cells in blood and bone marrow-
based autologous therapies. BioMed Res Int
6. Kasamon YL, Bolaños-Meade J, Prince GT, Tsai HL, McCurdy SR, Kanakry JA, Jones
RJ (2015) Outcomes of nonmyeloablative HLA-haploidentical blood or marrow transplan-
tation with high-dose post-transplantation cyclophosphamide in older adults. J Clin Oncol
33(28):3152
7. Chandradevan R, Aljudi AA, Drumheller BR, Kunananthaseelan N, Amgad M, Gutman
DA, Jaye DL (2020) Machine-based detection and classification for bone marrow aspi-
rate differential counts: initial development focusing on nonneoplastic cells. Lab Invest
100(1):98–109
8. Ratley A, Minj J, Patre P (2020) Leukemia disease detection and classification using machine
learning approaches: a review. In: 2020 first international conference on power, control and
computing technologies (ICPC2T). IEEE, pp 161–165
9. Radakovich N, Nagy M, Nazha A (2020) Machine learning in haematologi cal malignancies.
Lancet Haematol 7(7):e541–e550
10. ALzubi JA, Bharathikannan B, Tanwar S, Manikandan R, Khanna A, Thaventhiran C (2019)
Boosted neural network ensemble classification for lung cancer disease diagnosis. Appl Soft
Comput 80:579–591
11. Rosati S, Franco P, Fiandra C, Arcadipane F, Silvetti P, Gallio E, Balestra G (2020) Compar-
ison of different classifiers to recognize active bone marrow from CT images. In: 2020 IEEE
international symposium on Medical Measurements and Applications (MeMeA). IEEE, pp 1–5
12. Buturovic L, Shelton J, Spellman SR, Wang T, Friedman L, Loftus D, Lee SJ (2018) Evaluation
of a machine learning-based prognostic model for unrelated hematopoietic cell transplantation
donor selection. Biol Blood Marrow Transplant 24(6):1299–1306
13. Sharma H, Zerbe N, Heim D, Wienert S, Behrens HM, Hellwich O, Hufnagl P (2015) A
multi-resolution approach for combining visual information using nuclei segmentation and
classification in histopathological images. VISAPP (3):37–46
14. Kimura K, Tabe Y, Ai T, Takehara I, Fukuda H, Takahashi H, Ohsaka A (2019) A novel
automated image analysis system using deep convolutional neural networks can assist to
differentiate MDS and AA. Sci Rep 9(1):1–9
15. Dua D, Graff C (2017) UCI machine learning repository [Internet]. Available from: http://arc
hive.ics.uci.edu/ml
16. El-Hasnony IM, Barakat SI, Elhoseny M, Mostafa RR (2020) Improved feature selection model
for big data analytics. IEEE Access 8:66989–67004
17. ur Rehman MH, Liew CS, Abbas A, Jayaraman PP, Wah TY, Khan SU (2016) Big data reduction
methods: a survey. Data Sci Eng 1(4):265–284
Human Disease Prognosis and Diagnosis
Using Machine Learning

Sunil Kumar, Harish Kumar, Rashi Agarwal, and V. K. Pathak

Abstract The medical or healthcare sector is one of the trending and challenging
fields of machine learning (ML). Machine learning is a branch of artificial intelli-
gence (AI) that is concerned with machine training and learning. “It is the defining
technology of this decade, though its impact on healthcare has been meagre”—stated
James Collin of MIT in the USA. Healthcare-related machine learning encompasses a
broad range of techniques that can be used to assess efficiency, scope, versatility, and
ability in various machine learning applications. Many multinational corporations
and research organizations do more research and development work in the medical
sector, which can be associated with machine learning to find some fruitful solutions
for human beings, as humans face a variety of health-related issues. Disease prog-
nosis and diagnosis are new challenges of the medical or healthcare sector, in which
computer technology can utilize. Disease prognosis and diagnosis are critical tasks
in the medical or healthcare, in which the prediction and accuracy of the results must
be up to the mark. The significance of machine learning in medical or healthcare
systems has been discussed in this paper for further exploration.

Keywords Machine learning · Artificial intelligence · Healthcare · Disease

prognosis · Disease diagnosis

1 Introduction

Evolution is a general form which occurs in most of the living and nonliving thing.
Nature evolves living things, while nonliving or artificial things, such as technology,

S. Kumar (B) · H. Kumar

J C Bose University of Science and Technology, YMCA, Faridabad, Haryana, India
S. Kumar · R. Agarwal
University Institute of Engineering and Technology, CSJM University, Kanpur, UP, India
V. K. Pathak
CSJM University, Kanpur, UP, India
e-mail: vinay@vpathak.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 41
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_4
42 S. Kumar et al.

evolve as well. First of all, we have to understand how evolution takes place in
any terminology like biological or artificial. As far as we know, humans evolve
biologically, learning from their experiences and failures. But with artificial things
like computers, which follow instructions, they can provide some form of solution
or understanding. But, can we say machines learn from past experiences as humans
do? The answer is yes, the experience of the machine referred to as fed data into it;
this concludes that the machines can learn from past data rather than experience as
humans do. On this basis, we can say that the machines have started learning from
past data results machine learning comes into the picture. Interdisciplinary fields are
associated with each other to solve real-life problems. Machine learning is capable of
solving many computational problems that can be used for solving medical-related
issues in humans.
The prediction of the disease’s severity is an important issue in the healthcare
system, which may hurt the patient in future. The disease prognosis does this
prediction through the observations, symptom identification, and signs related to
the disease. With time limits, the disease prognosis keeps track of the disease and
checks the impacts of the disease on patients, like its health is stable, good, bad, or for
any complications. Using machine learning techniques, we can provide this function-
ality, which offers prediction, classification, and many more things by which we can
predict or classify different types of diseases like diabetes, heart attack, cancer, etc.,
according to their characteristics. Medical diagnosis is usually made based on the
doctor’s expertise and experience. Based on observations, the doctor draws certain
conclusions [1].

2 Machine Learning

Machine learning aims to improve the performance of an algorithm for a particular

task based on experience. In machine learning, experience can be considered as
training data. We take some sample data for observation from this training data.
Machine learning algorithms may develop mathematical or computational models
for making any decisions, forecasts, or classifications based on experience or training
data [2].
In machine learning, data plays an important role because data is the basic building
block of information as well as knowledge. If we have more data, then we can infer
more knowledge by which machine can learn more. But as the data size increases,
a large number of features also increases, which necessitates a large number of
samples to be accurate. For providing accuracy, the machine must be capable of
analysing, predicting, and making decisions on the past data. This functionality
provides machine learning, in which the system learns automatically from the given
past data.
Generally, machine learning is used for classification, prediction or making deci-
sions, etc. It can achieve based on some patterns and relationships within the given
Human Disease Prognosis and Diagnosis Using Machine Learning 43

datasets [3]. A concept of machine learning arises, which says from the given dataset,
we can develop an automatic inference of patterns or relationships [4].

2.1 Machine Learning Viewpoint

Machine learning can be seen as a different-different point of views:

ML as learner’s viewpoint—According to Tom Mitchell, machine learning is
[5]—“… said to learn from experience with respect to some class of tasks, and
a performance measure P, If (the learner’s) performance at tasks in the class, as
measured by P, improves with experience”.
Humans or non-machines like biological agents or learners learn from past experi-
ences, in which the learner repeats a set of tasks in order to assess their performance.
As a result, it can be check that what the learner’s performance is.
ML as Artificial Intelligence’s viewpoint—Generally, it is an assumption that
machine learning (ML) is a sub-domain of artificial intelligence (AI). This assump-
tion can be true or false, both from the viewpoint of intelligence. All of the ML is not
always performing intelligent behaviour, which is an aspect of artificial intelligence.
Anyone using ML can think he is using AI; it is not always correct because not all of
ML is AI, only some part of AI which belongs to ML and AI performing intelligent
behaviour [6].
ML as a new programming paradigm’s viewpoint—To understand the
programming paradigm of machine learning, we need to understand traditional
programming, which helps us in understanding the paradigm shift of programming.
In traditional programming, we have the input data and a coded program, which will
be run on a computer to produce the desired output.
In machine learning, we have the input and output data fed to the computer
system and generate a program. In this scenario, we can check the performance of an
algorithm. In machine learning, the general algorithms formulate the rules from the
given input data as the programming paradigm shifts from traditional programming
to machine learning that causes to deal with large datasets.

2.2 Machine Learning Approaches

Machine learning has different-different approaches to solve the different-different

type of problems. Using one of the machine learning approaches, most of the prob-
lems can be solved. The division of approaches is important because it aids in under-
standing the essential nature of a problem that you can later work on. This will help
you to provide a solution. Machine learning approaches have four broad categories
as given below [2].
Supervised Learning—This type of learning is appropriate when the dataset
contains labelled data; the labelled data have a tag or label and have some context or
44 S. Kumar et al.

meaning based on some observation [2]. Supervised learning works on a mathemat-

ical or computational model in which a dataset contains both input and the desired
output; it means you know the values of input and output [6, 7]. Supervised learning
consists of classification and regression.
Unsupervised Learning—This type of learning is appropriate when the dataset
contains no labelled data; data has no label given to the learning algorithm. This
type of learning approach searched for undetected patterns and relationships in the
given dataset. This approach generally draws inferences from the given dataset [2].
A medical or healthcare example of unsupervised learning methods is the prediction
of heart diseases using clustering [8].
Semi-Supervised Learning—This type of learning is appropriate when the
dataset contains both labelled and unlabelled data. This approach incorporates the
benefits of both supervised and unsupervised learning methods so it has the potential
to produce fruitful results in real-life applications. The real-life applications of semi-
supervised learning are human speech analysis, content classification of the Internet,
protein sequence classification and so on. In this approach, processing starts by clus-
tering the data, which is unlabelled and dealt with by unsupervised learning. Then,
next, for labelling the clustered unlabelled data, we use labelled data [2].
Reinforcement learning—This type of learning is appropriate in a dynamic envi-
ronment (real or virtual) in which the algorithm has an objective to gain a particular
goal. The algorithm or software agent acts as an action, and this action causes a
reward or penalty. The agent’s main objective is to learn how to perform perfect
actions to gain maximum rewards [2].

3 Machine Learning in the Healthcare System

In the modern age, interdisciplinary domains are altogether to solve real-life related
problems. ML and healthcare domains are also cooperating to solve real-life prob-
lems. The healthcare industry also uses ML to provide a better solution. The inter-
section of ML and healthcare is a new trending domain in which new possibilities
of finding new solutions to real-life problems. As the demand for AI is increasing,
the adoption of AI will be increasing soon. Over 50 executives from healthcare orga-
nizations that use or fund AI were surveyed by emerj [9]. This survey represented
an industry’s viewpoint for the adoption of AI in the USA. Emerj separated the
companies that applying AI from those that do not. Companies consider return on
investment (ROI) as a major factor that plays an important role in the adoption of AI.
In the survey, the majority of surveyed companies are focusing on improving disease
diagnostics and medical image analysis. As the field develops, disease diagnostic
is a new trend to explore [10]. AI or ML is associated with the healthcare systems
in which the disease prognosis and diagnosis is an important step to provide better
solutions to patient-centric systems. To use AI in the healthcare systems is not so
easy because of the complexities of both the systems [10].
Human Disease Prognosis and Diagnosis Using Machine Learning 45

Machine learning in the domain of medical or healthcare is an evolution of

machine learning in which algorithms and models interpret data (like medical records,
electronic health record, medical images, etc.), which can be used to identify and
improve the disease prognosis and diagnosis of humans. Numerous machine learning
applications will be used in the medical or healthcare systems [1], such as disease
identification, medical imaging diagnosis, smart health records, proactive health
management, drug discovery, clinical research and trial, customer complaint reso-
lution and many more. In this process, we have to develop new ML methodologies
or techniques that facilitate clinical decision-making and predictions. In clinical
diagnoses, ML methods provide valuable insights [11, 12].

3.1 Sources of Healthcare Data

As we can see from the preceding section, no data sources are involved in supporting
medical or healthcare systems. As a result, we must concentrate on medical or health-
care data sources in order to comprehend how data plays a significant role in the
medical or healthcare system. Technology has progressed from the computer to the
Internet. They work with a fundamental unit known as data. As data is a new fuel to the
industry, the healthcare system is no exception. Due to the increasing volume of clin-
ical data, operation performed on the clinical data is an important process. Patients’
documents are now transformed to digital form in both the private and public health-
care systems. That digital data is generated, modified, and maintained by healthcare
professionals. The healthcare data may have data from many sources like research
data, wearable devices data, laboratory result data, medical images, smart phone
data, environmental data, open access data and electronic health records, etc.

4 Disease Prognosis Using Machine Learning

The prediction of patients’ health conditions is a critical task in clinical or medical

decision-making. Machine learning is used not only for predicting the outcome or
severity of diseases but also for post-operative events. According to neuraldesigner (a
machine learning website/blog), the definition of the medical prognosis is as follows
[8]:
Medical prognosis consists of predicting the evolution of a patient so that the doctor can
select the most appropriate treatment for him or her.

Medical prognosis may have one or more criteria like patients’ life mortality,
survivability of the patient, progression of disease, treatment, sensitivity, etc. These
criteria are not the only factors that can help with prognosis; there are a number
of other features or variables that can also help. A few of the most significant are
mentioned below:
46 S. Kumar et al.

• Personal data of the patient like name, the age, sex, height, weight, etc.
• Data related to disease like symptoms of the patient, observations and test lab,
etc.
• Environmental data like the patient is drinking or smoking, the location from
which the patient belongs, etc.
• Additional information or condition which may help in disease prognosis.
• The benefits of disease prognosis using machine learning are that it provides:
• By the adjusting treatment level, we can improve the treatment structure.
• It prevents excessive treatment according to monitoring.
• It can prioritize patients according to their disease prognosis.
By using machine learning, predictive models also can be made. It helps doctors
or healthcare professionals to find out the outcome of the existing disease of the
patient. According to the outcome of the disease, appropriate treatment can be given
to the patients [11].

5 Disease Diagnosis Using Machine Learning

Machine learning-assisted disease diagnosis is a great challenge that will be accom-

plished efficiently and more accurately. Machine learning is becoming increasingly
popular for disease diagnosis (Fig. 1).
The medical or healthcare industry deals with a lot of data, including patient
details, medical records, history, and so on, which makes it difficult to evaluate and
maintain. To solve this issue, machine learning systems search for patterns in data
and predict outcomes. The healthcare practitioners will draw certain conclusions
based on the outcome, which helps in the patient’s personalized care. As a result,
we can see how machine learning systems can effectively manage huge amounts of
medical data.
There are numerous machine learning techniques available, each with its own set
of algorithms for data classification and prediction. As a result, we can say that it
contributes to more accurate and efficient decision-making [1].

Fig. 1 Trend of machine 2500

learning in disease diagnosis. 2000
The comparison is obtained
1500
through searching disease
diagnosis using machine 1000
learning on PubMed 500
0
1980 1990 2000 2010 2020 2030
Human Disease Prognosis and Diagnosis Using Machine Learning 47

6 Methodology

The methodology of the proposed system is to introduce a “human disease prognosis

and diagnosis using machine learning”. In general, machine learning begins with data
acquisition and continues with pre-processing by the system. Find specific diseases
and their relevant features in the form of numeric data that the machine learning
algorithm can process.
The proposed approach for the prognosis and diagnosis of human diseases using
machine learning (Fig. 2) in which some specific steps are presented are as follows:

Fig. 2 Proposed approach

for disease prognosis and Data Collection
diagnosis using machine
learning
Data preprocessing
of Datasets

Feature Extraction

Feature Selection

Testing Dataset Training Dataset

Machine Learning
Algorithm

Performance Evaluation
of Machine Learning Model

Predicted Output
48 S. Kumar et al.

6.1 Datasets Collection

The first step after identifying the specific disease is to collect a dataset. Datasets
can be obtained from the machine learning databases such as UCI, Kaggle, private
datasets, and so on.

6.2 Data Pre-processing of Datasets

In this step, we process the data obtained in the first stage, which may or may not be
perfect. This allows the data to be transformed into something useful and functional
[1]. For example, in an X-ray image, the pre-processing of the dataset can be done
as follows:
• On X-ray images, image resizing, image cropping, and dimension reduction are
a few of the pre-processing tasks used [13].
• If you are working on an image classification problem, you will need to identify
some issues before you can classify the images in the original dataset. Initially,
image resizing to a certain dimension as the source, because the machine learning
model demands that the size of the data being trained should be the same [13].
• The X-ray images can be in both two-dimensional images and in three-
dimensional images. Dimension reduction was needed in order to convert
three-dimensional images into two-dimensional images [13].

6.3 Feature Extraction

After selecting the region of interest (ROI), a mathematical calculation takes place for
computation, which gives information about the ROI called a feature. For obtaining
information from the ROI like the texture, a special kind of processing is performed
called feature extraction [14]. In this step, features extracted from the image using
machine learning techniques like grey level co-occurrence matrix (GLCM), LBP,
DWT, etc. GLCM is a methodology of mathematics for the analysis of texture that
provides the spatial ratio of pixels. Due to the large number of variables used in
computation, GLCM reduces the variables using only needed variables [15]. By
using feature extraction methods, all extracted features are passed to the next step.

6.4 Feature Selection

From extracted features, the only relevant features will choose for the study. Machine
learning simply works on the rule which says, if you input surplus data, you will get
Human Disease Prognosis and Diagnosis Using Machine Learning 49

surplus data. In machine learning, generally, algorithm prefers to feed only those
features that are important. Some important reasons to use feature selection are like
if the right subset of features is selected, it allows the machine learning algorithm to
train faster, reduces model complexity, and increases model accuracy [16]. Wrappers,
filters, and embedded methods are three major types of feature selection algorithms.

6.5 Division of the Datasets

Generally, in any machine learning models, the datasets are divided into two portions.
The first portion is used for training the model, while the second portion is used to
test the model, which may contain some biases in the test dataset, which affects the
accuracy. To avoid this problem, cross-validation is used to estimate a model’s ability
on an unknown dataset. In k-fold cross-validation, the dataset is divided into k, equal-
sized parts, such as 5, 10, and so on. Shuffle dataset with k-fold cross-validation to
ensure that the order of input and output is absolutely random [17].

6.6 Selection of Machine Learning Algorithms

In this step, identification or classification of disease is done, which tells that the
human has any ailment or not. Machine learning algorithms can perform this task. It
is also difficult to choose the right machine learning algorithm, such as SVM, random
forest, MLP, CNN, and so on. When we use machine learning algorithms, we can
achieve the following two cases:
1. If the human does not contain any disease, the machine learning algorithm
predicted that the human does not contain any disease, correctly.
2. If the human contains any disease, the machine learning algorithm predicted the
disease by which the human is affected.

6.7 Performance Evaluation of a Machine Learning Model

A confusion matrix, also known as an error matrix, is typically represented as a

table in machine learning (Fig. 3). A confusion matrix is used for visualization for
the performance of a model. Each column in the matrix represents instances in a
predicted class, but each row represents instances in an actual class (or vice versa)
[18].
The following are the fundamental concepts used in the confusion matrix [13]:
True Positive (TP)—The positive class is correctly labelled. For instance, if a
person has a disease, the model correctly classifies the disease existence.
50 S. Kumar et al.

Fig. 3 Confusion matrix [19]

True Negative (TN)—The negative class is correctly labelled. For instance, if a

person does not have a disease, the model correctly classifies the person as having
no disease.
False Positive (FP)—Positives are predicted incorrectly. For instance, the person
does not have any disease, but the system labels the person as having the disease.
False Negative (FN)—Negatives are predicted incorrectly. For instance, the
person has a disease, but the model says the person does not have any disease [13].
For measuring the performance of a machine learning model, we can choose
any performance metric from the performance metrics category. Some of the
performance.
metrics discussed below are accuracy, precision, sensitivity or recall, specificity,
and the F1 score:
Accuracy—This is a technique for determining the correct prediction value. To
determine the accuracy, TP and TN are divided by the total amount of test data. This
provides the correct classification results.

Accuracy = (TP + TN)/(TP + FP + FN + TN) (1)

Precision—Precision is evaluated by dividing the correct predictive value (TP)

by the actual TP and FP values. This is used to calculate the accurate prediction
outcomes of all positive class predictions. Precision can answer a question like how
many of those who we labelled have disease from the total of those who had.

Precision = TP/(TP + FP) (2)

Human Disease Prognosis and Diagnosis Using Machine Learning 51

Recall—Sensitivity is another name for recall. It is used to measure true positive

predictions (TP) on all true predictions. The proportion of correctly positive-labelled
disease to all disease cases in actual. Recall answers: Among all of the people who
have the illness, people who have the disease have been correctly predicted [20].

Sensitivity = TP/(TP + FN) (3)

Specificity—It is a metric for determining how many negatives are defined

correctly. The test’s specificity refers to its ability to correctly reject healthy people
who do not have a disease’s diagnosis, which means it indicates a healthy individual
[21].

Specificity = TN/(TN + FP) (4)

F1-Score—Precision and recall’s harmonic mean or average can be used to calcu-

late the F1-score. It is ideal if the model can find the right balance between precision
and recall [20]. The F1-score is computed by multiplying the precision and recall
values twice by the sum of the precision and recall values [22].

F1 - Score = 2 ∗ Precision ∗ Recall/Precision + Recall (5)

6.8 Predicted Output

The machine learning model correctly predicted or classified the diseases that the
patient is affected by. The algorithms expected performance is compared to the real
output. The predicted output can be verified by experts or doctors.

7 Conclusion

More work is being done on disease detection using machine learning, as we have seen
the trend for disease diagnosis using machine learning. The proposed methodology
will help in diagnosing diseases using machine learning on which we will work on.
In the current state of research and implementation, there was a lack of systematic
disease detection. The investigation of how machine learning is used in disease
detection which is essential for improving the efficiency of disease detection systems
and ensuring that future research remains relevant.
52 S. Kumar et al.

References

1. Kalaiselvi K, Deepika M. Machine learning for healthcare diagnostics. In: Jain V, Chatterjee
J (eds) Machine learning with health care perspective. Springer. ISBN: 978-3-030-40849-7.
https://doi.org/10.1007/978-3-030-40850-3
2. Machine Learning Approaches that Every Data Scientist Should Know, Medium
(2021) https://towardsdatascience.com/4-machine-learning-approaches-that-every-data-scient
ist-should-know-e3a9350ec0b9. Last accessed 2021/04/18
3. Yusuf M, Atal I, Li J et al (2020) Reporting quality of studies using machine learning models
for medical diagnosis: a systematic review. BMJ Open 10:e034568. https://doi.org/10.1136/
bmjopen-2019-34568
4. Hastie T et al (2005) The elements of statistical learning: data mining, inference and prediction.
The Mathematical Intelligencer 27:83–85
5. Mitchell TM (1997) Machine learning, vol 45, issue 37. McGraw Hill
6. Machine learning, Wikipedia. https://en.wikipedia.org/wiki/Machine_learning. Last accessed:
2021/04/11
7. Russell SJ, Norvig P (2010) Artificial intelligence: a modern approach, 3rd edn. Prentice Hall.
ISBN: 9780136042594
8. Medical prognosis using machine learning, Neuraldesigner.com (2021) https://www.neuraldes
igner.com/solutions/medical-prognosis. Last accessed 2021/04/24
9. Faggella D. Machine learning in healthcare: expert consensus from 50+ execu-
tives. https://emerj.com/ai-market-research/machine-learning-in-healthcare-executive-consen
sus/. Last accessed 2021/04/15
10. How to Get Started with AI in Healthcare - A 3-Phase Approach, Emerj (2021) https://emerj.
com/partner-content/get-started-ai-healthcare/. Last accessed 2021/04/20
11. Cleophas TJ, Zwinderman AH (2015) Machine learning in medicine—a complete overview.
Springer International Publishing
12. Topol EJ (2019) High-performance medicine: the convergence of human and artificial
intelligence. Nat Med 2019(25):44–56
13. Emhamed R et al (2020) Investigation of the performance of machine learning classifiers for
pneumonia detection in chest X-ray images. In: 2020 IEEE international conference on electro
information technology, EIT. IEEE, Chicago, USA, pp 98–104
14. Gupta et al (2020) Evolutionary algorithms for automatic lung disease detection. Measure-
ment140:590–608
15. Boban, Megalingam (2020) Lung diseases classification based on machine learning algo-
rithms and performance evaluation. In: International conference on communication and signal
processing, ICCSP, 28–30 July 2020. IEEE, Chennai, India, pp 315–320
16. Kaushik S. Feature selection methods, machine learning, analytics Vidhya. https://www.ana
lyticsvidhya.com/blog/2016/12/introduction-to-feature-selection-methods-with-an-example-
or-how-to-select-the-right-variables/. Last accessed 2021/04/28
17. The importance of k-fold cross-validation for model prediction in machine learning. Medium.
https://towardsdatascience.com/the-importance-of-k-fold-cross-validation-for-model-predic
tion-in-machine-learning-4709d3fed2ef. Last accessed 2021/04/24
18. Confusion matrix-Wikipedia. https://en.wikipedia.org/wiki/Confusion_matrix. Last accessed:
2021/05/01
19. Confusion Matrix. Image Source: https://2.bp.blogspot.com/-EvSXDotTOwc/XMfeOGZ-
CVI/AAAAAAAAEiE/oePFfvhfOQM11dgRn9FkPxlegCXbgOF4QCLcBGAs/s1600/con
fusionMatrxiUpdated.jpg
Human Disease Prognosis and Diagnosis Using Machine Learning 53

20. Towards Data Science. https://towardsdatascience.com/accuracy-recall-precision-f-score,

https://towardsdatascience.com/accuracy-recall-precision-f-score-specificity-which-to-
optimize-on-867d3f11124-specificity-which-to-optimize-on-867d3f11124. Last accessed
2021/04/30
21. Sensitivity and Specificity—Wikipedia. https://en.wikipedia.org/wiki/Sensitivity_and_specif
icity. Last accessed 2021/05/02
22. Rachmad A et al (2020) Mycobacterium tuberculosis images classification based on combining
of convolutional neural network and support vector machine. Commun Math Biol Neurosci
1–13. https://doi.org/10.28919/cmbn/5035
Incorporating Financial News Sentiments
and MLP-Regressor with Feed-Forward
for Stock Market Prediction

Junaid Maqbool, Preeti Aggarwal, and Ravreet Kaur

Abstract Stock market being very volatile depends on various political, environ-
mental, and internal factors. The stock price prediction using news data is an inter-
esting research topic. In this paper, an approach is proposed that represents textual
news data as sentiment metrics using VADER sentiment analysis and price data scaled
down between 0 and 1. The output predicted price of a stock on a particular day is
fed forward to the next level of MLP* -Regressor to train as well predict the prices of
following days. Experiments have been conducted on 10-year financial news as well
price data of Reliance Company using the proposed model. The results show that
the model because of feed-forward was able to learn the trend and the depths were
followed more closely than the spikes. The model was tested on the news data of the
same date as well as on the previous date separately. The model is an improvement
made to MLP-Regressor whose results are also compared. The MLP-Regressor with
feed-forward was able to learn long-term trends and also predict with an accuracy of
0.714 for the upcoming 7 days.

Keywords Stock prediction · News sentiment analysis · MLP-Regressor ·

Forecasting
*
MLP Multi-level Perceptron

1 Introduction

The price of a single stock fluctuates more frequently compared to the stock indices
and other metrics because of its dynamic and non-linear manner. The stock prices
are very unstable and uncertain due to the dependence of its price on an uncountable
number of parameters. The stock trade is done online dynamically. Traditionally,
the prediction decisions are based on the historical trend as known by the trader

J. Maqbool (B) · P. Aggarwal · R. Kaur

UIET, CSE Department, Panjab University, Chandigarh, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 55
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_5
56 J. Maqbool et al.

and the current market knowledge [1]. The stock market is often influenced by
political, financial, environmental factors, and the psychology of the investors [2].
The investors buy and sell their stocks based on all these events which let them predict
possible up or down in the market. As the financial news is one of the most important
factors which has the power to influence the psychology of traders to invest in a
particular stock. The traders can gain profits if they will be able to predict correctly
the stocks whose prices are going to increase in near future and buy them and sell
the stocks whose prices are going to decrease [3]. The prediction of stock prices has
attracted researchers from computer science, statistics, economic, and operations
research [4].
Different models of machine learning like Bayesian Networks [5], Artificial
Neural Networks [6], Support Vector Machines [7], Multi-Level Perceptron (MLP)
[8], and Recurrent Neural Network especially Long Short-Term Memory (LSTM)
[9] have already been utilized to predict the trends and the future stock prices. As
the stock market exhibits natural chaotic behavior as different world markets react
with different intensities to the period of crisis hence cannot be predicted by simple
trading strategy [10].
The motivation for this study is to build a model which will be able to learn and
predict the upcoming stock prices based on the sentiments calculated from financial
news articles. The stock market being one of top most money investing destination,
hence it will help investors to earn more profits and invest with more possibility of
gains. This will help and encourage new investors to invest in stock market with
limited information about market.
The main objective of this paper is to predict the future stock prices as well the
trend of the direction of the stock price in which it will move based on sentiments
[11] of news data provided for the same day and the previous day. This paper takes
into consideration a new parameter called label which signifies if that particular
date any news regarding the given stock is present. This paper will also check the
effect of this new parameter “label” on the stock prediction. The machine learning
models in [5–8] do not take into consideration the previous day prices thereby model
learns only the fluctuations based on the change in sentiments while [9] do not take
the sentiment of the news thereby predicting using historical trend and not taking
in consideration the sentiment. Our model MLP-Regressor with feed-forward takes
into consideration the previous day’s price which given short-term historical support
and news sentiments to predict the trend in the stock prices. Our model was able to
follow the trend in long term.
This model is tested and trained on the stock data of reliance industries. The model
is tested on the sentiments of the present day and the previous day separately. The
model is compared with the MLP-Regressor without taking the previous day’s price
into consideration. The model was also tested with and without the label parameter
and in turn, these models are evaluated in terms of Mean Absolute Percentage Error
(MAPE), precision, recall, accuracy, and F1 measure.
The problem statement of this paper is to check the prediction accuracy of stock
prices while incorporating financial news articles along with historical stock data
using MLP-Regressor.
Incorporating Financial News Sentiments and MLP-Regressor … 57

This paper is organized as follows: Section 2 briefly discusses existing work in

the related field. Section 3 discusses the approach followed in this paper, models
implemented, and the results derived from the experiments. Section 4 concludes and
discusses future scope.

2 Related Work

There are two traditional approaches for stock prediction: Technical Analysis and
Fundamental Analysis [12]. Most of the models made employ these two traditional
approaches for prediction where fundamental analysis takes into consideration the
overall financial conditions of the company, their management, and economy [1].
The models which employ a technical analysis approach generally take prediction
as a classification problem where historic time series data is used to learn the market
pattern. The models which work to predict exact stock price are termed as predictive
regression in economic literature [13]. Though simple and naïve approaches mostly
suffer from over-fitting when applied to real-world setups and are unable to learn the
long-term trend. The Recurrent Neural Networks especially LSTM can work better
on the long-term trend and prove to be superior to ARIMA (AR—autoregressive,
I—integrated, MA—moving average) (Table 1) [9, 14].
The LSTM models are now being extensively used for stock prediction because
of their ability to store the historical trend in architectural memory. LSTM models
clubbed with other deep learning techniques show promising results in stock
prediction.

3 Proposed Methodology

This section will introduce our methodology, feed-forward MLP-Regressor with

sentiment analysis. The framework is shown in Fig. 1. The training of the model
on time series data was done in a sequential manner without shuffling the data. The
output of each day was fed as an input to the next day too.
During training, the output of the first prediction was fed into the input for the
second prediction which helped the model to predict temporally for each day. The
predictions have shown because of feed-forward the model was able to predict the
trend for a longer period.

3.1 Problem Definition

To predict the stock price for an individual stock using the news sentiments, label,
and historical stock price data. The task is to predict the movement of the stock price
58 J. Maqbool et al.

Table 1 Related work in stock prediction

Research work Technique/Algorithm Dataset Observation
Stock Prediction RNN LSTM on NYSE [GOOGL]a The LSTM model can
using LSTM [9] historical time series predict the opening
data prices while learning
the long-term pattern
and performs better
than ARIMA [9]
Stock price prediction ARIMA, LSTM with Standard and Poor’s The models did fit the
using News sentiment stock price and textual 500 (S&P 500), data with some
analysis [14] information Online newspaper considerable accuracy
articles but did not perform
well when the stock
prices were less or
excessively volatile
[14]
Improved deep Empirical wavelet S&P 500, The hybrid novel stock
learning model for transform (EWT), Dow Jones Industrial prediction model is
prediction using time Outliner Robust Average (DJI) introduced using EWT,
series and sentiment Extreme Learning ORELM, LSTM, and
[15] Machine (ORELM), PSO. The proposed
LSTM, hybrid model has
Particle Swarm better accuracy than
Optimization (PSO) other deep learning
models and any single
model [15]
Ensemble of Deep Q-learning, S&P 500, The proposed model
Q-learning agents for Double German DAX doesn’t learn from
stock market Q-learning, market ups and downs,
forecasting [10] hence is flexible
against the chaotic
behavior of stock price
rather it learns to
maximize the return
over time [10]
An integrated Attention based Chinese stock The proposed model
framework of deep bidirectional LSTM, Exchange market data takes into
learning and Convolutional Neural consideration real
knowledge graph for Network transaction records
stock prediction [16] (CNN), data and utilizes CNN
Deep Stock-trend to extract daily group
Neural Network trading vectors.
(DSPNN) DSPNN outperforms
LSTM because of the
attention mechanism
and bidirectional
structure of DSPNN
[16]
(continued)
Incorporating Financial News Sentiments and MLP-Regressor … 59

Table 1 (continued)
Research work Technique/Algorithm Dataset Observation
A Multiplicative Volume Weighted NIFTY 50 When lag and
self-attention Average prediction exogeneous features
Bidirectional LSTM using multiplicative are considered, The
for stock prediction self-attention layer proposed model has
[17] higher accuracy over
LSTM and simple
RNN [17]
Stock prediction CNN, LSTM, new Historical data of Convolutional layer
using graph-based method named Stock stocks in Taiwan and used to extract financial
CNN-LSTM [18] Sequence Array America features combined with
Convolutional LSTM LSTM achieves better
(SACLSTM) performance than any
of method individually
[18]
Stock price pattern BP algorithm, Fuzzy Gree electric, Maotai With improved training
using Neural Network algorithm of shanghai speed of BP algorithm
and Back Propagation mainboard the accuracy is still
(BP) algorithm [19] better than fuzzy
algorithm to predict the
closing stock price
a NYSE = New York Stock Exchange, GOOGL—Google (historical stock price)

Fig. 1 Feed-forward MLP-regressor

after analyzing the sentiments of the financial news and the previous stock price for
the next trading day and also predict the trend of the stock prices for upcoming n
days. The problem has two tasks (1) what is the direction of the trend and (2) what
is the amount by which the trend changes.
Direction of trend: If the stock price for the next day will be more than the stock
price on the day preceding it, then the trend is positive, and if the stock price is less
on the next day than the preceding day, then the trend is negative. If the stock price
remains unchanged, then the trend is neutral.

T = closing(dayi + 1) − closing(day i) (1)

60 J. Maqbool et al.

Date Open Low High Close

Jan 19-2010 544.84 536.24 547.83 537.41
Jan 18-2010 548.30 540.85 552.98 544.59
Jan 15-2010 557.61 547.34 564.10 550.16

Fig. 2 Historical reliance stock data

⎧
⎨ T > 0, trend = positive
T < 0, trend = negative
⎩
T = 0, trend = neutral

3.2 Data Collection

The stock price data of reliance (Fig. 2) was collected from Jan 2010 to May 2020
from yahoo finance [20]. The closing price of each day along with the news sentiments
applied on the financial news data of the same day using VADER sentiment analysis
[21] from Natural Language Toolkit (NLTK) [22] and label was used to check the
effect of the news sentiments on stock prices or future price of the stock.
Label defines the presence of the news of individual stock in the news collected
on a particular day. Some important keywords related to the stock can be used to
check the label. Label equals 1 when news about the same company is present and
0 when not present.

3.3 Evaluation Metrics

The metrics are used to evaluate the way our model is performing as this model
classifies the data as per future predicted trend as well checks the amount of change.
So, this paper employs metrics like MAPE to check the error in change detection
which doesn’t take the direction of change into consideration and accuracy which
checks the classification correctness without considering the rate of change.
Accuracy In classification, accuracy is among the popular metric used. It repre-
sents the ratio of the number of correctly classified values to the total number of
classified values [10].

X (+)
Accuracy(X ) = (2)
|X |

Precision tells how many predictions are really positive among all positively
classified predictions.
Incorporating Financial News Sentiments and MLP-Regressor … 61

Recall represents the model’s ability to predict correct positives among all
positives in the list.
F1-score model score which is the function of precision score and recall score. It
is the harmonic mean of precision and recall score.

F1-Score = (2 ∗ precision ∗ recall)/(precision + recall) (3)

Mean Absolute Percentage Error (MAPE) It is average relative error of forecast

predictions in percentage. It is one of the popular forecast prediction accuracy metrics
[14, 23].

100  
n
MAPE = (y j − y j )/y j  (4)
n j=1

where yj is the original value and y j is the corresponding predicted value.

3.4 Model Used

This paper has used MLP-Regressor [24] which previously have not shown promising
results in learning the long-term trend [25]. The MLP models without feed-forward
tend to lose the historical information unlike LSTM models [26–28]. Our model
(Fig. 3) uses feed-forward to remember the trend.
The output of one prediction is used as an input to the next prediction helping
to preserve the previous trend and predict the next output in relation to the previous
output instead of making the model work independently on sentiment score. Senti-
ment score affects the stock price but the effect is always shown to the previous
price, not to a general or average price. Accuracy is measured by checking the real
trend and predicted trend for each prediction. MAPE is measured to understand the
percentage change with which the predictions vary to original output.

Fig. 3 Complete feed-forward MLP-Regressor with sentiment and stock input

62 J. Maqbool et al.

3.5 Evaluation and Results

The proposed model was checked against 10-year stock and news data. The training
and testing data were divided into an 8:2 ratio. The models were evaluated based
on the input data keeping the model specifications same. The model was tested with
label and without label to understand the effect of label parameter on the prediction
which showed slight variation in accuracy. Hence label parameter does play role in
improving the prediction accuracy of the model. Then, the model was tested against
the sentiments available. The sentiments for same day (Fig. 4) were used in one case
and sentiment score from the previous day (Fig. 5) was used in the other model to
understand the time effect of the news on the prediction.
The model without label (Fig. 6) was unable to predict as accurately as done by
with model with label. The models were tested for 7 days (Table 2), 30 days (Table
3), 300 days (Table 4), and 512 days (Table 5). Figures 4 and 5 show that both
the models are able to learn the trend and follow it for a long time while previous
day sentiment graph show more overlapping of predicted and original graph. The

Fig. 4 Prediction using

same day news sentiment

Fig. 5 Prediction using

previous day news sentiment
Incorporating Financial News Sentiments and MLP-Regressor … 63

Fig. 6 Feed-forward
MLP-Regressor without
label on same day news data

Table 2 Accuracy metrics for different models for 7 days prediction

Models Precision Recall Accuracy F1-score MAPE
MLP feed-forward with previous day data 0.80 0.80 0.714 0.714 4.62
MLP feed-forward with same day data 1.00 0.66 0.714 0.80 6.98
MLP without feed-forward 0.50 0.40 0.28 0.44 Infinite

Table 3 Accuracy metrics for different models for 30 days prediction

Models Precision Recall Accuracy F1-score MAPE
MLP feed-forward with previous day data 0.56 0.64 0.60 0.60 6.42
MLP feed-forward with same day data 0.58 0.50 0.60 0.53 7.18
MLP without feed-forward 0.47 0.57 0.50 0.51 Infinite

Table 4 Accuracy metrics for different models for 300 days prediction
Models Precision Recall Accuracy F1-score MAPE
MLP feed-forward with previous day data 0.55 0.52 0.52 0.53 6.56
MLP feed-forward with same day data 0.56 0.55 0.54 0.56 7.07
MLP without feed-forward 0.49 0.44 0.46 0.47 Infinite

Table 5 Accuracy metrics for different models for 512 days prediction
Models Precision Recall Accuracy F1-score MAPE
MLP feed-forward with previous day 0.53 0.53 0.53 0.53 5.07
data
MLP feed-forward with same day data 0.52 0.53 0.53 0.52 5.32
MLP feed-forwards without label for 0.50 0.51 0.50 0.50 9.16
previous day data
MLP feed-forwards without label for 0.50 0.50 0.50 0.50 8.09
same day data
MLP without feed-forward 0.51 0.49 0.51 0.50 Infinite
64 J. Maqbool et al.

Fig. 7 MLP-Regressor
without feed-forward on
same day news data

model, MLP-Regressor without feed-forward (Fig. 7), was also tested and it can be
concluded that these models are not able to follow the trend.
For predicting stock price of the upcoming 7 days, the model shows that using
sentiments of same day (Fig. 4) does not increase the accuracy but decreases MAPE
by 2.36%. But without feed-forward, the accuracy is as low as 0.28 and MAPE is
infinite (Table 2). The previous day prediction was used to predict the next day which
helped the models to remember and follow the trend. All models are plotted together
show models are able to learn the trend (Fig. 8). The results suggest the MLP model
with news sentiments from previous day shows better accuracy and lower MAPE
(Fig. 9).
When predicting for 30 days, the accuracy in both feed-forward models the accu-
racy decreases by 0.11 while the MAPE doesn’t show considerable change. And for
the long-term prediction of 512 days, the MAPE decrease by 1–2% and accuracy by
0.07.
While comparing the results with label and without label (Table 5) the accuracy
decrease by 0.03 and MAPE increase by around 4% showing that the label has a
considerable effect on the amount of change in price for a particular stock. The

Fig. 8 Predicted models**

Incorporating Financial News Sentiments and MLP-Regressor … 65

Fig. 9 MAPE comparison of models** (512 days) ** Same Day news data with Label (SDL), Same
Day news data without Label (SDNL), Previous Day news data with Label (PDL), Previous Day
news data without Label (PDNL)

model without feed-forward is able to detect trend with an accuracy of 0.51. MAPE
being infinite justifies Fig. 7 about the model without feedback not being able to
follow or predict trend anywhere.

4 Conclusion and Future Scope

In this work, the individual stock price was predicted using sentiments of financial
news and the time series data of the stock price. The graphs suggest that the label
helps to fit the trend more accurately by decreasing the MAPE and show a high
correlation between the stock price and news of a particular company than other
news. The models were able to detect and follow a downward trend precisely while
an upward trend was detected but the effect of the change was not followed when
stocks were highly volatile. This paper shows that MLP-Regressor when employed
with the feed-forward can provide promising results than MLP without feed-forward.
The feed-forward can be used with more modifications in future work to predict the
stock prices more accurately. There are various ways to predict stock price which
can be explored in future work. Some of these include an ensemble of LSTM and
other models where LSTM can be used to remember the trend and other models to
predict the change rate. Different sentiment analyzing algorithms can be tested to
check the compatibility of different sentiment scores against the ability to detect the
trend of stock prices more accurately and tested against various models.

References

1. Yadav A, Jha CK, Sharan A (2020) Optimizing LSTM for time series prediction in Indian stock
market. Proc Comput Sci 167(2019):2091–2100. https://doi.org/10.1016/j.procs.2020.03.257
2. Yudong Z, Lenan W (2009) Stock market prediction of S&P 500 via combination of improved
BCO approach and BP neural network. Expert Syst Appl 36(5):8849–8854. https://doi.org/10.
1016/j.eswa.2008.11.028
66 J. Maqbool et al.

3. Khan W, Malik U, Ghazanfar MA, Azam MA, Alyoubi KH, Alfakeeh AS (2020) Predicting
stock market trends using machine learning algorithms via public sentiment and political
situation analysis. Soft Comput 24(15):11019–11043. https://doi.org/10.1007/s00500-019-043
47-y
4. Zhang J, Cui S, Xu Y, Li Q, Li T (2018) A novel data-driven stock price trend prediction
system. Expert Syst Appl 97:60–69. https://doi.org/10.1016/j.eswa.2017.12.026
5. Malagrino LS, Roman NT, Monteiro AM (2018) Forecasting stock market index daily direction:
a Bayesian network approach. Expert Syst Appl 105:11–22. https://doi.org/10.1016/j.eswa.
2018.03.039
6. Chong E, Han C, Park FC (2017) Deep learning networks for stock market analysis and predic-
tion: methodology, data representations, and case studies. Expert Syst Appl 83:187–205. https://
doi.org/10.1016/j.eswa.2017.04.030
7. Yeh CY, Huang CW, Lee SJ (2011) A multiple-kernel support vector regression approach for
stock market price forecasting. Expert Syst Appl 38(3):2177–2186. https://doi.org/10.1016/j.
eswa.2010.08.004
8. Turchenko V, Beraldi P, De Simone F, Grandinetti L (2011) Short-term stock price prediction
using MLP in moving simulation mode. In: Proceedings of 6th IEEE international conference
intelligence data acquisition advance computer system technology application IDAACS’2011,
vol 2, no September, pp 666–671. https://doi.org/10.1109/IDAACS.2011.6072853
9. Moghar A, Hamiche M (2020) Stock market prediction using LSTM recurrent neural network.
Proc Comput Sci 170:1168–1173. https://doi.org/10.1016/j.procs.2020.03.049
10. Carta S, Ferreira A, Podda AS, Reforgiato Recupero D, Sanna A (2021) Multi-DQN: an
ensemble of deep Q-learning agents for stock market forecasting. Expert Syst Appl 164:113820.
https://doi.org/10.1016/j.eswa.2020.113820
11. Nti IK, Adekoya AF, Weyori BA (2020) Predicting stock market price movement using senti-
ment analysis: evidence from Ghana. Appl Comput Syst 25(1):33–42. https://doi.org/10.2478/
acss-2020-0004
12. Nti IK, Felix Adekoya A, Benjamin, Weyori A. A systematic review of fundamental and
technical analysis of stock market predictions. Artif Intell Rev 53:3007–3057, 123AD. https://
doi.org/10.1007/s10462-019-09754-z
13. Rapach DE, Strauss JK, Zhou G (2013) International stock return predictability: what is the
role of the United States? J Finance 68(4):1633–1662. https://doi.org/10.1111/jofi.12041
14. Mohan S, Mullapudi S, Sammeta S, Vijayvergia P, Anastasiu DC (2019) Stock price prediction
using news sentiment analysis. In: Proceedings—5th IEEE International Conference Big Data
Serv Application BigDataService 2019, Work Big Data Water Resour Environ Hydraul Eng
Work Medical, Heal Using Big Data Technol, pp 205–208. https://doi.org/10.1109/BigDataSe
rvice.2019.00035
15. Liu H, Long Z (2020) An improved deep learning model for predicting stock market price time
series. Digit Signal Process Rev J 102:102741. https://doi.org/10.1016/j.dsp.2020.102741
16. Long J, Chen Z, He W, Wu T, Ren J (2020) An integrated framework of deep learning and
knowledge graph for prediction of stock price trend: an application in Chinese stock exchange
market. Appl Soft Comput J. 91:106205. https://doi.org/10.1016/j.asoc.2020.106205
17. Hegde S (2021) Multiplicative self attention bi-directional long short-term memory network
for volume weighted average, vol M, no 05, pp 2265–2272
18. Ming-Tai Wu J, Li Z, Herencsar N, Vo B, Chun-Wei Lin J. A graph-based CNN-LSTM stock
price prediction algorithm with leading indicators, vol 1, p 3. https://doi.org/10.1007/s00530-
021-00758-w
19. Zhang D, Lou S (2021) The application research of neural network and BP algorithm in stock
price pattern classification and prediction. Futur Gener Comput Syst 115:872–879. https://doi.
org/10.1016/j.future.2020.10.009
20. Yahoo Finance, Relaince. https://finance.yahoo.com/quote/RELI/history?p=RELI
21. Hutto CJ, Gilbert E (2014) VADER: a parsimonious rule-based model for sentiment analysis
of social media text. In: Proceedings of the 8th international conference weblogs social media,
ICWSM 2014, pp 216–225
Incorporating Financial News Sentiments and MLP-Regressor … 67

22. Loper E, Bird S (2002) NLTK: the natural language toolkit. https://doi.org/10.3115/1225403.
1225421
23. de Myttenaere A, Golden B, Le Grand B, Rossi F (2016) Mean absolute percentage error
for regression models. Neurocomputing 192:38–48. https://doi.org/10.1016/j.neucom.2015.
12.114
24. Pimprikar R, Ramachandra S, Senthilkuma K (2017) Use of machine learning algorithms and
twitter sentiment analysis for stock market prediction. Int J Pure Appl Math 115(6):521–526
25. Ebrahimpour R, Nikoo H, Masoudnia S, Yousefi MR, Ghaemi MS (2011) Mixture of mlp-
experts for trend forecasting of time series: a case study of the tehran stock exchange. Int J
Forecast 27(3):804–816. https://doi.org/10.1016/j.ijforecast.2010.02.015
26. Selvin S, Vinayakumar R, Gopalakrishnan EA, Menon VK, Soman KP (2018) Stock price
prediction using LSTM, RNN and CNN-sliding window model. In: 2017 International confer-
ence advance computer communication informatics, ICACCI 2017, vol 2017-Janua, no January
2018, pp 1643–1647. https://doi.org/10.1109/ICACCI.2017.8126078
27. Roondiwala M, Patel H, Varma S (2015) Predicting stock prices using LSTM. Int J Sci Res
6(4):2319–7064
28. Jiang Q, Tang C, Chen C, Wang X, Huang Q (2019) Stock price forecast based on LSTM neural
network. Springer International Publishing
Framework for Context-Based Intelligent
Search Engine for Structured
and Unstructured Data

Ritendra R. Sawale, Lokesh Nagar, and Vaibhav K. Khatavkar

Abstract In today’s time, determining the user’s exact need through search queries
is really a significant challenge as there is a tremendous amount of structure and
unstructured data which is being produced daily. In structured data, it is easy to
extract what we need as compared to unstructured data. There is a need to understand
the semantics of the text given in such textual data. Natural language processing
helps us to extract useful information from such unstructured textual data. Word
embedding is one of the ways to overcome this issue. The implemented system
aims to build a framework that will search based on the context hidden in the user
query. As the context of keywords plays a vital role while extracting relevant search
results from the database, the implemented system works on determining the context
of the keyword in the query by using the GloVe word embedding technique. The
embedded query is used to find out the most relevant documents from the database.
This database consists of various text documents of different formats like pdf, word
files, excel sheets, online crawled data, etc. All this data is stored in a database name
as ElasticSearch. The proposed system can be used as an intranet searching system.
The most relevant data is sent as an output quickly. The existing entity-based search
engine is not having contextual capability which is provided by the implemented
system. The result for search queries are based on the combination of entity and
context-based search system.

Keywords Natural language processing · Artificial intelligence · Machine

learning · ElasticSearch

R. R. Sawale (B) · V. K. Khatavkar

Department of Computer Engineering, College of Engineering Pune, Pune, India
e-mail: sawalerr19.comp@coep.ac.in
V. K. Khatavkar
e-mail: vkk.comp@coep.ac.in
L. Nagar
SAS Research and Development, Pune, India
e-mail: Lokesh.Nagar@sas.com

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 69
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_6
70 R. R. Sawale et al.

1 Introduction

An unprecedented amount of data has been created daily. According to IDC, the
digital universe will reach over 40ZB by 2020, which is approximately 1.7 MB
of new information created for every human, every second, of every day. A large
portion of this information is unstructured and in text format. This created a need
for a system to analyze this unstructured data to get useful insights from it. Text
mining (also called text data mining or text analytics) is a method for extracting
useful information from unstructured data through the identification and exploration
of large amounts of text. In other definition, text mining is a method for extracting
structured data from unstructured text.
It becomes a need for various organizations to get useful information from such
unstructured data. This unstructured data in any IT organization includes feedback
from customers about specific products or services, comment on social media about
the product, tweets related to the company, etc. All this data is in the form of unstruc-
tured textual data. It is a headache for data analysts to go through each feedback
and comment manually to get the information about the updates or requirements of
clients and users. Text mining helps us to determine the context of the unstructured
data. Natural language processing plays a vital role in this scenario. The searching
interface helps the user to get the most relevant data concerning the search query.
Search query is the most used way for retrieving relevant data from the data reposito-
ries. As the size of data in just a repository is enormous, it becomes a challenge for a
searching system to get only the relevant document as it also depends on the context
of keywords used in the system. The proposed system’s main aim is to determine the
context of the keyword used in the search query by using artificial intelligence, NLP,
and machine learning. There are various ways to determine the semantic of the text,
like word embedding. Here, we will assign a vector value to the word according to
its context, and words with similar context will be placed close to each other. So after
getting the context of the text or keyword, the system will search for the most relevant
documents from the data repository. As the keyword context plays a vital role while
extracting relevant search results from the datasets or database, the proposed system
would mainly work on determining the context of the keyword in the query by using
different word embedding techniques.
The main motive of the implemented system is to get only the relevant documents
using ElasticSearch. But we are using AI over the search query to understand many
things like the context, spelling mistake, keywords, etc. Our system will also consider
the user’s role, depending upon which it will provide only authorized information,
so such an intelligent search system for structured and unstructured data is proposed
in this report. The system will also make the query more machine-readable so as
to get relevant data from the documents. Further literature survey of related work
is also done. The objectives and scope of this project are also defined along with
system design. The data repository consists of data from different data sources.
ElasticSearch is used to store all the documents as well as the sentences from the
Framework for Context-Based Intelligent … 71

documents. GloVe embedding is used to find the vector of the sentence which focuses
mainly on the context of the search query. So the sentence vectors are also stored in
the ElasticSearch.
The existing system named as text exploration framework which is the product
developed by SAS R&D is an entity-based search engine. The implemented system
provides contextual capability to the existing search engine using word embedding,
due to this relevancy of search result is improved efficiently. The combined effect of
this two model result into a better search engine.

2 Related Work

There are various implementations of the searching system by many researchers, but
AI-based searching systems are rarely done by anyone. Word embedding can be used
for various applications in NLP. Many researchers have implemented various systems
using different word embedding techniques. Word2vec, GloVe, Bert, fasText, etc.,
are the types of word embeddings.
Silva-Fuentes et al. [1] had implemented an information retrieval system using
Word2vec to determine the semantic of information. The terms in the search query
are used for searching along with the new terms added in the query, so the search
query expansion is done by using word embedding. Salim et al. [2] have proposed a
model which is a combination of recurrent neural network (RNN), long short-term
memory (LSTM) and word embedding. Firstly, the words are represented into a
vector by using word embedding. After this, the LSTM cells are initialized by using
the attribute features. The work of neural networks is more significant than any other
approach for detecting sarcasm. Within just 15 epochs, the model gives an accuracy
of 88%. Yenala et al. [3] have proposed a novel deep learning technique that is used to
detect the inappropriate language in the text. The inappropriate text includes various
messages, feedbacks, or comments on social media posts. A deep learning model is
proposed for detecting inappropriate suggestions in queries known as “Convolutional
Bi-Directional LSTM” (BiLSTM). BiLSTM is the combination of two techniques,
CNN and bi-directional LSTM. C-BiLSTM provides better results than LSTM, CNN,
or any other deep-learning-based model if used individually. Jeffrey Pennington et
al. [4] have developed the GloVe model. The model is trained on a massive corpus
like Wikipedia with million of word tokens. The GloVe is used to assign vectors
depending upon the global context of the word. It also outperforms related models
on similarity tasks and named entity recognition. The model produces a vector space
with a meaningful substructure as evidenced by its performance of 75% on a recent
word analogy task. Srinivas and Jairaj [5] have developed a Siamese network; it learns
the uncommon food dish names embeddings from scratch. Those names are out of
vocabulary for traditional word embedding. So, for improving the performance of the
food search system, this proposed method will help strongly. The dataset used for the
proposed method is in-house food taxonomy data. Timo Spinde et al. [6] presented a
system to analyze the context of the related (meaningly) words. They have used two
72 R. R. Sawale et al.

training models of word embedding. One model is used to train on left-wings news
articles, while another one is trained on right-wings news articles. The main idea is
to prove that the context of the biased word varies in different news outlets than one
of the non-biased words. This system is automated detection of media bias using the
implicit context of bias words that will derive through fine-tuned word embedding.
Two datasets are used—WordSim-353 and MEN and the Google analogy test set.
Han and Zhu [7] used a framework known as Flask, which is a Python framework,
to make a system for rapid retrieval and visualization of media data. The crawled
network’s raw data is initially stored in a non-relational database like MongoDB. A
cluster is set up by ElasticSearch, where the data is imported. After this, map for
data is created, and an inverted index is set up. Due to this, data retrieval becomes
more accurate as well as fast. Kibana is used to display the data and to present it.
Khatri et al. [8] have proposed a sarcasm detection system using machine learning
and two-word embedding techniques. BERT and GloVe embeddings are used in this
system. The GloVe is used to determine the response vector, while BERT is used to
get the context vector. GloVe and BERT embeddings, when used together, produce
a maximum F-measure score of 0.690 with logistic regression as a classifier.

3 Word Embedding: GloVe Embedding

The GloVe is one of the techniques used to find the word vector [10]. Word vector is
the representation of words in vector space such that words with similar contexts clus-
ter together while different words repel. The GloVe model productively influences
statistical data by using only non-zero values of word–word cooccurrence matrix for
training. Word–word cooccurrence is a sparse matrix. So only the non-zero elements
are considered.The GloVe is used in this system as it assigns the vector depends upon
the word’s global context, unlike Word2vec which is dependent on the local context.
The dimension of each vector can be of various dimensions, say 50, 100, 200, and
300. So GloVe provides a pre-trained model trained on big datasets like Wikipedia,
Twitter, etc., which can be downloaded directly from the Web. It contains millions of
words along with their vectors in various dimensions. But there can be some words
from the corpus that are not in the pre-trained model, so the vector assigns them zero.
There is a possibility that some words with vector zero are useful for users, and the
result of the search becomes irrelevant. So it is important to train the GloVe model
on the input corpus. So the implemented GloVe model is the combination of both
pre-trained and trained GloVe models. While training the GloVe model on the user-
defined corpus, various parameters like corpus length, vocabulary size, etc., need to
be changed, and thus, it directly affects the quality of embedding. The prerequisite
for getting good-quality embedding is the size of the input corpus. If the size of the
corpus is huge like 10 million words, then the embedding would be better than the
one with a corpus size of 10 thousand words. While training the model, one more
Framework for Context-Based Intelligent … 73

Fig. 1 GloVe training

parameter is important which is number of epoch. Figure 1 clearly depicts if more

epochs are given, then the loss will further reduce and the embedding will be much
better. This process is very time-consuming and requires huge computational power.

4 ElasticSearch

ElasticSearch is an open-source tool freely available on the Web; it is a full-text search

and investigation engine. ElasticSearch provides the capability to store, search, and
examine a vast amount of data in close to real-time. It is a highly efficient system
for searching through unstructured data. It is utilized as the fundamental engine that
powers applications with complex search requirements.
Elasticsearch is high-speed, scalable, and flexible, making the elastic stack a robust
solution for many different use cases. Elastic search is useful for security purposes
like finding threats and prevention, an enterprise search engine, and many other
use cases. As there is a great need for a document-oriented database, ElasticSearch
provides this capability. The indexing of the whole object graph needs to be done, but
it is important to denormalize it first. This denormalization is important to improve
the document retrieval performance as there is no need to join a query several times,
but keeping all the instances consistent is important, and updating is quite difficult.
Write-once-read-many-workload is suitable for it.
74 R. R. Sawale et al.

4.1 Why Indexing in ElasticSearch?

ElasticSearch searching for an index is much faster than searching for text into the
documents, which is why ES is capable of performing fast searching. Searching
through elastic search is like scanning the index of the documents related to the
keywords and then retrieving the relevant data from the documents. Search for text
on each and every page of a book is more difficult than searching for the index related
to the keyword. This is known as an inverted index because it inverts the page-centric
data structure, i.e., page -> words, to keyword-centric data structure, i.e., word to
pages. ElasticSearch utilizes Apache Lucene for managing and creating the indexes.
An index comprises at least one document, and the document is a collection of one
or more fields.

4.2 How Indexing Is Done in ElasticSearch?

There are two types of indexing: 1. Forward Index and 2. Inverted Index.
Forward Index
It is the most basic way to build up an index. In forward index, a list of all terms
from the documents are stored. It is very fast while indexing the documents. While
appending a new document, there is no need to rebuild the index. But it is not so
efficient when querying. The search engine needs to scan all the entries in the index
searching for the specific term, so as to return all the possible documents containing
that term. The example of the forward index is shown Fig. 2.
Inverted Index
It is a better option to sort the index by the term in the query so that the processing of
the query becomes quicker. As this is the inversion of the forward index, it is named

Fig. 2 Forward indexing example

Framework for Context-Based Intelligent … 75

Fig. 3 Inverted indexing

as an inverted index. By using the inverted index, the machine will need to look for
a term only once to successfully retrieve all the relevant documents containing that
term. So ElasticSearch makes use of inverted index as as to provide faster retrieval
of documents. An inverted index makes a list of all the unique words occurring
in the documents, and it also identifies the documents where each word occurs in
(Fig. 3).
The implemented system is divided into two parts Fig. 4. In the first part, all the
modules run as a batch process that continuously works to learn new vocabulary
from the fed documents. The user’s input is in the form of a text document that
includes PDF files, docs files, excel sheets, CSV files, etc. After accepting this input
file, the first process is to convert the file into a regular text file. There is a need
to extract only the text from the text file, and the images should be neglected as
there is no need for an image; our system is not extracting text from image files.
So the corpus is formed from the documents, but it contains a lot of unwanted
text and symbols that are of no use. It is necessary to clean our corpus; only the
required text should be there. So the next step is data preprocessing. In this step,
we are removing unwanted symbols, punctuation marks, multiple spaces replaced
with single space, converting whole text into lowercase, removing all the special
characters, stopwords elimination, lemmatization, etc. After the data preprocessing
step, the corpus contains only meaningful words. Now, the clean corpus is fed into
76 R. R. Sawale et al.

Fig. 4 Concrete system design

two modules simultaneously to paragraphs to sentences converter and GloVe model

for training.The paragraph to sentences converter makes use of a sentence tokenizer
from NLTK. The NLTK is a Python library and can be used for many purposes,
mainly for NLP.

5 Implementation Details

Sent-tokenize [9] smartly split the paragraph into sentences with the help of punctu-
ation marks. The sent-tokenize function uses an instance of PunktSentenceTokenizer
from the nltk.tokenize.punkt module, which is already being trained and thus very
well knows to mark the end and beginning of sentence at what characters and punc-
tuation. So after converting every paragraph into sentences, we need the vector for
the whole sentence, which can be done by using various word embedding techniques
like Word2vec, GloVe, Fast text, Bert, etc. The GloVe is used in this system as it
Framework for Context-Based Intelligent … 77

assigns the vector depends upon the word’s global context unlike Word2vec which
is dependent on the local context. The dimension of each vector can be of various
dimensions, say 50, 100, 200, and 300. With the help of GloVe embedding, the sen-
tence is embedded into a sentence vector and stored in a database for future use.
The dimension of vectors are also important, so all the possiblitiese are checked
while testing. ElasticSearch provides capabilities to search, store, and analyze vast
amounts of data. This data can be structured as well as unstructured in nature. Elas-
ticSearch is a high-speed, scalable, and flexible open-source tool, making the elastic
stack a robust solution for many different use cases. Elastic search is useful for secu-
rity purposes like finding threats and prevention, an enterprise search engine, and
many other use cases. ElasticSearch is used to store all the documents. It exhibits
useful properties like the ease in storing data, fast searching through the extensive
database, and quick data retrieval. So ElasticSearch is the best option for storing all
the user’s documents to perform its searching task more effectively and add up to
the search system’s overall performance. So two separate indices are created within
elasticsearch. The first index will store all the documents along with its names. Stor-
ing of data in ES is in json format, so document name and the extracted text from
document is stored in this index. In the second index, various fields are stored like
document names, sentences from the document, and their respective sentence vector.
ElasticSearch helps to store the whole data efficiently without any loss of data as
well as quickly. Retrieval of data is also very easy and fast as compared to traditional
databases like MySql, Oracle, etc.

6 Results and Discussions

The result of the search system is dependent on the context of the search query which
is the main motive of this project. With the help of word embedding, it was possible
to determine the requirements of the user, and the relevant documents as well as the
matching sentences from the document are displayed as output of the system. To test
the searching system, a crawl dataset of COVID-19-related news article from Times
Of India is created. The dataset consists of three columns, mainly headline, news
article, and link of the article. So the articles are first preprocessed, and then, the
GloVe model is trained on the corpus created by using the news data. After training,
the model can embed out of vocabulary words like COVID-19, remdesiver etc., which
are gathered from the news. Table 1 shows the query and the output documents names
(news headline) from the dataset. To visualize the embedding of out of vocabulary
words, a scatter plot of top-20 most similar words is created as shown in the figure.
The figure shows the words which are most similar to the words like “Covid vaccine”
shown in Fig. 5. It is clear from the figure that the embedding is based on context of
the word used in the corpus.
78 R. R. Sawale et al.

Table 1 Query with output document

Query 1. Oxygen Concentrator news?
1. Companies shut ops to supply oxygen, provide bikes to medics fighting pandemic
2. PM Narendra Modi reviews oxygen status, production to be increased
Query 2. Covid vaccine related news
1. Serum Institute asks government for 403 million to boost AstraZeneca vaccine output
2. Double mutant of coronavirus fast replacing dominant N440K variant in south India Scientists
Query 3. Remdesivir drugs?
1. Overuse of steroids could be counter productive, says AIIMS chief
2. No Covid appropriate behaviour in Kumbh, Ramzan, says Amit Shah
Query 4. Daily death poll due to covid?
1. Covid At 685 deaths, India records highest daily toll in 5 months
2. With 89,129 new cases, India s daily coronavirus infections hit six month high

Fig. 5 Contextually similar sentences

The proposed system is also tested by providing different search queries as the
input to the system. Later on, the result is checked manually to see whether relevant
results are given as output or not. So some of the results are shown in Table 1. The
table shows the queries and the output documents for the respective query.
From the result, it can be concluded that the system is considering the context of the
query, and depending on that, the searching is done. In order to check the contextual
similarity of sentences or words, a scatter plot shown in Fig. 6 is also built for some
sentences to see if the vectors are calculated using the context words of the sentence.
So the scatter plot easily shows that sentences with similar contextual meaning are
forming a cluster and are at a distance from the one which is not contextually similar
to them. To map the vectors which are of 50 dimensions into 2 dimensions, we make
use of PCA and t-SNE. And these two-dimensional vectors are then used to plot the
point on the scatter plot graph. Similarly, a scatter plot of contextually similar words
is also built using the same technique. And both the scatter plot graph is showing
appreciable output by clusteing similar words or sentences together to form a cluster
(refer Fig. 6).
Framework for Context-Based Intelligent … 79

Fig. 6 Contextually similar words

7 Conclusion

With the help of word embedding, the main objective of determining the context
of the search query is achieved. The trained GloVe model is able to embed out
of vocabulary words from the corpus, which makes it more efficient. The search
result gives relevant documents depending upon the context of the search query
as well as by matching the term in the query. ElasticSearch makes the searching,
storing, and retrieval of documents so faster and easier. The implemented system
works efficiently on any kind of textual data like news articles, reviews datasets, etc.
For testing purposes, the COVID-19 news article from TOI is used as a dataset by
using a Web crawler, and the performance of the system appreciable. The overall
performance of the implemented system is very appreciable; it is fetching relevant
documents with lightning speed with minimal latency, which is very necessary for
any search engine. The first future scope is integration with the existing system
named as TEF. Exploring new word embedding techniques like the latest one Elmo,
Bert, etc., but various other parameters are needed to be considered. Query expansion
techniques is also a good addition for any search engine. Next word prediction is one
of the important scopes for this project. The last one is to use other techniques like
knowledge graph instead of word embedding.
80 R. R. Sawale et al.

References

1. Silva-Fuentes MA, Calderon-Vilca HD, Calderon-Vilca EF, Cárdenas-Mariño FC (2019) Seman-

tic search system using word embeddings for query expansion. In: 2019 IEEE PES innovative
smart grid technologies conference-Latin America (ISGT Latin America). IEEE, pp 1–6. https://
doi.org/10.1109/ISGT-LA.2019.8894992
2. Salim SS, Ghanshyam AN, Ashok DM, Mazahir DB, Thakare BS (2020) Deep LSTM-RNN with
word embedding for sarcasm detection on Twitter. In: 2020 international conference for emerging
technology (INCET). IEEE, pp 1–4. https://doi.org/10.1109/INCET49848.2020.9154162
3. Yenala H, Jhanwar A, Chinnakotla MK, Goyal J (2018) Deep learning for detecting inappropriate
content in text. Int J Data Sci Anal 6(4):273–286. https://doi.org/10.1007/s41060-017-0088-4
4. Pennington J, Socher R, Manning CD (2014) Glove: global vectors for word representation.
In: Proceedings of the 2014 conference on empirical methods in natural language processing
(EMNLP), pp 1532–1543
5. Vijjali R, Mishra A, Nagamalla S, Sathyanarayna J (2020) Semantic embeddings for food search
using Siamese networks. In: Proceedings of the 4th international conference on natural language
processing and information retrieval, pp 138–143. https://doi.org/10.1145/3443279.3443303
6. Spinde T, Rudnitckaia L, Hamborg F (2021) Identification of biased terms in news articles
by comparison of outlet-specific word embeddings. In: Proceedings of the 16th international
conference (iConference 2021). Springer Nature, Virtual Event China
7. Han L, Zhu L (2020) Design and implementation of elasticsearch for media data.. In: 2020
international conference on computer engineering and application (ICCEA), IEEE. https://doi.
org/10.1109/ICCEA50009.2020.00036
8. Khatri A (2020) Sarcasm detection in tweets with BERT and GloVe embeddings. arXiv preprint
arXiv:2006.11512. arXiv:2006.11512
9. AnalyticsVidhya (2018) https://www.analyticsvidhya.com/blog/2018/11/introduction-text-
summarization-textrank-python/. Last accessed 1 Nov 2018
10. Pengyan510 (2021)https://github.com/pengyan510/nlp-paper-implementation. Last accessed
5 May 2021
Design and Development of Enhanced
Secured Video Communication System
for Medical Video Sensor Network
A. Sivasangari, D. Deepa, R. Vignesh, Suja C. Mana,
and B. Keerthi Samhitha

Abstract Privacy and security of healthcare data are becoming important issue
in healthcare field. The main issue to be considered during data transmission is
the security of the data that can be transmitted through wireless medium to the
remote server. The attacker can perform different types of security attacks on medical
information due to broadcast nature of wireless communication. Thus, an effective
security mechanism to safeguard the data against data alteration and deletion is
the need of the hour. This sort of data tamper would result in wrong diagnosis by
the physician, which is a life-threatening issue. WMSNs (Wireless Medical Video
Sensor Networks) have the ability to store, process, and collect heterogeneous devices
for multimedia data. Requirements such as high bandwidth, low packet loss ratio,
and minimal delay have also been increased. To meet the security requirement, a
secure key exchange method in WMSN is proposed in this paper. The main issue
to be considered during data transmission is the security of the data, which can be
transmitted through wireless medium to the cloud server. The attacker can perform
different types of security attacks on medical information due to broadcast nature
of wireless communication. Additionally, the collected multimedia data is stored
in the cloud. Thus, an effective security mechanism to safeguard the data against
data alteration and deletion is the need of the hour. In order to provide the security in
three levels of data transmission in WMSN, the secure key exchange in WMSN (SK-
WMSN) model is proposed in the present work. Also, we analyzed other proportional
assessment issues in WMSN, for instance, incorporation and security issues. Finally,
we inspected and masterminded the current off-the-rack contraptions, models, and
testbeds executed for WMSNs.

Keywords Key exchange · WMSN · Queue scheduler

A. Sivasangari (B) · D. Deepa · R. Vignesh · S. C. Mana · B. Keerthi Samhitha

Sathyabama Institute of Science and Technology, Chennai, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 81
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_7
82 A. Sivasangari et al.

1 Introduction

For a little while now, researchers and technology experts have been drawn to a
popular use of WMSNs (wireless medical sensor networks), which consists of a
large number of portable sensors with low memory space, processing capacity, and
latency. Several study organizations and programmers have concentrated their efforts
on the electronic-healthcare surveillance infrastructure, using CodeBlue, UbiMon,
SPINE, LiveNet, MobiHealth, Alarm-Net, etc. For WSNs, the role of WMSNs in
healthcare systems has resulted in significant developments in the healthcare sector
in the twenty-first century. WSNs became a science fiction/movie-fantasy subject
for the healthcare industry a few years earlier, but now they’ve become a fact and
would provide much excellent performance. According to a survey taken place in
2008, the people worldwide (≥65 years) were projected to be 502 million, and that
number is expected to reach 1.4 billion by 2030. When more people reach that age,
the risk of developing such serious and chronic conditions increases dramatically. For
instance, Alzheimer’s signs usually occur after the age of 65 years, cardiovascular
disease and stroke symptoms rise just after age 60, and diabetes, such as those of
several other conditions, increases after age of 60 (e.g., blood glucose, blood pressure
levels, etc.). As a result, the elderly population wanted to bring independence and
high-quality treatment without jeopardizing their convenience while lowering their
medical expenses. In this sense, wireless sensor technologies could offer extremely
useful resources for tracking the wellbeing of elderly and patients who require
constant monitoring. As a result, healthcare using wireless technologies is an inno-
vative and expanding area of research inquiry. In reality, the future of contemporary
healthcare in an aging world would necessitate constant health screening with little
actual contact between patients and doctors. Recent times, the concept WMSN (wire-
less medical sensor network) has been coined to pull together many experts from
various backgrounds such as electronics, bioengineering, medicine, computer, etc.
The architecture of a wireless healthcare framework presents several problematic
situations, such as efficient data transfer, node transport protocols and fast inci-
dent tracking, accurate data distribution, power storage, node computing, and web
services. However, introducing emerging technology in healthcare systems without
understanding potential risks endangering patient safety. For example, the patients’
health vital signs are especially susceptible (i.e., whether a patient does have a humil-
iating disease), so any leakage of specific access to healthcare can disgrace him/her.
In reality, disclosing disease details will often lead to a person destroying his or
her employment or making it difficult for her or him to seek health cover. Moreover,
wireless medical sensor networks include a wide variety of healthcare uses, including
medical data analysis, movement tracking in fitness clubs, player position recording,
etc. As a result, WMSNs share the patient data with insurance providers, doctors,
wellness coaches, and family. Such sharing of data can cause problems to and make
patients vulnerable for external threats. In a clear case, a patient’s sensors relay body
information to a nursing staff; however, an intruder could be spying on the patient
information as it is transmitted, compromising the patient confidentiality and the
Design and Development of Enhanced Secured Video Communication … 83

attacker can post the details of the patient to public using social media and it risks
the privacy of the patient. After all, wireless healthcare can provide many benefits to
monitoring patients; nevertheless, while an individual’s physical and mental data is
extremely sensitive, data protection becomes major issues for healthcare implemen-
tations, especially where new communication is used. More specifically, if HIPAA
laws are not enforced correctly, a healthcare worker faces stringent criminal and civil
sanctions (i.e., a penalty or prison sentences). As a result, medical confidentiality and
safety are the primary concerns in smart healthcare.
Data security seems to be the method of securing servers, records, and identities on
a system by implementing a range of controls, programs, and procedures that define
the relative value of various databases, their vulnerability, and regulatory enforcement
criteria, and afterward apply necessary safeguards to defend those properties. Data
security, like other methods such as security controls, file safety, or user behavioral
security, is not really the be-all and end-all of a security practice. It is one way of
assessing and mitigating the risks associated with storing some kind of information.
The data management technologies mentioned below are being used to deter attacks,
minimize damage, and maintain safeguards.
The concern is not when a security violation will occur, nor when it might occur.
Once forensic analysis is called on to investigate the underlying problem of a hack,
getting a data auditing framework in place to monitor and document on access
management modifications to data, who has access to confidential data, when this
was obtained, file route, and so on is critical to the review. Alternatively, by imple-
menting appropriate data auditing methods, IT managers can achieve the exposure
required to avoid unwanted modifications and possible breaches.
Usually, this takes several days for a company to find a hack. Companies also
learn about violations from their clients or third party rather than from their own
IT teams. Through tracking data activity and unusual activities in full detail, you
can detect security violations more easily, resulting in accidental damage, failure,
modification, unintended exposure, or acceptance of data which is personal.
Data Risk Evaluation
Data vulnerability evaluations assist businesses in identifying their most vulnerable
confidential data by providing dependable and repeatable measures to prioritize and
address severe security threats. The procedure begins by defining confidential data
obtained by global networks, expired data, and conflicting permissions. Risk analyses
outline key conclusions, reveal data flaws, explain each flaw in depth, and provide
prioritized remediation suggestions. Data Minimization.
The understanding of data has shifted over the last generation of IT management.
Initially, having additional data always seemed to be better than having less. You will
never know in advance what you would like to do about it. Data is now a liability.
The possibility of a reputation-damaging data leak, damage in billions of dollars, or
substantial regulatory penalties both perpetuate the notion that obtaining something
other than the bare minimum of confidential data is highly risky.
At the end, follow best practices for data minimization and evaluate all data
processing needs and processes from a market perspective.
84 A. Sivasangari et al.

Since the accessibility nodes send private data through wireless networks, any
person may behave as an attacker and interrupt and tamper with the sensitive infor-
mation being transmitted on the WSNs. Furthermore, the attacker can effectively
decrypt data exchanges between clients and data centers. This might lead to a variety
of threats, such as repeat, key imitation, stolen authenticator, and so on. Related types
of threats are regarded, with an emphasis on the insecurity of wireless technology, in
team to address the cloud-centric, multistage encryption as just a service scheme and
also to investigate IoT-based post-disaster management mechanisms. Body sensor
network (BSN) software has been one of the newest Internet of Things (IoT) innova-
tions in healthcare. A patient could be tracked using this software by using a series
of tiny controlled and portable wireless sensors. When such a method is seen in
terms of protection, nevertheless, patient privacy is compromised. The main secu-
rity standards in BSN-based medical systems is a stable IoT-based welfare system
that introduces biometric-based approaches as the third encryption element, as well
as a modern user-anonymous authentication method relying on WMSNs. To ensure
safe and allowed communication in WMSNs, the symmetric key-based authenti-
cation mechanism is built. The following are the practical protection properties of
WMSNs.
Mutual authentication: It enables the sensor, gateway, and medical practitioner to
mutually verify one another.
Session-key agreement: It distributes the session key between patients and medical
specialist in order to secure contact in WMSNs.
Known-key security: And if the attacker tampers with one user’s session key, the
attacker cannot breach the session key of yet another user.
Client anonymity: It masks the identity of people, such as a medical specialist
and a patient. Stolen smart card attack unless the user wants to change his or her
credential, he or she can include the login credentials and identity during the session-
key-update process so that SC and GW access can determine if the inputs are correct
or not.

2 Related Work

Iqtidar et al. [1] talk about a novel intrusion detection system to prevent attacks on
personal medical devices. Authors talk about an efficient intrusion detection system
which can identify the intrusion into personal medical devices. Authors reveal about
some of the possibilities of vulnerabilities for these devices and how the proposed
system effectively detects these intrusions in advance. In paper [2] Liu et al. discuss
about regional medical data sharing using big data technologies. Quantin et al. [3]
describe the implementation of an information system to facilitate the keep tracking
of electronic medical records of people in European regions. Teodor Sumalan et al.
[4] present a solution for the surveillance system which is based on emulating the
data generated by some sign sensors. Paper [5] describes a video based monitoring
based on wireless sensor network technology. This video monitoring station will act
Design and Development of Enhanced Secured Video Communication … 85

as a gateway between various processing nodes and the network. The software and
hardware implementations are studied in detail in the paper by Yifan et al. Sodhro
et al. [6]. They talk about energy efficient video transmission in wireless body sensor
networks. Authors have proposed a sustainable algorithm to address this energy
management issues. This algorithm proved to be effective in increasing the life of
wireless body sensor networks.
Singh et al. [7] propose an energy efficient video streaming architecture appli-
cable to wireless media sensor networks. This implementation tries to reduce the
energy consumption in the network [7]. A novel energy consumption approach is
proposed by the author [7]. In paper [8] by Abbas et al., authors describe about a
video surveillance system suitable for wireless sensor networks. Authors also imple-
mented a dynamic queue scheduler. An open source platform based on raspberry pi is
the core component of this implementation [8]. Qian et al. [9] describe about a video
based wi-fi sensor node. The core function of this node is the air target detection.
This sensor is capable of detecting the air quality index which is a key measure to
determine the level of air pollution [9]. A zigbee based WSN model is described
in paper [10]. This work by A. Jasim et al. describe about centralizing the network
management task at the coordinator of the network. It tries to remove loads from
other nodes [10]. In paper [11], authors describe about traffic violation detection
algorithm. In paper [12], authors talk about a wearable device applying the principle
of wireless sensor networks. Yi et al. [13] propose a method to reduce the number
of blackout nodes. This implementation will increase the data collected by selecting
an appropriate video encoding methods. This proposed architecture is particularly
applicable in solar powered wireless video sensor network. Raj et al. [14] talk about
an iot-based electronic-healthcare system. In this work, authors have presented a
cost-effective health sensor platform for rural health monitoring. Jamshed et al. [15]
describe about a node scheduler which is capable of scheduling tasks depending on
the priorities of tasks. This scheduler helps to attain better quality of service for the
network.

3 Proposed Work

It involves three phases. They are (i) Cluster Formation (ii) data transmission between
the SH and the BS and (iii) data transmission between the BS and the cloud storage.

3.1 Sensor Node Design

A sight and sound sensor hub comprises of an outsized number incorporating a

detecting unit with a camera, a foreign correspondence unit, and a preparing unit;
the whole sensor needs to be controlled by a unit of measurement as demonstrated in
Fig. 1. The preparing unit is particularly vital for the sensor hub. The handling unit
86 A. Sivasangari et al.

Fig. 1 Processing sequence sensor node

needs to have low force utilization, high velocity estimation, and furthermore need to
be little. Thus, we use raspberry pi (RPi) (Raspberry Pi Establishment, Cambridge,
Joined Realm) as a handling unit as demonstrated in Fig. 2a.The Raspberry pi 2 model
B incorporates 1 GB of Smash and incredible 900 MHz quad-center ARM Cortex-A7
(ARM Holding, Cambridge, Joined Realm) central processing unit. It depends on
Broadcom BCM2835 framework on a chip. Its HDMI attachment, a SD space card, a
couple of USB connectors, and an Ethernet connector. With these associations, rasp-
berry pi may be a good instructive work area PC. We use raspbian working framework,

Fig. 2 Input to output queue steps

Design and Development of Enhanced Secured Video Communication … 87

which depends on a debian conveyance. Raspbian working framework (raspbian-

wheezy, Raspberry Pi Establishment, Cambridge, Joined Realm, 2015) may be a
linux/GNU rendition (3.18) created for raspberry pi equipment. Our foundation
comprises of seven raspberry pi model 2B sensor hubs.
The remote correspondence unit has the duty of sending video information in
WMSNs. For remote correspondence, we utilize monetarily accessible EDIMAX wi-
fi dongle as demonstrated in Fig. 2b. EDIMAX wi-fi dongle (EW-7811Un, Edimax,
Taipei, Taiwan) is connected to USB port of each raspberry pi board. The dongle has
information rates up to 150 Mbps, agrees with remote 802.11 b/g/n, and it upholds
shrewd communicate power control and auto-inactive change. With low force utiliza-
tions, these dongles are especially appropriate for quick prototyping for remote sight
and sound sensor network applications.

3.2 The Development and Transmission of Video Bundle

Our point is to send video transfers with high casing rate and great quality. Conse-
quently, IVSP utilizes H.264 standard for video pressure. The worker application
produces a video transfer from either live or put away source and converts video
information in H.264 transfers. As indicated by H.264 standard, there are I, P, and B
outlines. We do not utilize B outlines in tests, on the grounds that the deciphering and
encoding of those edges are reliant upon the subsequent edge. In our analyses, we
use I edges and P outlines. I outline need is in every case tons above P outline. The
length of a bundle needs to be changed in order that it can satisfy the necessities of
transmission. On the off chance that a parcel has extremely limited quantity of data,
each hub must send rapidly. This may expand the likelihood of clog, energy waste,
and more crashes in remote channel. On the off chance that a parcel has enormous
measure of data, which can comprise of various edges, each bundle must stand by
until all the video outlines are made. This may create more setbacks. Hence, length
of parcel should be picked in suitable manner. We accept that the majority extreme
parcel length is l. within the event that the length goes past this edge, the video
information is going to be separated into numerous bundles. The parcel is going
to be I outline bundle, if a bundle contains the knowledge of I outline. Something
else, parcel is going to be P outline bundle. We realize that remote transmission is
problematic, so parcel misfortune may be a conspicuous element. An I outline parcel
misfortune will cause loss of all P outline bundles, since they will not be decoded
needless to say and can be futile. Within the event that P outline parcels are lost,
video deciphering will continue surely aggravations in video information. Within
the event that I outline bundles are retransmitted rapidly, the results of misfortune
parcels won’t be conspicuous. Along these lines, retransmission of I outline bundles
is significant for acceptable quality video information.
88 A. Sivasangari et al.

3.3 Queue Scheduler

As we talked about in Segment 4.1, various sorts of parcels have distinctive retrans-
mission prerequisites. As per the extent of importance, parcels are often separated
into four needs from high to low: The I outline bundles have most elevated need,
P outline bundles have high need, lost I outline parcels that require retransmission
have low need, and lost P outline bundles that needs retransmission have least need.
Figure 4 shows the bundle stockpiling and transmission measure at a sensor hub. At
the purpose when a hub produces or gets an information bundle, the parcel is moved
to the I-Casing parcel line or P-Edge bundle line contingent upon its quality. These
two lines are utilized to store the unverified parcels incidentally, sitting tight for the
retransmission if necessary. I outline bundles are of critical significance in supporting
the constant hour gridlock. Thusly, we give the best got to I outline bundles, on the
grounds that our fundamental point is to make sure I outline parcels. Within the event
that I outline parcel is lost, it goes into lost I outline line, and if a P outline bundle
is lost it goes into lost P outline line. In our examinations, to manage traffic classes
with various needs during a proficient manner, weighted cooperative scheduler is
employed within the weighted cooperative scheduler, we dispensed a weight.

3.4 Congestion Identification

To quantify clog in remote sensor organizations, we present a mix of two diverse

blockage tips that could precisely recognize blockage at every sensor hub. The most
markers are support inhabitance T, and second pointer is cradle inhabitance change
rate δ. The support inhabitance may be a vital marker of clog. It’s received to assess the
measure of cushion involved by the parcels at a hub over its start to end transmission.
The upside of support inhabitance is that clog at every hub is often straightforwardly
and immediately identified. Cradle inhabitance T are often communicated into three
states: typical state, moderate state, and important state as demonstrated in above
figure to acknowledge diverse cushion states, we utilize two unique limits T1 and
T2, as demonstrated in figure Support inhabitance change rate δ is employed to
differentiate the organization clog and may be characterized as:

δi = T c I − T L I Tmax − T c I, I e {I frame, P frame} (1)

in which T c I and T L I, I e{I Edge, P Frame} show the support inhabitance in

current and last round, individually. From above condition, we will get the (Fig. 3)

Fig. 3 Buffer states for congestion detection

Design and Development of Enhanced Secured Video Communication … 89
 
Cushion inhabitance change rate δ = iwi × δi iwi, Ie {I frame, P frame}
(2)

All need lines have an identical support size Tmax . The cradle inhabitance change
rate δ mirrors the inclination of cushion inhabitance. For a given support inhabitance,
the larger the price of δ, the upper the likelihood of line flood is. Also, a negative δ
shows that clog has been lightened and cushion inhabitance is diminished. As per
these two blockage pointers, the sensor hubs have an aggregate of three states.
In the proposed approach, each SH is assumed to be a network connector that
controls several sensor nodes utilizing a distributed clustering technique. The SH
can proceed with an independent decision without the presence of any centralized
management. The clustering approach can be manually exploited by taking the loca-
tion and communication range between the SH and BS into account. The sensor
nodes with minimum energy levels become members in a cluster while remaining
nodes become CHs. The elected CHs are responsible for authentication of member
nodes and receiving and forwarding of data to a BS (Base Station).
Secure Communication between SH and BS.
After electing the SH, the proposed scheme performs the operations are compres-
sion, encryption, and data transmission between the SH and BS. The multimedia
data compression consists of four stages such as transform calculation, thresholding,
quantization, and entropy encoding. The signal transformation is done with wavelet
transform (WT). The Coefficients are quantized by using the uniform step size which
depends on the maximum and the minimum values of signal matrix and the number
of quantization levels. The Huffman encoding can be used for obviating this problem.
Data can be encrypted by using RC7 encryption algorithm. Inverse transform and
decompression are applied at the receiver side to reconstruct the original data.
Secure Communication between BS and Cloud storage.

100
90
80
Buffer occupancy(%)

70
60
50
40
30
20
10
0
2 4 6 8 10
TIme(Minutes)

Fig. 4 Buffer occupancy with IVSP

90 A. Sivasangari et al.

In this phase, a novel lightweight Feistel Cipher structure is presented for encryp-
tion. The key management for the presented work is achieved through the use of
Attribute based Key-Exchange (ABKE).
The BS nodes are preloaded with a set of attributes, which are ID, type of data
and type of users. Each user is assigned an access structure and corresponding secret
key SK. The session key is derived by applying hash value for the attributes, which
are preloaded in BS nodes. The BS collects the information from SH and sends it
to the cloud server. The users who satisfy the access policy are allowed to access to
the patient information. The user derives the session key from their attributes and
decrypts the master key, which is used for deriving the medical information. The key
generation module provides a secure key exchange for ensuring authentication and
is achieved by a trusted third party who is responsible for health care information
monitoring and key distribution.

4 Performance Analysis

To assess entirely unexpected conventions and calculations relating for different

systems administration layers (transport, network layer, MAC, or physical layer)
of remote transmission locator organization or check various applications over the
WMSN, specialists may perform scientific examination, direct tests, or use reen-
actments. Once in a while, Analytical examination neither offers a right model for
such modern remote framework nor truly portrays the conduct of period remote
organizations. Likewise, in a few cases, tests and examinations in remote identifier
network for the most part and in remote transmission indicator network over all
square method some way or another muddled and long, and grave to be re-conducted
by various specialists. WMSN test beds square measure utilized for higher arrange-
ment and fulfilling the reasonable and specialized difficulties of organizations sent in
period frameworks. While testbeds turned into the popular technique for testing and
assessing with remote transmission locator network applications, they furthermore
give proposes that for bunch activity numerous individual sensors on an ordinary
remote stage in an incredibly controlled and instrumented environmental factors.
Subsequently, investigation on test testbeds with current equipment and PC code
stages, licenses clients not exclusively to exhibit pertinence and assess application-
level and organization level execution measurements (Ex. Jitter, streams, detection
likelihood, end to end delay, etc.) in genuine conditions, anyway also to approve
investigation models. Contrasted and leading period trials and field organizations,
testbeds offer wide power in testing without doubt long-lasting analyses that are
indispensable in troubleshooting, approval, and combination periods of dependable
remote transmission indicator organizations. WMSN testbeds are ordered into 2
classes, PC code testbeds and Hardware testbeds. Table about six delineates the
overall PC code and equipment testbeds found inside the writing and sums up their
particulars and fundamental alternatives.
Design and Development of Enhanced Secured Video Communication … 91

Hardware testbeds incorporate passing on of Hardware contraptions, like different

sorts of cameras with absolutely different objectives and picture taking care of capaci-
ties, and remote correspondence equipment that will maintain various rules and abso-
lutely special data rates. Other than that, hardware testbeds offer supporting code for
data insight and program. Dependent upon the class wise organization maintained
and supported by different network, Hardware testbeds are more segregated into
single and multi-tier testbeds.
(a) Software Testbed:
Camera Resolution: WiSNAP, AER Emulator.
Wireless Mote: Added library like Agilent ADCM-1670,OV7649@30fps, 60fps.
Features: Matlab based testbeds, VisualC++ based testbeds, AE recognition.
(b) Hardware Testbed:
Camera Resolution:Meerkat, SenseEye, IrisNet.
Wireless Mote:Logitech QuickCam Pro, PTZ Sony SNC-RZ30N.
Features: Energy efficient, Multi level resolution, Surveillance applications.
Figures 4 and 5 shown about node packets and congestion probability ratio
respectively,

100
IVSP
90 ECODA

80 NoIVSP

70
Packet delivery ratio(%)

60

50

40

30

20

10

0
2 4 6 8 10
Time(Minutes)

Fig. 5 Packet delivery ratio of frame packets

92 A. Sivasangari et al.

5 Conclusion

In this paper, here discussed about the issues and secure challenges in WMSNs.
As well as examined overviewed and detailed design about WMSNs and also
uses explained clearly. In this paper added security challenges based on criteria
also analyzed detailed assessment carried on WSMS (Wireless Multimedia Sensor
Networks) and investigation challenges and their issues in arranging estimations,
shows, plans, and gear to wireless sensors. And here discussed about enormous
segment of the current responses for WMSN at the different layers of the correspon-
dence stack: physical, MAC, coordinating, transport, and application close by the
possible cross layer execution. Feistal light weight cipher was used for encryption
and decryption.

References

1. Newaz AI, Sikder AK, Babun L, Uluagac AS (2020) HEKA: a novel intrusion detection system
for attacks to personal medical devices. In: 2020 IEEE conference on communications and
network security (CNS), Avignon, France, pp 1–9
2. Liu Y, Cai H, Li G (2018) Research on sharing and application of medical big data. In: 2018
3rd international conference on smart city and systems engineering (ICSCSE), Xiamen, China,
pp 771–774. https://doi.org/10.1109/ICSCSE.2018.00166
3. Quantin C, Allaert FA, Fassa M, Avillach P, Fieschi M, Cohen O (2007) Interoperability issues
regarding patient identification in Europe. In: 2007 29th annual international conference of the
IEEE engineering in medicine and biology society, Lyon, France, pp 6160–6160
4. Sumalan T, Lupu E, Arsinte R, Onaca E (2015) Low-cost wireless system for patient monitoring.
In: 2015 E-health and bioengineering conference (EHB), Iasi, Romania, pp 1–4
5. Yifan K, Peng J (2008) Development of data video base station in water environment monitoring
oriented wireless sensor networks. In: 2008 international conference on embedded software
and systems symposia, Chengdu, China, pp 281–286
6. Sodhro AH, Fortino G (2017) Energy management during video transmission in wireless body
sensor networks. In: 2017 IEEE 14th international conference on networking, sensing and
control (ICNSC), Calabria, pp 655–660
7. Singh JP, Mishra MK, Khan MA (2017) Energy-efficient approach towards video-based
sensor networks (wireless) beneath barrier coverage. In: 2017 8th international conference
on computing, communication and networking technologies (ICCCNT), Delhi, India, pp 1–5
8. Abbas N, Yu F (2018) Design and implementation of a video surveillance system for linear
wireless multimedia sensor networks. In: 2018 IEEE 3rd international conference on image,
vision and computing (ICIVC), Chongqing, China, pp 524–527
9. Qian C, Cai W (2014) The design and implementation of video wireless WIFI sensor nodes
with the function of air targets detection. In: 2014 international conference on wireless
communication and sensor network, Wuhan, China, pp 222–225
10. Jasim A, Çeken C (2015) Video streaming over wireless sensor networks. In: 2015 IEEE
conference on wireless sensors (ICWiSe), Melaka, Malaysia, pp 63–66
11. Mana SC, Keerthi Samhitha B, Jose J, Swaroop MV, Reddy PCR (2019) Traffic violation
detection using principal component analysis and viola Jones algorithms. Int J Recent Technol
Eng (IJRTE) 8(3). ISSN: 2277-3878
12. Ramamoorthy V, Divya S, Mana SC, Keerthi Samhitha B. Examining and sensing of artificial
knee with multi sensors networks. J Adv Res Dyn Control Syst (JARDCS) 10(4):115–120
Design and Development of Enhanced Secured Video Communication … 93

13. Yi JM, Yoon I, Noh DK (2017) Adaptive video coding selection scheme for solar-powered
wireless video sensor networks. In: 2017 4th international conference on computer applications
and information processing technology (CAIPT), Kuta Bali, Indonesia, pp 1–4
14. Raj C, Jain C, Arif W (2017) HEMAN: health monitoring and nous: an IoT based e-health care
system for remote telemedicine. In: 2017 international conference on wireless communications,
signal processing and networking (WiSPNET), Chennai, India, pp 2115–2119
15. Jamshed MA et al (2015) An energy efficient priority based wireless multimedia sensor node
dynamic scheduler. In: 2015 12th international conference on high-capacity optical networks
and enabling/emerging technologies (HONET), Islamabad, Pakistan, pp 1–4
Detecting Twitter Hate Speech Using
Sentiment Analysis

Arpita Tripathy, Anshika Goyal, Urvashi Tyagi, and Poonam Tanwar

Abstract In current era due to the rapid increase in the consumption of the Internet
by people of different cultural backgrounds, malicious content on the Internet has
become an endless problem. In the automatic detection of malicious text content,
distinguishing between hate speech and profanity will be a major issue. The police
investigating strategies for positive or negative emotions in the text. Companies use it
extensively to perceive emotions in social data, measure overall reputation, and under-
stand customers. For example, if you use sentiment analysis to mechanically analyze
more than 4000 reviews related to your products, you can more easily determine
whether customers are satisfied with your customer service and pricing plans. The
Twitter information set is usually used for comparative analysis of the model. Infor-
mation science provides useful data based on large amounts of complex information
or big data. Information science/engineering combines different fields of statistics
and computer science to interpret information for decision-making purposes. The
purpose of this article is to provide the method for detecting twitter the speech using
Support vector machine and Machine Learning algorithm.

Keywords Sentiment analysis · Machine learning (ML) · Support vector machine

(SVM)

1 Introduction

The global availability of the Internet has changed the way we greatly perceive
the world. The social network (SM) is one of the youngest representatives of the
global network, and it is also a gift in various forms: online gaming platforms,
chemical analysis applications, forums, online news services, and social networks [1,
2]. Completely different social networks represent different goals: opinion delivery

A. Tripathy · A. Goyal · U. Tyagi · P. Tanwar (B)

Department of CSE, MRIIRS, Faridabad, India
e-mail: poonamtanwar.fet@mriu.edu.in
Faculty, CSE, FET, MRIIRS, Faridabad, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 95
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_8
96 A. Tripathy et al.

(Twitter or Facebook), business contact (LinkedIn), image exchange (Instagram),

video delivery (YouTube), chemical analysis (Meetic), etc. [5].
Among the diverse existing social networks, Twitter presently ranks with inside
the live performance of the leading structures and is one of the predominant crit-
ical information reasserts for researchers. Twitter can be a famous real-time public
microblogging community where, frequently, the information seems earlier than then
on legitimate fourth estate. Characterized via way of means of its brief message limit
(now 280 characters) and unfiltered feed, its utilization has fast escalated, specifically
amid events, with a median of 500 million tweets published in line with day.
In current era, social media (mainly twitter) can be used to spread hate
posts/messages. Hate post/speech refers to a form of communication that makes
one or more people who maintain their own sense of group belonging to the nose. It
is mostly presented by race, sexual orientation, personality, disability, etc., religion,
political affiliation, or opinion [14].

1.1 Twitter

Twitter can be a micro blog site can be used to share their posts in the form of
opinions, experiences, and feelings using text messages (tweets). Tweet on behalf
of users from various desktop and mobile clients and even other applications. Who
posted the tweet [1–4]. Each user has unique identity their name and an identification
number. Users can have followers; get the latest information about what users have
posted. Other users can view, share, or in other words, repost the content of any
posted tweet [1–4].
Data Retrieval
Tweets and related information (such as username, posting date and location etc.,
are usually retrieved from Twitter through several API. Access via authentication
request [1, 2, 4, 10]. If you request access to one or more APIs, you will need to
submit credentials issued by Twitter. To receive them, you need to register a request
on Twitter [21].
• Rate Limits
Streaming API allows to send 5000 user IDs at the same time [11]. The duration of
the API rate limit window is 15 min [12]. A user represented by an access token can
make 180 requests/requests [11] per time window. Submit 450 enquiries/inquiries
on your behalf every quarter.
Detecting Twitter Hate Speech Using Sentiment Analysis 97

1.2 Data Cleaning

Cleaning was performed with some iterations of regex syntax to induce obviate re-
tweets, handles, and special characters. Duplicate tweets were also removed and
lemmatization with a part of speech was also applied to any or all tweets. The last
stage involved removing stop words and also words shorter than three characters as
they are doing not usually carry very valuable meaning. However, stop words like
‘but’ and ‘not’ were kept for neural networks sequencing. We created two additional
columns, ‘tweet_length’ and ‘handle_count’ to analyze whether these two factors
have any impact on positive/negative language [5].

1.3 Text Mining

Intelligent text analysis is a method to retrieve the patterns and their relationships
in a collection of unstructured data [9]. Text analysis is usually divided into three
stages: pre-processing: delete, export, and mark stop words. Knowledge extraction-
use machine learning tools to find hidden dimensions. And relationships that are
difficult to detect manually.
Eliminating the ignored words will delete meaningless words, and they often
appear in every text, such as B. articles and pronouns. Derivatives help to express
features more densely by shortening the roots of curved words. Tokenization removes
punctuation marks and breaks down sentences into simple combinations of words
and letters. After the mandatory pre-processing is completed, statistical methods are
used to analyze the content of the text to determine its attributes [2, 3]. Text data can
be processed in different ways and at different levels.

1.4 Exploratory Data Analysis

The EDA section provided some useful insights into the very fabric of the words
utilized in these tweets. Word clouds were created to showcase the foremost common
1, 2, and 3 g present within the text. Attaching below a comparison between positive
and negative lexicon, larger words correspond to a higher frequency [1–4, 17, 19].
The duration distribution between positive and negative tweets is also analyzed,
as shown in Figs. 2 and 3 [19], negative tweets are on average shorter than positive
tweets. 1 represents negative tweets and 0 represents positive tweets as represented in
Fig. 1. A simple t test confirms that when the p-value is less than 001, the difference
in means is important.
In previous section, the association between the controls and aggressiveness is
obtained by sketching the number of positive/negative results against the total number
98 A. Tripathy et al.

Fig. 1 Negative tweets and positive tweets

Fig. 2 Graph showing positive and negative tweets per length [19]

Fig. 3 Graph showing offensive language and number of handles [19]

of controls. Mainly, tweets contain 0–3 controls whereas 0 and 1 handler tweets. One
can explain this by the fact that a curator can direct small talk to someone.
Detecting Twitter Hate Speech Using Sentiment Analysis 99

1.5 Modeling

The only predictor variable we use for modeling is the text itself, and we use the
TFIDF method to vectorize a refined version of the text. TFIDF is more popular than
word data packets, because in this case, the rarity of words has a certain meaning
[7, 8]. Except for CNN (which only accepts token string input) and Naive Bayes, all
models use the TFIDF array. This is where we tried to use the arched structure to
see if power is needed. The performance of CNN is not added to the figure below,
but the performance is very slow, only exceeding 50% accuracy. The neural network
part requires a different job. You can try to implement RNN on an existing CNN at a
price lower than the standard cost. As mentioned above, the most effective model is
logistic regression with f-1 score. The test suite is 98.2%. Logistic regression is also
very useful for describing the meaning of each word in the prediction result [15–17].

1.6 Twitter API and MySQL Storage

New tweets are periodically collected by making the request access to the Twitter
portal in the Realm MySQL database, which is only created for the incoming tweet
stream [5]. The Python Tweepy library was not used to create the Twitter API link.
How to restore SQL database from unformatted JSON stream:
• Twitter ID, Time and Tweet Text.

1.7 Model Evaluation and Dashboard

The basic pipeline was developed by dragging and dropping recently uploaded tweet
packets onto a separate laptop and placing them where the exact same pre-cleaning
was applied (Fig. 4).
In addition, the final model and the objects selected by tf can be retrieved in
fresh_tweets_df.ipynb, and the file can also be found in the ‘notebook’ section of
the repository. 9000), and the dictionary of words is simply discarded, and the rest is
left in the matrix. Then the probability prediction function can be used on vectorized
matrix and greater than 0.9 are filtered to collect only the tweets that the model finds
to be very offensive. As the last step, the new forecast data frame will be uploaded
to another repository, which is located. All heroku implementations are sorted here,
including creating configuration files and ‘tweetresources.txt’ files. Over time, tweets
that mention the topic will appear in the final application, the most repeated words
and the ratio of negative to neutral/positive tweets [21–23].
100 A. Tripathy et al.

Fig. 4 Twit pipeline [19]

2 Existing Work and Problems Identified

The Internet allows you to easily access and share information at super fast speeds.
This set of anonymous functions also makes it an effective means of distributing
hateful or offensive content. In addition, many competitions are held to solve
similar problems (e.g., publicly insulting, offensive or hateful/objectionable content
in German, Spanish, or English).The basic method is to use a simple template. The
most common method is a combination of feature extraction and the problem oriented
machine learning algorithms [1–5]. The BoW (bag of words) method is used in
combination with the naive Bayes classifier. Use BoW with Support Vector Machine
(SVM). However, due to the many false positives of BoW where as other researchers
have applied more complicate methods to provide the functionality of classic cars.
Other deep learning techniques in this field include the use of convolution neural
networks (CNN), recurrent neural networks (RNN), and deep learning. Another
important step is the introduction of converters, especially the BERT converter,
which provides seven. One of the ten most effective models in the subtask. In fact,
this method has recently performed best in competitions between more than fifty
participating teams (based on the average performance of all subtasks) [11–15].
The limitations of the method used are choices that are often difficult to under-
stand, and it is difficult for people to explain the reasons for making these decisions.
This is usually a practical problem, as systems that automatically review personnel
voices may require manual complaint procedures. A new method of categorizing
hate speech can better understand the selection content and show that it can even
outperform existing methods for certain data sets. This may be effective, but these
sources need to be maintained and updated so that you know that it may be a problem
[23, 24].
Detecting Twitter Hate Speech Using Sentiment Analysis 101

3 Proposed Work and Research Method

The data pre-processing stage includes two methods: Bag of words (BoW) and Term
Frequency Inverse Document Frequency (TFIDF) [1, 2].
The bag-of-words technique is often a simplified illustration employed in language
process and data retrieval. In this method, due to its word pocket(s), text such as
sentences or documents is displayed, regardless of grammar and uniform order,
while maintaining the plural.
TFIDF can be a number of statistics whose purpose is to reflect the meaning of
words in documents in an overwhelming collection. The use of weights considers
knowledge search, text mining, and user-defined modeling [1, 2].
Before we archive computer files using various algorithms, we will clean them up
because these tweets contain a lot of tenses, grammatical errors, unknown characters,
# tags, and Greek characters [2].
Problem can be solved using abbreviations, stop word deletion, and skipping.
Tonalization eliminates the inflection of the word ending and restores the word to its
basic or lexical form. Keep promises [2, 20–24]
Following methods are used for data.
• Eliminate stop words: they can be articles or prepositions. These words can be
deleted without any negative impact on the ending. A feasible model. The words
‘es’, ‘is’, etc., are usually used in West Germanic.
• Greek characters or special character deletion.
• Slang Language words deletion: example: ‘luv’, ‘thnx’, etc.
• Stemming: Through stemming, a word is abbreviated to the word root, so that
the uppercase or beginning of the word is separated. Consider the list of common
prefixes and suffixes contained in words with variable angles. Determine how the
conditions should be reduced. Porter’s algorithm is used in our data set.
• Lemmatization: Lemmatization is a method of categorizing different types of
phrases in a word so that they can be analyzed together. Therefore, he connects
words that are roughly similar to the words. The word ‘run’, so ‘run’ is the motto
of these words.
Main classification algorithms like Decision Tree, Logistic Regression, Random
Forest (RF) etc., can be used for bag of words and TFIDF.
In order to spot hateful tweets machine learning algorithms works well. 2 use
cases were used to implement the system.
1. Positive tweet so as to properly classify it as a positive tweet.
2. Negative tweet so as to properly classify it as an annoying tweet.
After providing several labeled coaching SVMs for every category, we are able
to classify the new text. Next, we will contend with the text classification problem.
We will refine our training data and even try to fill it. Then, we continue to use SVM,
which is a fast and powerful classification algorithm that can handle the limited
amount of information that needs to be checked. The main goal of the project is to
102 A. Tripathy et al.

Fig. 5 Flowchart showing

proposed system

determine the hate speech, which will be a binary classification task. It is to classify
that every sample into a non-hate speech or hate speech category. We estimate the
likelihood of ranking each post based on the likelihood that each post contains hate
speech. Therefore, we will test some of the most popular software and choose the one
that works best. We will choose the naive Bayesian method because it can classify
spam well and can similarly identify hate speech. You choose SVM and RF because
they work well in the most difficult tasks (Fig. 5).

4 Results and Observation

The dataset consists of 20 K posts from Gab and Twitter. Each data point is annotated
with one of the hate/offensives/normal labels, target communities mentioned, and
snippets (rationales) of the text marked by the annotators who support the label. The
proposed work has been analyzed using Kaggle record where a.csv file containing
31,962 tweets. The data set is severely skewed: 93% of Twitter data does not contain
hate tags, while 7% contains Twitter data with hate tags.
Detecting Twitter Hate Speech Using Sentiment Analysis 103

The functions that cannot be called after data deletion will be called first. Those
are:
1. Number of empty words
2. Pound signs used
3. Characters used
4. Number of uppercase letters.
After deleting the text, some features are extracted because they are more
meaningful at this time. These are:
1. Word count: This feature indicates how many words are in the revision.
2. Number of characters: How many letters does the revision contain?
3. Average word length: The average number of letters in a word during over scan.
Although the performance of the final model is very good even when testing our
data set, one of the biggest limitations of the project is that the performance of the
model is measured in a new tweet. From a pragmatic point of view, people can simply
view a few tweets that have been marked as negative, and subjectively regard them as
unwelcome. In this context, this latest view raises another important question related
to the inherent tendency of individuals who manually flag tweets.
The judgment also tried to incorrectly define the original basic facts, because
offenses to some people might not be the case to others. The wordless dictionary
may be one of the main shortcomings of this model. There are 9 million words in our
original vocabulary. When it comes to word processing outside of the vocabulary, a
recurrent neural network may be the easiest choice.

5 Conclusion

Cyber bullying and violent speeches on social platforms are some of the shortcomings
of our time. Freedom of expression on the Internet can easily be reduced to insulting,
unfounded and unconstructive criticism of sex, political, and religious beliefs. Both
machine learning and the large amount of data available on social platforms are
beneficial. An effective solution to alleviate this problem. We use SVM, which is a
very simple but powerful supervised machine learning algorithm that can be used
for classification and regression. It is suitable for small and medium data sets and is
very easy to set up.

References

1. Pitsilis GK, Ramampiaro H, Langseth H (2018) Effective hate-speech detection in Twitter data
using recurrent neural networks. Appl Intell 48:4730–4742. https://doi.org/10.1007/s10489-
018-1242-y, https://www.researchgate.net/publication/326517890
104 A. Tripathy et al.

2. Gaydhani A et al (2018) Detecting hate speech and offensive language on Twitter using machine
learning: an N-gram and TFIDF based approach. In: IEEE international advance computing
conference 2018. https://arxiv.org/pdf/1809.08651.pdf
3. Watanabe B, Ohtsuki (2018) Hate speech on Twitter: a pragmatic approach to collect hateful and
offensive expressions and perform hate speech detection, 15 February 2018, pp 13825–13835.
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8292838
4. Zimmerman, Kruschwitz (2018) Improving hate speech detection with deep learning ensem-
bles. In: Proceedings of the eleventh international conference on language resources and
evaluation (LREC 2018), pp 2546–2553. https://www.aclweb.org/anthology/L18-1404.pdf
5. Pereira-Kohatsu JC (2019) Detecting and monitoring hate speech in Twitter. Published in
Special issue sensors for affective computing and sentiment analysis. Sensors 19(21):4654.
https://doi.org/10.3390/s19214654
6. Ziqi et al (2018) Hate speech detection using a convolution-LSTM based deep neural network.
In: Proceedings of ACM the web conference (WWW’2018). ACM, New York, NY, USA, p
10. Article 4, https://doi.org/10.475/123
7. MacAvaney S et al (2019) Hate speech detection: challenges and solutions. PLOS One 14(8).
https://doi.org/10.1371/journal.pone.0221152, https://journals.plos.org/plosone/article?id=10.
1371/journal.pone.0221152#sec019
8. Salminen J et al (2018) Anatomy of online hate: developing a taxonomy and machine learning
models for identifying and classifying hate in online news media. In: Proceedings of the twelfth
international AAAI conference on web and social media (ICWSM 2018). https://ojs.aaai.org/
index.php/ICWSM/article/view/15028/14878
9. Lee K, Palsetia D, Narayanan R, Patwary MMA, Agrawal A, Choudhary A (2011) Twitter
trending topic classification. In: 2011 IEEE 11th international conference on data mining
workshops (ICDMW), December 2011, pp 251–258. https://doi.org/10.1109/ICDMW.201
1.171
10. Chih-Chung (2021) A library for support vector machines, 14 April 2021. https://www.csie.
ntu.edu.tw/~cjlin/libsvm/
11. Kumar S, Morstatter F, Liu H (2013) Twitter data analytics. Springer, New York, NY, USA
12. Kamal O et al (2021) Hostility detection in Hindi leveraging pre-trained language models. In:
CONSTRAINT 2021. https://arxiv.org/pdf/2101.05494v1.pdf
13. Leite JA et al (2020) Toxic language detection in social media for Brazilian Portuguese: new
dataset and multilingual analysis. In: IJCNLP 2020. https://arxiv.org/pdf/2010.04543v1.pdf
14. Mollas I et al (2020) ETHOS: an online hate speech detection dataset. In: CIKM2020. https://
arxiv.org/pdf/2006.08328v1.pdf
15. Raghad, Hend (2020) A deep learning approach for automatic hate speech detection in the
Saudi Twittersphere. In: Proceedings of the fifth Arabic natural language processing workshop,
December 2020, pp 12–23. https://www.aclweb.org/anthology/2020.wanlp-1.2.pdf
16. Mathew B et al (2021) HateXplain: a benchmark dataset for explainable hate speech detection.
Assoc Adv Artif Intell. www.aaai.org, https://arxiv.org/pdf/2012.10289v1.pdf
17. Twitter Inc. Twitter developers. Accessed on 18 October 2019. 2019 Available online https://
developer.twitter.com/en.html
18. Fan L et al (2020) Stigmatization in social media: documenting and analyzing hate speech
foCOVID-19 on Twitter. Proc Assoc Inf Sci Technol 57(1). https://doi.org/10.1002/pra2.313.
Epub 22 October 2020. https://asistdl.onlinelibrary.wiley.com/doi/10.1002/pra2.313
19. Tortella (2020) How to build a (quasi) real-time hate speech classifier for Twitter, June 2020
20. Vedanshu et al (2020) Live Twitter sentiment analysis. In: Proceedings of the international
conference on innovative computing and communications (ICICC), May 2020. Available at
SSRN https://ssrn.com/abstract=3609792 or https://doi.org/10.2139/ssrn.3609792
21. Tanwar et al (2016) A tour towards sentiments analysis using data mining. Int J Eng Res Technol
5(12)
22. Tanwar, Rai (2020) A proposed system for opinion mining using machine learning, NLP and
classifiers. IAES Int J Artif Intell (IJ-AI) 9(4):726–733. ISSN 2252-8938. https://doi.org/10.
11591/ijai.v9.i4.pp726-733
Detecting Twitter Hate Speech Using Sentiment Analysis 105

23. Kaustav et al (2020) A facial expression recognition system to predict emotions. In: 2020
international conference on intelligent engineering and management (ICIEM), June 2020
24. Tanwar P, Priyanka (2019) Spam diffusion in social networking media using latent
Dirichlet allocation. Int J Innov Technol Explor Eng (IJITEE) 8(12). ISSN 2278-3075
25. Tanwar, Prasad, Dutta (2018) Natural language processing for hybrid knowledge representation.
Int J Adv Paradigms 10(3)
Comparative Evaluation of Semantic
Similarity Upon Sentential Text of Varied
(Generic) Lengths

Richa Dhagat, Arpana Rawal, and Sunita Soni

Abstract Semantic similarity plays a vital role in natural language processing,

information retrieval, text mining, question and answering system, text-related
research, biomedical informatics, and plagiarism checking. Measuring similarity
between sentences means estimating the degree of closeness in meaning between
these sentences. Very recently, NLP research has observed much-inclined shift from
word-lexicon and syntactic feature-based techniques to structural feature-based tech-
niques of exploring NLP semantics. Rather, no stone is unturned by NLP community
working on computation of text semantic similarity to formulate supporting feature
spaces, say, word-to-word co-occurrences, lexical corpora and natural language
grammar rule-based word order vectors as well as semantically parsed structures.
The most popularly used WordNet lexical dictionary was exploited as hierarchical
structures (semantic trees) to fetch all combinations of noun and verb families of word
phrases, predominantly occupying free texts. In this paper, the computation of text
semantic similarity is addressed by devising a novel set of generic similarity metrics
based on both, word-sense of the phrases constituting the text as well as the grammat-
ical layout and sequencing of these word phrases forming text with sensible meaning.
The experiments performed on benchmark datasets created by Li and group explore
semantically parsed dependency structures in feature extraction step and achieve
very promising values of semantic closeness within sentence pairs. The model’s
performance is indicated achieving highest value of Pearson’s correlation coefficient
(0.89) with mean-human similarity scores against those obtained through closely
competent structured approach models. However, these very promising results are
contributed by word-sense similarity component of sentential similarity measure.
A future scope of work is also incepted that shall also improve upon dependency
(grammatical relations) component to raise the text similarity performance metric.

Keywords Natural language processing · Semantic similarity · Structured

approach · Similarity matrix · Lexical dictionary

R. Dhagat (B) · A. Rawal · S. Soni

Department of Computer Science and Engineering, Bhilai Institute of Technology, Durg, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 107
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_9
108 R. Dhagat et al.

1 Introduction

Advances in NLP, preferably combining the field of computational linguistics, along

with a large amount of text-based resources nowadays are publicly available in digital
formats (e.g., online encyclopedias). These repositories allow newer NLP prototype
or commercial tools to become more reliable and robust in accomplishing many
natural language applications such as semantic search, summarization, question
answering, document classification, sentiment analysis-cum-classification, plagia-
rism detection tasks and much more while performing machine learning tasks on
Web and image datasets. All these are possible only if syntax-semantic-based intel-
ligent techniques are employed to capture the meaning, concept and idea revealed
within the text content. Diversified school of thoughts have been triggering various
approaches and algorithms for over nearly two decades to find robust methods to
arrive at precisely effective semantic similarity measures, and the work is been
pioneered in the early decade of 2000 by Erkan and Radev [1], Li et al. [2], Mihalcea
[3], Miranda [4]. Their works invariably differed in using feature spaces for learning
the textual pattern similarities and machine learning approaches. In addition, they
differed in using computational expressions for measuring text similarities that were
used for varied learning (classification labeling or grouping) tasks with their own
decided thresholds settings.
A major breakthrough in recently pursued research works on text similarity was
unanimously agreeable opinion of representing the whole text document as sentences
posing to reflect the actual meaning of the text in itself or augmented by adjacently
lying sentences. By analyzing such works, we arrived at some of the major findings
outlined as follows:
• Computations on word vectors, in any combinations, whether POS-tagged or not,
were very computationally expensive;
• The conventional techniques were highly domain or corpus dependent, while a
lot effort was being attempted to arrive at generic approach of semantic similarity
computation;
• There was a need to adopt a robust method that maps the correctness of word
order and validates the word-sense borne by word (phrases) in and around the
sentences reflecting a theme or topic or idea or context narrated through.
All the above-mentioned observations were welcomed and drilled into by NLP
research community as they could see sentential similarity approach of text mining,
easily invading into sufficiently higher information retrieval effectiveness in appli-
cation realms like web page retrieval, image caption retrieval from web, web page
ranking and document summarization from the topic-relevant web pages at hand.
The paper is outlined into sections as follows: Sect. 2 presents the related
research carried out using structural features of free text, albeit, various research
groups explored these structural features with varied research objectives pertaining
to different realms. Section 3 outlines the refinements in the methodology undertaken
to compute word-sense similarity and grammatical similarity by taking a sample
Comparative Evaluation of Semantic Similarity Upon … 109

instance of sentence pair. Section 4 tabulates the experimental computations for a

finite set of sentence pairs upon which results are already available due to approaches
already dealt in the past. The concluding remarks on the result comparisons are
presented in Sect. 5.

2 Semantic Similarity Computations: A Survey

2.1 Structured Approaches: State-of-the-Art

As rightly observed by Alzahrani and group [5, 6] that extraction of lexical, syntactic
and semantic features from the experimental text is governed by some or the other
pre-defined list of lexical dictionaries (databases) like some commonly available
ones from the English language corpora are WordNet [7], Brown Corpus, Wikipedia
Rewrite Corpus [8], METER corpus [9] and Webis Crowd Paraphrase Corpus [10].
In reality, these external knowledge resources may not account for capturing the
thematic sense reflected in that text. Hence, they attempted to explore the structural
features in two different perspectives, namely block-specific and content-specific
textual features. There have been several works that have attempted to compute
inter-document similarities by representing them into hierarchical levels of para-
graphs, sentences and word phrases. It was observed by NLP research community
that much more promising results were obtained when they performed text similarity
experiments upon these structural feature space representation types.
The work done by Li et al. [2] claims to have devised hybrid approach to text
semantic computation method based on semantic similarity (using cosine metric)
over the information content associated with each participating sentence of the
sentence pair. However, their work leaves an element of doubt unresolved, as
initially they formulate the logic stating that their method does not have any domain-
dependency issues, while, their experiments did make use of pre-compiled dictio-
naries, namely, WordNet for computing semantic similarity and Brown Corpus for
calculating information content values.
Meanwhile, Islam and Inkpen [11] attempted to formulate a variant text semantic
similarity model by incorporating string, word co-occurrence and optimal word order
matching steps, however, augmented by enormous time consumption in constructing
intermediary joint semantic word similarity matrices. Lee’s [12] contribution differs
from the above piece of work in the sense that the former emphasizes on corpus
dependent word-to-word association weight (information content), while the latter
supports the usage of distance-based (wup) similarity metric to arrive at overall
semantic similarity of sentence pairs. Also, Croft et al. [13] replicated the work done
by Li et al. by simply using cosine vector similarity over formulated weighted word
vectors, where weights on these vectors were computed as an outcome of WordNet-
based path similarity measure. The candidate texts were only caption and header titles
and were not valid grammatically constructed sentences. Taieb et al. [14] proposed
110 R. Dhagat et al.

measures that aggregate noun and verb fragments together along with common word
order parameter to obtain overall sentence similarity metric. The work extends the
concept of other related works by also including matching of modality (in form of verb
tenses) between candidate pair of sentences along with compound nouns. Although
Pawar and Mago [15] proposed almost similar approach as Li et al. that considers both
semantic and syntactic structures for text similarity using lexical knowledge base,
their work distinguishes with others in the sense that they have used totally novel
similarity metric. Vani and Deepa [16] have borrowed Li and Lee’s contributions in
hybridizing the text semantic similarity metrics with a variation in formulating logic
for second contributive metric (sim2 ) in computing over all sentence similarity. Here,
the word-to-word association weight is replaced by proportionate count of common
(matched) words (or allied synsets) found in the sentence pairs. Such related works
motivated our research group to investigate further using multiple combinations of
both syntactic and semantic similarity metrics and pursue comparative studies, if
any.

2.2 Structure-Based Approach Revisited

Ozates et al. [17] claim to pioneer the exploring of a novel structure-based approach
based on grammatical relations among all combinations of participating words in the
sentence pairs. However, their work restricted to usage of only those grammatical
(dependency) relations contributing to predominantly noun and verb family of word
phrases and ignoring unimportant dependency relations, for instance, prep (prepo-
sitional), punct (punctuation), det (determiner) and possessive and many more, as
stated by them. They emphasized on the merits of using dependency tree structures
as they reflect the neighboring (may be adjacent or non-adjacent) word-senses within
individual sentential units better than word vector-based or syntactic-based feature
representation of texts. They used dependency grammatical structures in bigram
structured formats as textual feature spaces in order to compute sentence semantic
similarity. Each bigram structure is comprised of a dependent word, a head word
and a typed-dependency tag expressing the type of relationship between them. The
consequent semantic similarity measure was expressed in terms of Kernel expression
formulations, namely simple approximate bigram kernel (SABK), TF-IDF-based
approximate bigram kernel (TABK) and its variants. The sentence-similarity metric
assigns weights to the grammatical relation tags using degrees of grammatical rela-
tion tag overlap. Also, weights to word phrases are assigned by computing tf-idf
measures upon head as well as dependent arguments.
Wali et al. [18] also claims to develop a hybrid approach of computing sentence
similarity upon dependency-parsed structural features extracted from sentence pairs.
They name it as syntactico-semantic similarity based on thematic roles and semantic
classes of distinct as well as common words and their synonymous equivalents, been
matched using Jaccard coefficient metric.
Comparative Evaluation of Semantic Similarity Upon … 111

Under this paradigm, Vakare et al. [19] have devised a novel metric to compute
sentence similarity using dependency parsing. Their work attempts to learn grammat-
ical tag relations by training the similarity scores on pre-defined datasets. This leaves
a query as to what will be the learnt weights, if at all, the background corpus gets
changed and indicates an element of uncertainty aroused due to domain-dependency
nature of the undertaken problem objective.
The authors appreciate much closer work done very recently by Zhang et al. [20]
who have proposed a sentence similarity computation model which uses a hybrid
approach combining both syntactic-cum-semantic similarity between sentence pairs
using grammatical dependency relations obtained from an appropriately selected
semantic NL dependency parser. The sentential semantic similarity metric is named
as kernel function that itself is expressed in terms of another (corpus-based) semantic
‘wup’ metric. However, the filtering out step of some of the dependency rela-
tions: {‘det’, ‘expl’, ‘goes with’, ‘possessive’, ‘preconj’, ‘prep’, ‘punct’, ‘ref’} as
unimportant ones need further justifications.
Farouk [21] distinguishes its work from Zhang’s by formulating sentence graphs
as structural features obtained from grammatical relations of parsed dependencies.
Consequently, the semantic similarity component of overall text sentence similarity
is computed through these semantic relation matrices.
After undertaking an exhaustive survey upon the above-mentioned similarity
metrics used in both unstructured and structured approaches of semantic similarity
computation, it was found that NLP researchers, who explored inter-sentential gram-
matical relations for feature space representations of sentential datasets, obtained
competently better results in semantic similarity computations than those who did
not use these structures.

3 Methodology

In this paper, we proposed a refined similarity score between two input sentential texts
with a generic perspective to use this similarity for any machine learning task in NLP
domain. In order to keep the configuration settings to the problem objective simple,
we continue to borrow thematic (word) sense from the most popularly used WordNet
(lexical database of semantic relations between words) dictionary. The methodology
adopted in this paper can be considered as a refinement to the approach followed by
Lee [12]. The current work chooses Li’s benchmark datasets [22] following the same
protocol of NLP scientists who have been working in this realm. It may be noted
that the previous works had computed the noun vector by comparing each noun of
the candidate sentence with the union of noun word-sets from both the sentences
of sentence pair. The same applies for verb vector computation, making the steps
more cumbersome and confusing. We have straight forwardly compared noun word-
sets of candidate sentences with one another. Similar approach is repeated for verb
word-sets of the candidate sentences.
112 R. Dhagat et al.

3.1 Semantic Similarity Computation (Phase I)

We investigate the importance of the proposed similarity by a simple example. Let

us consider a pair of sentences. Let SA = ‘A cemetery is a place where dead people’s
bodies or their ashes are buried.’ SB =‘A graveyard is an area of land, sometimes
near a church, where dead people are buried.’
Step 1: The words in each sentence after part-of-speech tagging are categorized
into noun and verb sets. (N A , N B ), (V A , V B ) are noun and verb sets of sentence A
and sentence B, respectively. In this example, N A = [‘cemetery’, ‘place’, ‘people’,
‘bodies’, ‘ashes’] and N B = [‘graveyard’, ‘area’, ‘land’, ‘church’, ‘people’]. V A =
[‘buried’] V B = [‘buried’]. ‘m’ and ‘n’ are count of noun sets in vector N A and N B .
‘p’ and ‘q’ are count of verb sets in vector V A and V B .
Step 2: For noun pair of each candidate sentence, (nA , nB ) in the noun sets ∀i, j :
n i A N A , n j B N B , we construct a semantic similarity matrix of dimensionality size, (m
× n) such that if niA = = njB , path_similarity = 1 else extract synset lists for niA and
njB from WordNet lexical corpus. Similar steps of computation ply for all verb pair
(vA , vB ) in the verb sets ∀i, j : vi A V A , v j B VB belonging to sentences A and B. In this
way, two synset lists are formed N A_syn = synsets (niA , POS(niA )), N B_syn = synsets
(njB , POS(njB )); hence, there shall be four synset lists for each candidate word pair
denoted by: N A_syn , N B_syn , V A_syn and V B_syn .
Step 3: Semantic similarity matrix of dimensionality size: (p × q) between V A
and V B is computed similar to semantic similarity matrix obtained between N A and
N B of dimensionality size: (m × n). For example, pair of sentences stated that the
semantic similarity matrix ‘S’ is denoted as follows (Fig. 1).
Step 4: This step details out the sequence of calculations that are needed
to arrive at each of the synset-similarity values S_sim that form the semantic
similarity matrix of the two sentences as a whole. This value is calcu-
lated as the maximum of the path similarity among all synset pairs of
both, participating noun and verb word-sets. For instance, let synset lists
of the word pair {niA = ‘church’, niB = ‘ashes’} be [Synset(‘church.n.01’),
Synset(‘church.n.02’), Synset(‘church.n.04’), Synset(‘church_service.n.01’)]and
[Synset (‘ash.n.01’) Synset(‘ash.n.02’), Synset(‘ash.n.03’)], respectively. The synset
pair [Synset (‘church.n.04’), Synset (‘ash.n.01’)] provides the maximum path simi-
larity (S_sim) out of all synset pairs as ‘0.1’. Similarly, the S_sim value for word
pair (‘area’, ‘place’) is ‘0.5’ (Fig. 2).

Fig. 1 5 ✕ 5 Semantic similarity matrix (for noun vectors)

Comparative Evaluation of Semantic Similarity Upon … 113

Fig. 2 Semantic noun vectors from noun vector semantic similarity matrix

Step 5: Finally, semantic vectors SvecN A , SvecN B are computed from semantic
similarity matrices for both noun and verb sets of the two sentences. The vectors seek
the maximum of the synset_similarity values in order of row and column dimensions
to obtain the two vectors; in our example, SvecN A = [1.0, 0.5, 1, 0.5, 0.166] and
SvecN B = [1.0, 0.5, 0.5, 0.5, 1], SvecV A = [1] and SvecV B = [1]; the calculations
of example pair of sentences are shown in the above figure.
Step 6: As a result, semantic noun and verb portions of similarity scores can be
expressed as
m n
SvecNi A j=1 SvecN j B
Sem_N A,B = i=1
+ (1)
m+n m+n

where m and n are length of semantic vectors SvecN A and SvecN B , respectively.
p q
SvecVi A j=1 SvecV j B
Sem_V A,B = i=1
+ (2)
p+q p+q

where p and q are length of semantic vectors SvecV A and SvecV B , respectively. In
our case, semantic noun score, Sem_N A,B = 1+0.5+1+0.5+.0166
5+5
+ 1+0.5+0.5+0.5+1
5+5
=
0.66 and semantic verb score, Sem_V A,B = 2 + 2 = 1.
1 1

Step 7: We retain the same expression (as related works discussed above) to
compute overall word-sense semantic similarity between sentence pairs (S A , S B ) =
ζ × Sem_N A,B + 1 − ζ × Sem_V A,B ; the reason behind is to compare the sentential
similarity scores with the values obtained in the previous works with similar kind of
experimental setup. The authors did not drill into the insights of ‘ζ ’ parameter (also
called exponential balance coefficient (EBC)) which is borrowed from the previous
works and is usually set in the range [0.5, 0.1], and for our experiments, value of
EBC is set to 0.65. In our case, the overall word-sense similarity (S A , S B ) = 0.65 ×
0.666 + 0.35 × 1 = 0.78.
114 R. Dhagat et al.

3.2 Semantic Similarity Computation: Phase II

The concept of grammatical dependencies pioneered by Stanford research group led

by manning describes grammatical relations augmented with word arguments lying
in 2-tuple format. These binary arguments hold a governor argument (also known as
a regent or a head) and a dependent argument (also known as tail). It may sometimes
happen that two or more adjacently or non-adjacently lying words with in a sentence
may jointly reflect a different meaning in thematic sense. Recently, lots of work
is being done in carrying out text mining tasks using variant versions of semantic
(dependency) parsers. One such novel method of obtaining this portion of semantic
similarity is being discussed in the algorithm below.
Step 1: For all candidate sentence pair, (S A , S B ) the sentences S A , S B are parsed 
into a set of grammatical (binary) relations denoted by S A = T A1 , T A2 , . . . , T Am ,
S B = {TB1 , TB2 , . . . , TBn }. T i denotes the ith relation, where 1 ≤ i ≤ m, 1 ≤ j ≤ n, m
and n denotes
 the number
 of participating
 relations
 in sentence A and B, respectively.
T Ai = h iA , t Ai , d Ai and TBi = h iB , t Bi , d Bi , where h iA is the head, d Ai refers to
dependent, t Ai denotes the participating grammatical tag.
Step 2: The grammatical portion of semantic similarity between two relations can
be defined as
      
j j j
sim T Ai , TB = α × arg _sim h iA, h B + β × arg _sim d A, i
dB
 
j
× tag_sim t Ai , t B (3)

Here, ‘arg_sim’ denotes WordNet-based (path) similarity between corresponding

head and dependent nodes, respectively. We have borrowed the expression (3) from
the work by Zhang et al. [20] but with a disagreement toward assignment of parame-
ters α and β. According to any language scholar, as none of the words can be ignored
for reflecting semantics of a sentence, we provide equal importance to the sets of head
and dependent arguments, thus providing equal weights (= 0.5) to both head and
dependent arguments. Tag_simis easily  computed over corresponding grammatical
j j
relation tags such that tag_sim t A , t B = 1 if t Ai == t B else 0.
i

Step 3: The final semantic similarity computation is done at sentential level. Here,
the greatest value of inter-sentential grammatical relation similarities is averaged over
the count of participating grammatical relations as shown in Eq. (4).

max   i j 
m
sim T A, TB
1≤ j ≤n
i=1
Grammatical similarity =
m+n
n max   i j 
sim T A, TB
j=1
1≤i ≤m
+ (4)
m+n
Comparative Evaluation of Semantic Similarity Upon … 115

4 Experiments and Results

We have evaluated our methodology using preliminary baseline dataset comprising

nine sentence pairs, borrowed by Lee [12]. The semantic similarity values obtained
due to word-sense and grammatical-sense were tabulated in separate columns (3 and
4), some of which are shown in Table 1. The computed similarity measures now can
be compared with the equivalent results obtained by Lee [12] and Alzahrani et al. [6].
As seen in the experimental tabulations of Table 1, for the sentence pairs that were
found meaningfully highly similar, their high word-sense similarity scores justified
their most similar semantics; however, their less grammatical-sense similarity scores
indicated that while the first candidate of the pair was written in a clausal form,
the second candidate was completely scribed in phrasal form. Such observations
were made for sentence pairs: {1, 7}. Some other sentential pairs exhibited total
dissimilarity and were assigned very low word-sense similarity scores, for instance,
sentential pair numbers: {2, 4, 5, 8}. According to human judgments, these sentences
narrated totally different domain of context. There was still another category of
sentence pairs that were assigned equally high word-sense similarity scores and
grammatical-sense similarity scores; such sentential pairs narrated on similar topic
or context and also were found written in similar grammatical styles. For instance,
in sentential pair ID: {3, 6, 9}, both the sentences were found scribed in clausal form
of sentence construction.
The only way to authenticate the performance of the algorithm used in this paper
was to experiment our methodology upon Li et al. [2] datasets, given that Islam and
Inkpen [11], Croft et al. [13], Taieb et al. [14], Pawar and Mago [15], Zhang et al.
[20] and Farouk [21] too had compared their methodologies upon the same dataset,
setting it as benchmark standard for the aspiring semantic NLP researchers. It was
observed that all these works followed WordNet-driven knowledge-based approach
except by Islam and group. All of them used varied part-of-speech combinations of
word vectors and/or word-embeddings as feature vector variants except that latter
two work groups which used sentential parsed dependencies as structural features to
explore sentential semantics.
Even between the last two closely competent approaches, Farouk [21] was able
to measure the similarity of participating lexicons reflecting semantic inter-relations
using parsed dependencies from C & C parser as against Stanford dependency parser
used by Zhang et al. [20]. The overall similarity measures obtained from all the above-
mentioned participating models were tabulated as shown in Table 2. These variations
could only be reflected as minor deviations in the form of Pearson’s correlation
coefficient values as the only feasible performance metric to compare our model.
Their result comparisons were done by finding Pearson’s correlation coeffi-
cient between both (word-sense similarity and mean-human similarity) as well as
(grammatical-sense similarity and mean-human similarity) measures. It was found
that our methodology yielded better semantic similarity scores than those obtained by
116 R. Dhagat et al.

Table 1 Semantic similarity computations of sentence pairs (courtesy: Lee’s, 2011 datasets)
Sentence A Sentence B Word-sense Grammatical-sense Lee Alzahrani
similarity similarity (2011) (2015)
If she can be She is not 0.65 0.46 0.9125 0.75
more considerate
considerate to enough to be
others, she will more popular to
be more popular others
If she can be You are not 0.08 0.05 0.0195 0
more supposed to
considerate to touch any of the
others, she will art works in this
be more popular exhibition
I will not give If you could 0.68 0.69 0.938 0.933
you a second promise to be
chance unless careful, I would
you promise to consider to give
be careful this you a second
time chance
I will not give The obscurity of 0.22 0.18 0.419 0.357
you a second the language
chance unless means that few
you promise to people are able
be careful this to understand
time the new
legislation
About 100 The army 0.47 0.37 0.695 0.877
officers in riot entered in the
gear were forest to stop the
needed to break fight with
up the fight weapon
Your digestive Stomach is one 0.90 0.62 0.918 0.777
system is the of organs in
organs in your human body to
body that digest digest the food
the food you eat you eat
I do not think it It is an illegal 0.72 0.544 0.918 0.777
is a clever idea way to get what
to use an illegal you want, you
means to get should stop and
what you want think carefully
The powerful Political person 0.38 0.17 0.591 0.718
authority is sometimes
partial to the abuses their
members in the authority that it
same party with is unfair to the
it citizen
(continued)
Comparative Evaluation of Semantic Similarity Upon … 117

Table 1 (continued)
Sentence A Sentence B Word-sense Grammatical-sense Lee Alzahrani
similarity similarity (2011) (2015)
The fire An organization 1 0.92 1 1
department is an which has the
organization job of putting
which has the out fires is the
job of putting fire department
out fires

all six competent models. This was evident from the very promising value of corre-
lation score as 0.891 (for word-sense similarity) as compared to similarity compar-
isons put forth in Zhang’s [20] and Farouk’s [21] work. These correlations can be
seen tabulated in Table 3. The results obtained from the proposed methodology can
be illustrated in scatter plots shown in Fig. 3.

5 Conclusion

The salient promising feature of the current work that can be drawn at the end is that
the experiments are not performed by setting any kind of constraints on the input
(free) text nor narrowing the semantic feature spaces by removal of function or stop
words or filtering out certain specific grammatical relations representing the context
of the topic narrated in the sentences.
In other words, there is a fair attempt to find a generic method to compare texts
in the direction of NLP semantics. Unlike past pieces of works, the authors are still
on the move to find a suitable expression to compute over all sentential semantic
similarity contributed from the perspective of word-sense and grammatical-sense in
totality. Moreover, the less correlation value (0.677) of grammatical-sense similarity
measures for highly similar sentence pairs in experimental datasets needs further
investigations that is being undertaken as the next scope of research in this direction.
Table 2 Varied sentential semantic similarities for performance comparisons
118

R&G number Mean-human Li 2006 Islam 2008 LSS 2013 Pawar et al. 2018 Zhang et al 2020 Farouk 2020 Proposed Proposed
similarity similarity similarity
(Word-sense) (Grammatical)
1 0.01 0.33 0.06 0.18 0.02 0.04 0.11 0.12 0.13
5 0.01 0.29 0.11 0.20 0.07 0.07 0.13 0.11 0.27
9 0.01 0.21 0.07 0.28 0.01 0.03 0.07 0.08 0.13
13 0.11 0.53 0.16 0.17 0.29 0.07 0.20 0.25 0.28
17 0.13 0.36 0.26 0.32 0.36 0.14 0.28 0.33 0.37
21 0.04 0.51 0.16 0.32 0.23 0.10 0.21 0.33 0.31
25 0.07 0.55 0.33 0.22 0.28 0.13 0.27 0.39 0.33
29 0.01 0.33 0.12 0.22 0.13 0.07 0.21 0.13 0.33
33 0.15 0.59 0.29 0.32 0.76 0.08 0.33 0.35 0.42
37 0.13 0.44 0.20 0.28 0.1 0.09 0.24 0.39 0.30
41 0.28 0.43 0.09 0.32 0.05 0.11 0.23 0.30 0.21
47 0.35 0.72 0.30 0.20 0.16 0.46 0.30 0.25 0.23
48 0.36 0.65 0.34 1 0.54 0.42 0.35 0.28 0.44
49 0.29 0.74 0.15 1 0.30 0.39 0.34 0.30 0.36
50 0.47 0.68 0.49 0.8 0.25 0.49 0.28 0.36 0.24
51 0.14 0.65 0.28 0.8 0.3 0.1 0.21 0.30 0.26
52 0.49 0.49 0.32 1 0.84 0.31 0.40 0.46 0.29
53 0.48 0.39 0.44 0.47 0.89 0.40 0.40 0.52 0.47
54 0.36 0.52 0.41 0.8 0.78 0.05 0.26 0.30 0.09
55 0.41 0.55 0.19 0.8 0.31 0.07 0.32 0.54 0.29
(continued)
R. Dhagat et al.
Table 2 (continued)
R&G number Mean-human Li 2006 Islam 2008 LSS 2013 Pawar et al. 2018 Zhang et al 2020 Farouk 2020 Proposed Proposed
similarity similarity similarity
(Word-sense) (Grammatical)
56 0.59 0.76 0.47 0.8 0.98 0.38 0.49 0.61 0.37
57 0.63 0.70 0.26 1 0.48 0.37 0.32 0.49 0.35
58 0.59 0.75 0.51 0.8 0.89 0.56 0.44 0.48 0.50
59 0.86 1 0.94 1 0.86 0.86 0.87 1 0.87
60 0.58 0.66 0.60 0.8 0.90 0.43 0.51 0.52 0.40
61 0.52 0.66 0.29 0.8 0.93 0.37 0.30 0.48 0.34
62 0.77 0.73 0.51 1 1 0.52 0.51 0.78 0.55
63 0.56 0.64 0.52 1 0.7 0.45 0.44 0.50 0.38
64 0.96 1 0.93 1 0.87 0.93 0.87 1 1
65 0.65 0.83 0.65 1 0.85 0.36 0.55 0.71 0.33
Comparative Evaluation of Semantic Similarity Upon …
119
120 R. Dhagat et al.

Table 3 Pearson’s
Algorithm Pearson’s correlation
correlation comparisons to
coefficient
human judgment scores
Word-sense similarity (proposed 0.891
work)
Grammatical similarity (proposed 0.677
work)
Sentence similarity (Farouk M.) 0.881
Overall sentence similarity 0.877
(Zhang et al.)
Sentence similarity (Pawar et al.) 0.781
LSS (Croft et al.) 0.807
STS (Islam et al.) 0.853
STASIS (Li et al.) 0.816

1.50
r=.891
Word-sense
Similarity

1.00

0.50

0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Mean Human Similarity
1.50
r =.677
Grammatical
similarity

1.00

0.50

0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Mean Human Similarity

Fig. 3 Pearson’s correlation (r-value) comparisons with mean-human similarity scores (Datasets:
Li et al. 2006)

Acknowledgements The paper’s findings are the part of an ongoing research being carried out at
Research and Development cell of Bhilai Institute of Technology, Durg, Chhattisgarh. The authors
express gratitude to the sponsoring organization for providing an opportunity and infrastructure for
pursuing the experiments.
Comparative Evaluation of Semantic Similarity Upon … 121

References

1. Erkan G, Radev R (2004) Lexrank: graph-based lexical centrality as salience in text

summarization. Artif Intell Res 22:457–479
2. Li Y, McLean D, Bandar ZA, O’Shea JD, Crockett K (2006) Sentence similarity based on
semantic nets and corpus statistics. IEEE Trans Knowl Data Eng 18(8):1138–1150
3. Mihalcea R, Corley C, Strapparava C (2006) Corpus-based and knowledge based measures of
text semantic similarity. In: Proceedings of the 21st national conference on artificial intelligence.
AAAI, Boston, Massachusetts, pp 775–780
4. Miranda Chong MY (2013) A study on plagiarism detection and plagiarism direction iden-
tification using natural language processing techniques. Doctoral dissertation, University of
Wolverhampton, UK, pp 1–326. http://rgcl.wlv.ac.uk/papers/chong-thesis.pdf
5. Alzahrani S, Valade P, Salim N (2012) Using structural information and citation evidence
to detect significant plagiarism cases in scientific publications. J Assoc Inform Sci Technol
63(2):286–312
6. Alzahrani S, Valade P, Salim N (2015) Uncovering highly obfuscated plagiarism cases using
fuzzy semantic-based similarity model. J King Saud Univ Comput Inform Sci 27(3):248–268
7. Fellbaum C (ed) (1998) WordNet: an electronic lexical database. MIT Press, Cambridge, MA
8. Clough P, Stevenson M (2011) Developing a corpus of plagiarized short answers. Lang Resour
Eval. Special Issue Plagiarism and Authorship Analysis 45(1):5–24
9. Gaizauskas, Foster J, Wilks Y (2001) The METER corpus: a corpus for analyzing journalistic
text reuse. In: Proceedings of the corpus linguistics. Lancaster, UK, pp 214–223
10. Burrows S, Potthast M, Stein B, Eiselt A (2013) Webis crowd paraphrase corpus 2011 (Webis-
CPC-11). Zenodo
11. Islam A, Inkpen D (2008) Semantic text similarity using corpus based word similarity and
string similarity. ACM Trans Knowl Discovery Data (TKDD) 2(2):1–25
12. Lee MC (2011) A novel sentence similarity measure for semantic based expert systems. Exp
Syst Appl 38(5):6392–6399. https://doi.org/10.1016/j.eswa.2010.10.043
13. Croft D, Coupland S, Shell J, Brown S (2013) A fast and efficient semantic short text similarity
metric. In: 13th international proceedings on UK workshop on computational intelligence.
IEEE, USA, pp 221–227
14. Hadj Taieb MA, Ben Aouicha M, Bourouis Y (2015) FM3S: features-based measure of
sentences semantic similarity. In: 10th international proceedings on hybrid artificial intelligent
systems. Springer International Publishing, Bilbao, Spain, pp 515–529
15. Pawar A, Mago V (2018) Calculating the Similarity between words and sentences using a
lexical database and corpus statistics. IEEE Trans Knowl Data Eng 1–14
16. Vani K, Gupta D (2017) Detection of idea plagiarism using syntax—semantic concept
extractions with genetic algorithm. Exp Syst Appl 73:11–26
17. Özates S, Ozgur A, Radev D (2016) Sentence similarity based on dependency tree kernels
for multi-document summarization. In: 10th international proceedings on language resources
and evaluation. European Language Resources Association (ELRA), Portorož, Slovenia, pp
2833–2838
18. Wali W, Gargouri B, Ben Hamadou A (2017) Sentence similarity computation based on
WordNet and VerbNet Comput y Sistemas 21(4):627–635
19. Vakare T, Verma K (2019) Sentence semantic similarity using dependency parsing. In: 10th
international proceedings on computing, communication, and networking technologies. IEEE,
Kanpur, India, pp 2089–2092
20. Zhang P, Huang X, Zhang L (2020) Information mining and similarity computation for semi-
/un-structured sentences from the social data. IEEE Internet Things J. 2352–8648.https://doi.
org/10.1016/j.dcan.2020.08.001
122 R. Dhagat et al.

21. Farouk M (2020) Measuring text similarity based on structure and word embedding. Cogn Syst
Res 63(11):1–10
22. O’Shea J, Bandar Z, Crockett K, McLean D (2008) Pilot short text semantic similarity bench-
mark data set: full listing and description. http://www2.docm.mmu.ac.uk/STAFF/J.Oshea/
TRMMUCCA20081_5.pdf
The Role of Artificial Intelligence
and Data Science Against COVID-19

Saumya Tripathi and Dilip Kumar Sharma

Abstract Today, a massive outbreak of a deadly virus has challenged humanity in

each aspect; we deemed ourselves powerful. Millions of people are at risk, and in
some cases, even if the person has reported negative, it is relapsing. Globally, data
scientists are also working to tackle this issue by diagnosing, predicting it with the
help of machines. Artificial intelligence (AI) serves as a potential tool in fighting
against the COVID-19 pandemic. Science the onset of a pandemic, the use of AI has
exponentially increased to detect patients suffering or those who are critical. In this
paper, the authors discussed the role of AI and the selective review of constraints
associated with the functioning of AI. We will also discuss the shortcomings due to
the excessive data available between the private and public organizations related to
the health sector and their rigorous use without taking note of the data validity.

Keywords COVID-19 · Coronavirus · Data science · Artificial intelligence

1 Introduction

The virus named SARS-CoV-2 has been spreading terror in the hearts of people
due to its deadly nature. It was first identified in Wuhan, China, as pneumonia of
unknown cause on December 8, 2019 [1]. The spread of this disease has been so fast
and so wide that the world is struggling to stabilize itself. As of April 25, 2020, there
has been 2,802,118 cases of coronavirus and 195,443 deaths globally. The nations
are fighting the coronavirus as best they can by expanding their healthcare facilities,
stopping people from collecting at a particular place and in that they have also started
utilizing the power of data science, artificial intelligence (AI) and machine learning.
Since December 2019, a new type of coronavirus called novel coronavirus (2019-
nCoV, or COVID-19) was identified in Wuhan, China. The COVID-19 has then

S. Tripathi (B) · D. K. Sharma

GLA University, Mathura, India
D. K. Sharma
e-mail: dilip.sharma@gla.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 123
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_10
124 S. Tripathi and D. K. Sharma

rapidly spread all over China and the world. It can cause symptoms including fever,
difficulty in breathing, cough and invasive lesions on both lungs of the patients [1]. It
can spread to the lower respiratory tract and cause viral pneumonia. In severe cases,
patients suffer from dyspnea and respiratory distress syndrome [2, 53]. There are
total 175,622,44,582,692 cases, recovered case 159,171,963 and 3,788,723 death
according to World O meter [16]. The World Health Organization (WHO) has
declared that COVID-19 is responsible for the death of many individuals and termed
a pandemic. Nations are responsible for the spread and one’s own country wellness.
The countries use AI and machine learning (ML) tools to avoid and secure their
nations from this pandemic.
Many people who are suffering from COVID-19 infection are developing mild
to moderate respiratory tract infection and uneasiness and inhaling oxygen. Older
people suffering from another disease like blood pressure and diabetes other chronic
illness are experiencing more threat to their lives. To avoid this disease, the best way
that can be used is to prevent the spread of this disease as it is spreading from touch
and smell. When an infected person comes in contact with a healthy person and if
he sneezes or coughs that there are 99% chances that the healthy person might get
infected from the virus. The droplets from the saliva of the infected person will get
into the respiratory tract of a healthy person and may make him ill and suffered from
COVID-19. Therefore, it is necessary to maintain hygiene and keep sanitizing hands
as the guidelines provided by WHO indicates that either masking off and proper
sanitization can keep you off from the virus.
The governments from various countries have a hard time-fighting coronavirus
as they are putting all efforts into preventing the virus from spreading. The critical
conditions are managed by applying emergencies and using technology from equip-
ment to fight from it. The kits and their use contain artificial intelligence, data science,
data analytics, machine learning, thermal sensors and other techniques. Furthermore,
in Sect. 2, we have written about applications of data science, Sect. 3 is about the role
of artificial intelligence in predicting COVID-19 patients, Sect. 4 discusses guide-
lines issued by WHO, Sect. 5 discusses the role of data science in managing critical
patients, in Sect. 6, we listed some available datasets, Sect. 7 consists of the things
that could be done to control COVID-19, Sect. 8 consists of a summary of techniques
and challenges in applying them, and Sect. 9 deals with conclusion and future work.

2 Application of Data Science

In these times, many people feel helpless in the face of the pandemic. However,
people could focus on how they can give back to society by helping people in need.
Given below are some of the things that the data science community is doing in the
face of the novel coronavirus.
The Role of Artificial Intelligence and Data Science … 125

2.1 Real-Time Visualization

Some of the people are making interactive visualizations by utilizing the data given
by different governments and World Health Organization (WHO) with the help of
various tools available. This is helping many people track accurate and up-to-date
statistics.

2.2 Forecasting and Prediction

For making the database more accurate and suitable for finding results efficiently
and provide proper suggestion and treatment regarding the avoidance of the spread
of this virus.

2.3 Discovering Treatments

At Columbia University, two graduates are using machine learning to find treatment
for this virus. They screen therapeutic antibodies, and the probability of success is
quite high [3].

2.4 Calculating Risk Factors

The teams are trying to find various factors that can be responsible for the spread of
COVID-19. They are also trying to predict how can we stop the spread and which
measures are being helpful and which are not very helpful by using the data of
different regions and precautions taken in those regions.

2.5 Identifying Who is Dying and Why

People are trying to find the pattern among the people who are dying because of
COVID-19. This includes respiratory diseases, heart diseases, genetic component,
environmental factors, etc. But this data is challenging to access because of HIPAA
restrictions [4].
126 S. Tripathi and D. K. Sharma

2.6 Drug Discovery

The overall research is going on in order to find the most vital drug that can fight
the coronavirus. Though the scientist is still not successful but with the help of data
science, the scientist can track what all discovery using some similar salts in creating
a more appropriate drug is taking place.

2.7 Monitor Patients Health

Using artificial intelligence and data science, the health of patients can be monitored
using a similar pattern of cell structure and changes in the patient’s body and anatomy.

2.8 Track and Prevent Disease

By tracking the patients who are showing some signals regarding the flu and keep
monitoring them till they are negative can help prevent the virus and stop its wide-
spreading nature.

2.9 Robotics

In China, robots were used to help deliver medicines to patients so as to stop the
spread of the coronavirus among doctors. They were also used to disinfect hospitals
by using ultraviolet light.

2.10 Predictive Analytics

Making the patients aware of the symptoms and helping them to take certain kind of
salt and another remedy to help cope up with the virus and prevent it from spreading.

2.11 Contact Tracing

Data science-based tools are used to track patients or people who are quarantined
across the country. It is also used to trace contacts of the patients.
The Role of Artificial Intelligence and Data Science … 127

3 Role of Artificial Intelligence in Predicting COVID-19

Patients

The health industry needs a superpower to diagnose and make the decision to handle
the virus efficiently and provide proper suggestion and treatment regarding the avoid-
ance of the spread of this virus. AI serves as a boon to the health industry as it is
capable of handling numerous amount of data and use its intelligence and diagnose
the patient having certain issues common to categorize whether he/she is infectious
of COVID-19 or not. It also helps in understanding the development of vaccine and
their availability. This technology is used for prediction, analysis and tracking of
patients currently suffering from the virus and helps the sector to be aware of the
vast illness due to pandemic [56].

3.1 Early Detection

AI helps in monitoring and detecting disease using prior knowledge and through its
experience. As the data is supervised to the machine, therefore it is speedy and more
accurate rather than any medical doctor detection and diagnosis.

3.2 Decision Making

AI helps the patients to make the better decision during any critical situation as the
patients are moped derived toward the access and dealing it using dataset monitoring.

3.3 Treating Disease

Healthcare workers seek help from the machine and the Internet to find accurate
medicine by providing the symptoms of the patient. If the patient is well educated
about the COVID-19 virus, the patient illness is slowed down from spread.

3.4 Associated Care

A small study in Wuhan [18] has shown that the risk of dying increases with age.
Elderly people have more diseases such as heart disease, respiratory problems,
diabetes and clotting of blood. The place visited by him should be sanitized and
locked up, and people around him should be tested immediately.
128 S. Tripathi and D. K. Sharma

3.5 Checking Health Through Wearable

This includes respiratory diseases, heart diseases, genetic component and envi-
ronmental factors. It is made publicly available by AWS and is a repository that
constitutes of datasets related to machines.

3.6 End of Life Care

If the patient is not treated well using in intelligence and machine learning tools
gently and with care and all the medication is used well, then the infected person
may lose his life. But if the tools are used efficiently, and the medication and care is
taken properly, then the well-being of the patient can be maintained.
AI serves as a platform to train machines with huge dataset and analyze health
sector data and categorize accordingly. It also helps doctors and health vectors to
train machines and set algorithm to optimize the data related to the virus with speed
and accuracy. These machines help in retrieving the more precise data and support
the health workers by diagnosing the issue more efficiently.

4 Precautions to be Taken to Avoid Getting Infected

In order to avoid getting infected, whether from indoor or outdoor, certain measure
and guidelines are published by World Health Organization (WHO). If a member is
infected from one family, then avoid getting in touch with him, and keep necessary
action when getting in touch with him. If one is moving out, then avoid getting in
touch with the infected person and maintain proper distance. These are some of the
limitations and guidelines to be followed to avoid virus outbreak in the country.
The guidelines issued by WHO indicates that a person should keep washing his
hand whenever he is going outdoor or if there is an infected person indoor that wash
hand while being indoor. Always have a mask on, and in the second wave, the double
mask was changed in the guidelines. Always disinfect the area where the ill person
is, and do not share your personal item with them. Social distancing is the key to
preventing the virus from getting into one’s respiratory tract. For children, special
guidelines are indicated in which children above three years of age have to keep the
mask on and keep sanitizing their hands. They are not allowed to go outdoors. All the
schools and educational institutions are temporarily shut for children. For traveling
in a four-vehicle like a car, try to keep proper ventilation, and if there are two-person
traveling that tries to sit socially distant to avoid getting infected.
The Role of Artificial Intelligence and Data Science … 129

5 Role of Data Science in Managing Critical Patients

People affected by COVID-19 are requiring or critical respiratory care. Data science
has come up to care for these patients. In [7], they have studied various researches
and come up with a three-stage process that will categorize the search result in three
states to help the patients in ICU. They have read studies from [7–11].
We can also determine who the people most vulnerable are so that they could be
separated beforehand from the rest [38]. It will also help us in identifying the people
who are more susceptible to death. A small study in Wuhan [37] has shown that the
risk of dying increases with age. Elderly people have more diseases such as heart
disease, respiratory problems, diabetes and clotting of blood.

6 Available Datasets

There are various datasets freely available on the Web to help combat COVID-19
(Table 1).
These are only some of the datasets that are available from the sea of them available
online. Countries have shared their own datasets online for everyone to analyze and
forecast. There are related to age groups, counts of confirmed, recovered and dead,
hospital beds and testings, etc., depending on the country.

7 Controlling COVID-19

Worldwide, countries are quarantining people, imposing lockdown, sanitizing to stop

community spread. Public health has been the top priority for countries currently,
and they are willing to sacrifice their economy to save lives. Although there is no
single measure that could be taken to stop the pandemic worldwide, there are still
some steps that could be taken like.

7.1 Social Distancing

People should follow proper protocols to ensure a safe distance from each other. A
minimum of 1 m distance is required.
130 S. Tripathi and D. K. Sharma

Table 1 Available datasets for the COVID-19 data

S. No Dataset name Description
1 “Google Covid-19 Public Datasets” [39] This dataset comprises various datasets
that are provided by Google for free so that
it is more accessible to researchers, data
scientist and analysts. This contains JHU
CSSE (Johns Hopkins Center for Systems
Science and Engineering) dataset,
OpenStreetMap data and global health data
from the World Bank.
2 “Covid-19 Open Research Dataset” [40] At the Allen Institute, the semantic scholar
team is providing CORD-19 by partnering
with groups that are leading in COVID-19
research. It is also available for free on the
Web site
3 “The COVID tracking project collects and As the name suggests, this has all the data
publishes the complete testing data published from US states and territories of
available for US states and territories” [41] testing constitutes of positive, negative and
pending, deaths and total test results
4 “LitCovid” [42] It is an organized hub of literature to track
researches on COVID-19. It takes into
account how many papers are getting
published every week and from which
country. It is the resource of NCBI
5 “ECDC data on Covid-19 cases ECDC is a European Centre for Disease
Worldwide” [43] Prevention and Control, and it delivers data
on all geographical distributions of novel
coronavirus
6 “COVID-19 Open Research Dataset It is a Kaggle dataset, constituting 57,000
Challenge (CORD-19)” [44] scholarly articles, with over 45,000
including full text, about COVID-19,
SARS-CoV-2, and related coronaviruses.
To cope up with this, the medical research
community needs different AI experts to
find answers
7 “A public data lake for analysis of It is made publicly available by AWS and
COVID-19” [45] is a repository that constitutes of datasets
related to characteristics and spread of
COVID-19
8 “COVID19 Global Forecasting (Week 4)” It is a dataset launched by Kaggle and is
[46] one of the challenges to answering WHO
and NASEM questions on the pandemic
9 “Research Resource Repository Related datasets for research and analytics.
COVID-19” [50] The data may be directly or indirectly
related to coronavirus. It deals with some
subsidiary issues also
(continued)
The Role of Artificial Intelligence and Data Science … 131

Table 1 (continued)
S. No Dataset name Description
10 “Complete our world in Data for It is a crucial metric to understand the
COVID-19 dataset” [51] pandemic from a positive perspective
11 “Activity, population and Deals with the location and population
location-COVID-19” [52] table in compliments and COVID-19
dataset
12 “Research initiatives in HRS for Data resource including the data for
COVID-19” [48] COVID-19 patients and questionnaire with
the US State policyholder
13 “Covid-19 OPEN DATA–FIGSHARE” It deals with sharing the initiative of
[47] collaboration of universities and medical
colleges
14 “ICPSR: Data Repository for COVID-19” It deals with the health industry and
[53] maintains a centralized repository for the
medical record of the COVID-19 patients
and store a secure enclave to study those
records
15 “VIVLI-COVID Data” [54] It deals with the clinical trials on
COVID-19

7.2 Testing

Countries should start testing people in large volumes as the virus is spreading expo-
nentially in the world. People should also ensure that if any symptoms occur, they
should go and have themselves checked out.

7.3 Travel Measures

People should avoid traveling and getting in touch with people. Countries have sealed
their borders, but also in-country travel should be avoided.

7.4 Containment

If a person has tested positive, he/she should be contained and quarantined so as to

not spread the virus in other places. The place visited by him should be sanitized and
locked up, and people around him should be tested immediately.
132 S. Tripathi and D. K. Sharma

7.5 Personal Hygiene

Every person should wear a mask whenever going out and should wash hands regu-
larly. Personal hygiene is necessary in this case as the virus enters our body from the
eyes, nose, mouth, etc., and people should avoid touching their face. They should
wash or sanitize their hands multiple times in a day.

8 Summary of Works Done by Researchers

See Table 2.

9 Conclusion and Future Work

AI serves as a potential tool in fighting against COVID-19 and many similar

pandemic. Obviously data serves as the main center for such pandemic to give
optimum results. The increase of documents from hospitals and patients patholog-
ical data has made us rely more on artificial techniques and machine learning to
extract a more accurate data so that the diagnosis can be done rapidly. Extraordinary
surveillance by the government to keep the track of records to get over the pandemic
takes into lots of dataset and perfect coordination with the health workers. Due to the
evolving pandemic, we are dependent on deep learning approaches and are devel-
oping a model that is more prognostic and empirical. These models help to retrieve
datasets using AI models. These models help the health workers to guide and help
patients to recover and establish a more reliable bond with their health workers as
due to the crises patients and their relatives are more apprehensive. It is of great
importance for the health workers to take care of their personal data and be justified
with that. AI serves as a boon in the present crises to accelerate the digitization of
economy and production activities.
In this paper, we have discussed about how data science, artificial intelligence
and machine learning are used against the spread of COVID-19. It also helps doctors
and health vectors to train machines and set algorithm to optimize the data related
to the virus with speed and accuracy. These machines help in retrieving more of
the various datasets that could be used to do the predictions and forecasting, real-
time visualizations, finding contacts of people that are exposed to this virus and
various other applications as discussed above in the paper. For future work, there
can be various other applications that could be added to this paper. Every country
has issued its own datasets. So, details about them can also be added to a paper.
Furthermore, there are many things that could be done to stop the spread of them we
have listed some of the main ones. All could be listed in detail.
The Role of Artificial Intelligence and Data Science … 133

Table 2 Authors and there related work regarding COVID-19

Authors Data type Technique/method Challenges
Wouters et al. [32], Population density, Deep learning, Low quality of data
Mathur et al. [9] strategic data clustering algorithm, and issues with
DBSCAN privacy factors
Pham et al. [55] Public announcements Naïve Bayes model, Insufficient data and
and distribution data machine learning and ambiguity, and there
is no truthfulness of
data being correct
Vaishya et al. [20], Radiology data (X-ray Camera system and Not able to explain
Wehbe et al. [26] and CT scan) image recognition the results and
technique results taking a long
time to train and
Vulnerabilities need
to involve medical
experts to give
correct results
Cresswell et al. [29], Camera images AI-based computer Need a high-quality
Abu-Raddad et al. (infrared images) vision and FR (facial system to capture
[35] recognition system) images and noise
present may make it
difficult to train
Zhou et al. [11] Satellite data (various Non-parametric Noises in satellite
location using GPS) process using Gaussian images, phone data
technique not synchronized
Soltan et al. [27] Text data from social Deep learning, Insufficient data
media machine learning, from time series,
speech recognition and Privacy issues and
synthesis of certain invasion, deep
linguistic data learning models take
Information extraction a long time and are
using machine learning very heuristic
Social media reports
might not be of good
quality and be
multidimensional
Vijayashanthi et al. Healthcare reports, Dynamic programming To establish trust
[12] travel data heuristic data with autonomous
system
Rahmatizadeh et al. Text data on various CNN-based model Results to be
[7] COVID-19 details data, supervised verified, but it will
learning, text extraction become extremely
tedious
Efforts to train
workers to work with
autonomous systems
134 S. Tripathi and D. K. Sharma

References

1. Who.int, 2020. [Online]. Available https://www.who.int/docs/default-source/coronaviruse/sit

uation-reports/20200226-sitrep-37-covid-19.pdf?sfvrsn=2146841e_2. Accessed 17 April 2020
2. Song F, Shi N, Chan F, Zhang Z, Shen J, Lu H, Shi Y et al (2020) Emerging 2019 novel
coronavirus (2019-nCoV) pneumonia. Radiology 295(1):210–217
3. Datascience.columbia.edu. 2020. DSI alumni use machine learning to discover coronavirus
treatments. Data Science Institute. [online]. Available at https://datascience.columbia.edu/dsi-
alumni-use-machine-learning-discover-coronavirus-treatments. Accessed 18 April 2020
4. HealthITAnalytics, 2020. Understanding the COVID-19 pandemic as a big data analytics
issue. [online]. Available at https://healthitanalytics.com/news/understanding-the-covid-19-
pandemic-as-a-big-data-analytics-issue. Accessed 18 April 2020
5. MarkTechPost, 2020. List of COVID-19 resources for machine learning and data science
research Marktechpost. [online]. Available at https://www.marktechpost.com/2020/04/12/list-
of-covid-19-resources-for-machine-learning-and-data-science-research/. Accessed 18 April
2020
6. Petropoulos F, Makridakis S (2020) Forecasting the novel coronavirus COVID-19. PloS One
15(3):e0231236
7. Rahmatizadeh S, Valizadeh-Haghi S, Dabbagh A (2020) The role of artificial intelligence in
management of critical COVID-19 patients. J Cell Mol Anesth 5(1):16–22
8. Alimadadi A, Aryal S, Manandhar I, Munroe PB, Joe B, Cheng X (2020) Artificial intelligence
and machine learning to fight COVID-19
9. Mathur P, Burns ML (2019) Artificial intelligence in critical care. Int Anesthesiol Clin 57(2):89–
102
10. Chaplain X, Huet O (2019) Is artificial intelligence (AI) at the doorstep of intensive care units
(ICU) and operating room (OR)? Anaesth Crit Care Pain Med 38(4):337–338
11. Zhou W, Li H, Tian Q (2017) Recent advance in content-based image retrieval: a literature
survey. arXiv preprint arXiv:1706.06064
12. Vijayashanthi M, Krishna VV, Reddy G (2018) Survey on recent advances in content-based
image retrieval techniques. J Innov Comput Sci Eng 7(2):41–48
13. Gutierrez G (2020) Artificial intelligence in the intensive care unit. Crit Care 24:1–9
14. Mupparapu M, Wu CW, Chen YC (2018) Artificial intelligence, machine learning, neural
networks, and deep learning: futuristic concepts for new dental diagnosis. Quintessence Int
(Berlin, Germany: 1985) 49(9):687–688
15. Bini SA (2018) Artificial intelligence, machine learning, deep learning, and cognitive
computing: what do these terms mean and how will they impact health care? J Arthroplasty
33(8):2358–2361
16. Coronavirus Update (Live): 175,622,445 cases and 3,788,723 deaths from COVID-19 virus
pandemic—Worldometer. Worldometers.info, 2021. [Online]. Available https://www.worldo
meters.info/coronavirus/. Accessed 11 June 2021
17. Lin E, Tsai SJ (2019) Machine learning in neural networks. Adv Exp Med Biol 1192:127–137
18. Hashimoto DA, Witkowski E, Gao L, Meireles O, Rosman G (2020) Artificial intelligence
in anesthesiology: current techniques, clinical applications, and limitations. Anesthesiology
132(2):379–394
19. Ghassemi M, Celi LA, Stone DJ (2015) State of the art review: the data revolution in critical
care. Crit Care 19(1):1–9
20. Vaishya R, Javaid M, Khan IH, Haleem A (2020) Artificial intelligence (AI) applications for
COVID-19 pandemic. Diabetes Metab Syndr 14(4):337–339
21. Jamshidi M, Lalbakhsh A, Talla J, Peroutka Z, Hadjilooei F, Lalbakhsh P, Jamshidi M et al
(2020) Artificial intelligence and COVID-19: deep learning approaches for diagnosis and
treatment. IEEE Access 8:109581–109595
22. Zhou Y, Wang F, Tang J, Nussinov R, Cheng F (2020) Artificial intelligence in COVID-19 drug
repurposing. Lancet Digit Health
The Role of Artificial Intelligence and Data Science … 135

23. Lalmuanawma S, Hussain J, Chhakchhuak L (2020) Applications of machine learning and arti-
ficial intelligence for COVID-19 (SARS-CoV-2) pandemic: a review. Chaos Solitons Fractals
25:110059
24. Summers RM (2021) Artificial intelligence of COVID-19 imaging: a hammer in search of a
nail
25. Mushtaq J, Pennella R, Lavalle S, Colarieti A, Steidler S, Martinenghi CM, Palumbo D, Esposito
A, Rovere-Querini P, Tresoldi M, Landoni G (2021) Initial chest radiographs and artificial intel-
ligence (AI) predict clinical outcomes in COVID-19 patients: analysis of 697 Italian patients.
Eur Radiol 31(3):1770–1779
26. Wehbe RM, Sheng J, Dutta S, Chai S, Dravid A, Barutcu S, Katsaggelos AK et al (2021)
DeepCOVID-XR: an artificial intelligence algorithm to detect COVID-19 on chest radiographs
trained and tested on a large US clinical data set. Radiology 299(1):E167–E176
27. Soltan AA, Kouchaki S, Zhu T, Kiyasseh D, Taylor T, Hussain ZB, Peto T, Brent AJ, Eyre DW,
Clifton DA (2021) Rapid triage for COVID-19 using routine clinical data for patients attending
hospital: development and prospective validation of an artificial intelligence screening test.
Lancet Digit Health 3(2):e78–e87
28. Agarwal M, Saba L, Gupta SK, Carriero A, Falaschi Z, Paschè A, Danna P, El-Baz A, Naidu
S, Suri JS (2021) A novel block imaging technique using nine artificial intelligence models for
COVID-19 disease classification, characterization and severity measurement in lung computed
tomography scans on an Italian cohort. J Med Syst 45(3):1–30
29. Cresswell K, Tahir A, Sheikh Z, Hussain Z, Hernández AD, Harrison E, Williams R, Sheikh
A, Hussain A (2021) Understanding public perceptions of COVID-19 contact tracing apps:
artificial intelligence–enabled social media analysis. J Med Internet Res 23(5):e26618
30. Vasileiou E, Simpson CR, Robertson C, Shi T, Kerr S, Agrawal U, Sheikh A et al (2021)
Effectiveness of first dose of COVID-19 vaccines against hospital admissions in Scotland:
national prospective cohort study of 5.4 million people
31. Perrotta D, Grow A, Rampazzo F, Cimentada J, Del Fava E, Gil-Clavel S, Zagheni E (2021)
Behaviours and attitudes in response to the COVID-19 pandemic: insights from a cross-national
Facebook survey. EPJ Data Sci 10(1):1–13
32. Wouters OJ, Shadlen KC, Salcher-Konrad M, Pollard AJ, Larson HJ, Teerawattananon Y, Jit M
(2021) Challenges in ensuring global access to COVID-19 vaccines: production, affordability,
allocation, and deployment. Lancet
33. Fan G, Yang Z, Lin Q, Zhao S, Yang L, He D (2021) Decreased case fatality rate of COVID-19 in
the second wave: a study in 53 countries or regions. Transboundary Emerg Dis 68(2):213–215.
Fallahgoul H (2021) Inside the mind of investors during the COVID-19 pandemic: evidence
from the StockTwits data. J Financ Data Sci 3(2):134–148
34. Bendavid E, Oh C, Bhattacharya J, Ioannidis JP (2021) Assessing mandatory stay-at-home and
business closure effects on the spread of COVID-19. Eur J Clin Investig 51(4):e13484
35. Abu-Raddad LJ, Chemaitelly H, Butt AA (2021) Effectiveness of the BNT162b2 COVID-19
vaccine against the B. 1.1. 7 and B. 1.351 variants. New England J Med
36. Cosgriff CV, Celi LA, Stone DJ (2019) Critical care, critical data. Biomed Eng Comput Biol
10.1179597219856564
37. Zhou F, Yu T, Du R, Fan G, Liu Y, Liu Z, Cao B (2020) Clinical course and risk factors for
mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study.
Lancet 395(10229):1054–1062
38. Arxiv.org, 2020. [Online]. Available https://arxiv.org/pdf/2003.07347.pdf. Accessed 25 April
2020
39. Google Cloud Platform. Console.cloud.google.com, 2020. [Online]. Available https://console.
cloud.google.com/marketplace/browse?filter=solution-type:dataset&filter=category:covid19.
Accessed 29 April 2020
40. CORD-19 | Semantic Scholar. Pages.semanticscholar.org, 2020. [Online]. Available https://
pages.semanticscholar.org/coronavirus-research. Accessed 29 April 2020
41. Most recent data. The COVID Tracking Project, 2020. [Online]. Available https://covidtrac
king.com/data. Accessed 29 April 2020
136 S. Tripathi and D. K. Sharma

42. LitCovid—NCBI—NLM—NIH. Ncbi.nlm.nih.gov, 2020. [Online]. Available https://www.

ncbi.nlm.nih.gov/research/coronavirus/#data-download. Accessed 29 April 2020
43. Download today’s data on the geographic distribution of COVID-19 cases world-
wide. European Centre for Disease Prevention and Control, 2020. [Online]. Avail-
able https://www.ecdc.europa.eu/en/publications-data/download-todays-data-geographic-dis
tribution-covid-19-cases-worldwide. Accessed 29 April 2020
44. COVID-19 Open Research Dataset Challenge (CORD-19). Kaggle.com, 2020. [Online]. Avail-
able https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge. Accessed
29 April 2020
45. A public data lake for analysis of COVID-19 data | Amazon Web Services. Amazon
Web Services, 2020. [Online]. Available https://aws.amazon.com/blogs/big-data/a-public-
data-lake-for-analysis-of-covid-19-data/. Accessed 29 April 2020
46. COVID19 Global Forecasting (Week 4) | Kaggle. Kaggle.com, 2020. [Online]. Available
https://www.kaggle.com/c/covid19-global-forecasting-week-4/data. Accessed 29 April 2020
47. COVID-19.zbmed.de, 2020. [Online]. Available https://covid-19.zbmed.de/externalReso
urces/literature/. Accessed 25 April 2020
48. Coronavirus outbreak is the world’s worst crisis since Second World War. Dailytimes.com.pk,
2020. [Online]. Available https://dailytimes.com.pk/586812/coronavirus-outbreak-is-the-wor
lds-worst-crisis-since-second-world-war/. Accessed 29 April 2020
49. 2020. [Online]. Available http://www.unitedindiversity.org/information/425/coronavirus-or-
covid-19/en. Accessed 29 April 2020
50. 2020. [Online]. Available https://datascience.nih.gov/covid-19-open-access-resources.
Accessed 25 May 2020
51. Our World in Data, 2020. [Online]. Available https://ourworldindata.org/coronavirus-source-
data. Accessed 25 May 2020
52. COVID-19 activity—location and population table, 2020. [Online]. Available https://data.
world/covid-19-data-resource-hub/covid-19-activity-location-population-table. Accessed 25
May 2020
53. OPENICPSR, 2020. [Online]. Available https://www.openicpsr.org/openicpsr/covid19.
Accessed 25 May 2020
54. VIVLI-Center for Global Clinical Research Data, 2020. [Online]. Available https://vivli.org/
how-to-share-your-research-on-the-vivli-covid-19-portal/. Accessed 27 May 2020
55. Pham QV, Nguyen DC, Hwang WJ, Pathirana PN (2020) Artificial intelligence (AI) and big
data for coronavirus (COVID-19) pandemic: a survey on the state-of-the-art
56. Agbehadji IE, Awuzie BO, Ngowi AB, Millham RC (2020) Review of big data analytics,
artificial intelligence and nature-inspired computing models towards accurate detection of
COVID-19 pandemic cases and contact tracing. Int J Environ Res Public Health 17(15):5330
57. Zhu H, Wei L, Niu P (2020) The novel coronavirus outbreak in Wuhan, China. Glob Health
Res Pol 5(1):1–3
Spectral Analysis of DNA on 1-D
Hydration Enthalpy-Based Numerical
Mapping Using Optimal Filtering

Subhajit Kar, Madhabi Ganguly, and Aryabhatta Ganguly

Abstract The paper proposed a new numerical coding technique of DNA, based
on hydration enthalpy, to find the coding regions of a gene. The genetic sequences
when converted into the numerical sequence can be applied to a predesigned least
square FIR filter to capture period three property of coding regions. Moving average
filter is then used to eliminate high-frequency noise present in the spectrum due to
long range correlation between nucleotide bases. The proposed algorithm has been
applied successfully to various standard datasets to predict the exon boundaries.
The performance of the proposed method is compared with other popular numeric
encoding schemes, namely dipole moment and pseudo-EIIP, and found superior in
terms of accuracy and other evaluation parameters. The proposed method achieved
an accuracy of 92% on the benchmark gene F56F11.5 of C Elegans (Accession
number AF099922).

Keywords Genomic signal processing · Exons · Digital filter · Hydration

enthalpy · Coding regions

1 Introduction

DNA or deoxyribonucleic acid which consists of four nucleotide bases, namely

adenine (A), cytosine (C), thymine (T), and guanine (G), is responsible for trans-
ferring heredity information from generation to generation. It is made up of genes
and intergenic spaces (Fig. 1a). Each gene contains a particular set of instruction.
Eukaryotic genes are further divided into coding and non-coding regions.
Discrimination of protein coding regions from non-coding regions is called
splicing. Signal processing methods rely on period three property for that purpose
[1]. The property states that a protein coding region of length N exhibits a relatively

S. Kar (B) · M. Ganguly

Department of Electronics, West Bengal State University, Kolkata, India
A. Ganguly
Department of Multimedia, Brainware University, Kolkata, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 137
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_11
138 S. Kar et al.

Fig. 1 a Eukaryotic and prokaryotic gene structures, b double helix structure of DNA

large value at discrete frequency k = N/3 but near zero value at other frequencies
corresponding to magnitude spectrum.
Mapping of the DNA character sequence into its corresponding numeric sequence
is eventual requirement of applying DSP mechanism. Numerical sequence mapping
from DNA is classified into fixed mapping, DNA physicochemical property-based
mapping and statistical property-based mapping. Each of these classifications comes
with their own advantages and disadvantages [2–4]. The detection of correct exons
vastly depends on these mapping techniques. Voss proposed a 4-D binary represen-
tation of DNA sequence which can be effectively used in exon finding problem
[5]. But the method is computationally complex since it employed four unique
binary sequences to solve the exon finding problem. To ease the time complexity, we
have proposed the hydration enthalpy-based mapping method where the hydration
enthalpies of the four nucleotide bases are derived using Monte Carlo simulation.
The technique has yielded superior result in identifying these coding areas with great
accuracy in eukaryotic genes. Since hydration enthalpy depends on various physic-
ochemical property of nucleotide bases including size of the bases, their dipole
moments and number of hydrophilic center in the particular base, so it can take
account of multiple properties of DNA at a time.
In the DNA helix, the bases: adenine (A), cytosine (C), thymine (T) and guanine
(G) are each linked with their complementary bases by hydrogen bonding (Fig. 1b)
between themselves. All hydrogen bonds of nucleotide bases are strengthened and
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based … 139

protected from solvent hydration by the hydrophobic stacking of the bases. At the time
of hydration, the nucleotide bases would quickly form bridges with water molecules
and hydrate with solvent water. Recent study proved that water molecules are respon-
sible for the double helical structure of DNA, whereas hydrogen bonds are responsible
for sorting the base pairs so that they can link together in correct order [6]. Thus,
hydration plays a crucial role in stability of DNA. As the coding regions of DNA
are more stable and conserved compared to the non-coding counterparts, hydration
enthalpy can be a good discrimination property between these two regions.

2 Materials and Method

The proposed method to determine protein coding region in eukaryotic gene employs
two stage digital filtering for spectral analysis. The purpose of the spectral analysis of
nucleotide sequence is to compute its period three spectral components located at a
frequency equal to 2π/3 in the PSD spectrum. The nucleotide sequence is converted
into numerical sequence using the hydration enthalpy of nucleotide bases. Generated
sequence is now applied to least square FIR filter designed to recognize period three
property. Removal of high-frequency noise present in the PSD is done by moving
average filter. The output power spectra of designed filter will indicate the exons
by certain peaks. The flowchart of the implemented algorithm comprises six steps
which is depicted in Fig. 2.

2.1 DNA Sequence Database

In this work, several datasets such as HMR195, GENESCAN and Burset and Guigo
are considered for analysis. The genes of NCBI accession numbers AF071552,
AF061327, M62420, AF059734, AF099922 and D10914 were downloaded from
NCBI Web site (https://www.ncbi.nlm.nih.gov/) in FASTA format. The nucleotide

Fig. 2 Prediction of protein coding region using proposed numeric representation

140 S. Kar et al.

sequences are chosen so that sequence length must be smaller than 10,000 bp and
greater than 100 bp. Genes with different numbers of exons are chosen to detect
the efficiency of the method. The encoded proteins by the coding regions can be
observed from UNIPROT Web site (https://www.uniprot.org/uniprot/P18440).

2.2 Numerical Mapping of DNA Sequence

The mapping technique maps the DNA sequence into a one-dimensional numerical
sequence. In this work, the enthalpy of hydration of the nucleotide bases is calculated
using Monte Carlo simulation [7]. Enthalpy of hydration is defined as the quantity
of energy produced upon dilution of one mole of gaseous nucleotide bases. The
solvation of DNA bases into water constitutes couple of steps. In the first step,
the hydrogen bonds between the water molecules break apart. It is an endothermic
process requiring very small amount of heat energy. In the second step, the small water
molecules bond to the various hydration sites of nucleotide bases. During the process,
water molecules form hydrogen bonds with the hydrophilic centers of nucleic acid
bases as well as between themselves. This step includes interaction between solute–
solvent and solvent–solvent particles. New hydrogen bonds are created in the process
releasing heat energy. Thus, the second step is an exothermic process. The total
hydration energy of a nucleotide base can be found by the relation.

E Total = E Soln + E Solv + E int (1)

where E Soln , E Solv and E int represent energy contributions from solute–solvent,
solvent–solvent and intermolecular interactions, respectively.
Hydration energy of nucleotide bases is always negative due to the release of
energy during this process. The solvation energy of DNA bases in water is in the
following order depending upon the number of hydrophilic center present in the
bases:
Etotal(Guanine) > Etotal(Thymine) > Etotal (Cytosine) > Etotal(Adenine).
The adapted hydration enthalpy of the four nucleotide bases is as follows.

Nucleotide base Hydration enthalpy (Kcal/mol)

A −39.28
C −40.58
T −40.63
G −47.99

In this way, a sequence x(n) = ‘ACTAGAT’ is transformed into its equivalent

numerical representation as x(k) = [−39.28 −40.58 −40.63 −39.28 −47.99 −
39.28 −40.63].
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based … 141

2.3 Linear Phase Least Square FIR Filtering

The critical function of the algorithm is to design a precise band pass filter to capture
the three base periodicities present in the coding regions. Designing linear phase FIR
filters applying window methods is relatively simple yet lack of precise control of the
critical frequencies such as wp and ws (pass-band and stop-band cut-off frequencies).
Optimized filters are preferred to overcome the problem. One of them is least square
FIR filter method which minimizes the weighted, integrated squared error between
a desired and actual magnitude response of the filter over a set of desired frequency
bands [8]. In our method, the desired signal is the ideal piecewise linear output of
the designed filter. In order to predict the true exon locations in gene, a direct form
FIR filter of order N = 270 has been designed. The goal of the filter is to detect three
base periodicity, and hence, the center frequency of the filter should be 2π/3 = 0.667
π radian.
The designed filter has the pass-band between 0.664 π radian and 0.672 π radian.
Another very important criteria needed to be set is weight value of each band which
controls the magnitude response of the filter.
A weighted moving average filter is implemented using Gaussian window of
length 30 which gives output free of noise.

2.4 Calculation of Power Spectral Density

Power spectral density of filtered output with respect to different base position is
obtained by squaring the filter response. For instance, if Y (k) is the filtered output of
a numerical sequence x(n), then PSD of the sequence can be generated by [abs(y)2 ]
in MATLAB environment. Thus, the equation of power spectrum of output filter can
be written as:

S(k) = [abs(Y (k))2 ] where k = 0, 1, 2, . . . , N − 1 (2)

2.5 Evaluation Criterion

Various evaluation parameters such as specificity (S p ), sensitivity (S n ), accuracy,

Matthews correlation coefficient (MCC) and F1 score are measured to assess the
performance of the coding scheme at nucleotide level. Similarly, various evaluation
parameters at exon level like discrimination factor, miss rate and wrong rate are also
measured and taken into consideration.
Finally, area under the curve of ROC is considered for evaluation purpose. ROC
curve is a graph that measures performance and accuracy of a binary classifier system.
142 S. Kar et al.

Here, the true positive rate (sensitivity) is plotted as a function of the false positive
rate (100-Specificity) for different threshold points.

3 Results

The encoded genes from different dataset with the help of proposed numerical repre-
sentation method are passed through specifically designed bandpass filter with center
frequency 2π/3 to get desired result. The output PSD plot of the experimented genes
is given in Figs. 3 and 4. The dotted lines describe the boundaries of actual coding

Fig. 3 Output power spectrum plot for genes: a AF071552, b AF061327

Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based … 143

Fig. 4 Output power spectrum plot for different genes: a M62420, b AF059734, c AF099922, d
HAMHG5
144 S. Kar et al.

regions of genes as specified in NCBI Website. The blue, black and yellow plots
are representing the PSD obtained from pseudo-EIIP, hydration enthalpy and dipole
moments, respectively.
To assess the efficacy of the proposed numerical representation based on enthalpy
of hydration, various evaluation parameters are computed from the output power
spectral density plot and compared with two well-established coding measure,
namely pseudo-EIIP and dipole moment [9, 10]. The detail results of predicted
boundaries of gene F56F11.5 are provided in Table 1. Here, NA suggests there
is no exon in the NCBI database. The detail evaluation results at nucleotide level
are depicted in Fig. 5 with the help of bar graphs. The sensitivity and specificity for
the benchmark gene having NCBI accession number AF099922 found to be 88%
and 93%, respectively, which are best when compared to the other two encoding
schemes. Receiver operating characteristic curve for gene AF099922 is plotted for
all the three numerical representations, namely pseudo-EIIP, dipole moment and
hydration enthalpy, for further evaluation purpose.
An exon level study of different evaluation parameter performed through Table 2
which reveals the efficacy of the proposed numerical model.
The experiment result showing great improvement in results as the miss rate and
wrong rate for most of the exons is zero. The discriminating factors are also greater
than one for most of the genes indicating that the exons and introns could be easily
identified in the output PSD with respect to the threshold value.
We have measured the area under the curve (AUC) in case of hydration energy
from Fig. 6, and it is found to be 0.8631 which is very good considering the fact that
AUC close to 0.9 or above indicates excellent accuracy. A wide number of numerical
representations have been evaluated in Akhtar et al. through AUC criterion, found
best of the value of 0.81 using paired numeric representation of DNA [11].

4 Discussions

To show the superiority of the new hydration enthalpy-based mapping over the previ-
ously adopted encoding schemes, a comparative study is adopted for various genes
according to the findings in previous literatures.
Gene Accession No AF071552:
For the gene AF071552, the output of the designed algorithm is comparable with the
previous findings [12]. The proposed method produces MCC value of 0.55 which
suggest good prediction accuracy of coding and non-coding regions. The sensitivity
obtained is 100% means all the nucleotides in the coding regions predicted correctly.
Gene Accession No AF061327:
The specificity of 77% computed using the proposed algorithm is very good
comparing to the others [13]. High specificity signifies that the algorithm successfully
captured the non-coding regions i.e. introns.
Table 1 Predicted boundaries of exons for the selected genes. Figures in bracket enlist the original boundaries of exons as specified by NCBI
Sequence (NCBI ACCN. NO) Gene description Exon-1 Exon-2 Exon-3 Exon-4 Exon-5 Exon-6
AF071552 Single exon 428–1459 NA NA NA NA NA
(441–1313)
AF061327 Two exons 39–282 1333–1656 NA NA NA NA
(13–153) (1245–1604)
M62420 Three exons 278–591 1134–1220 2686–2908 NA NA NA
(185–566) (1112–1219) (2608–3044)
AF059734 Four exons 59–466 1016–1424 1712–1865 1975–2185 NA NA
(335–491) (1296–1495) (1756–1857) (1953–2051)
AF099922 Five exons 912–1147 2455–2948 4062–4428 5470–5703 7429–7738 NA
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based …

(928–1039) (2528–2857) (4114–4377) (5465–5644) (7265–7605)

HAMHG5 Six exons 4108–4919 5269–5348 5903–6161 6173–6444 – 9252–9682
(4028–4855) (5399–5516) (6098–6187) (6473–6596) (7354–7440) (9143–9437)
145
146 S. Kar et al.

Fig. 5 Performance analysis of different coding rules using various matrices

Table 2 Different evaluation

Gene Accn. No Discrimination Missing rate Wrong rate
parameters at exon level
measure
AF071552 1.66 0 1
AF061327 4 0 0.5
M62420 1.22 0 0
AF059734 0.72 0 0.2
AF099922 1.6 0 0
HAMHG5 0.65 0.17 0.4
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based … 147

Fig. 6 ROC plot of exon prediction method based on three different mapping techniques

Gene Accession No M62420:

For the gene M62420, all the evaluation parameters at nucleotide and exon level are
compared with other numerical representation method [10]. The miss rate and wrong
rate for this gene are zero suggesting the algorithm have identified all the exons and
introns correctly. Discrimination factor D > 1 signifies all exons identified without
ambiguity. The sensitivity of the designed algorithm for this gene is 98% which is
highest comparing to the previous findings [10].
Gene Accession No AF059734:
Proposed method derived significant improvement in exon prediction for the gene
AF059734 which contain four exons. From Zhang et al. study, it is clear that the best
method employing frequency of nucleotide occurrence generates sensitivity of 69%
and specificity of 32%, while the proposed method resulted in 78% sensitivity and
48% specificity [12].
Gene Accession No AF099922:
The findings from the proposed hydration energy produce results of sensitivity 88%
and specificity 93%. The proposed representation produces discrimination factor of
unity. When comparing to the other numerical methods like tri-nucleotide, entropy,
trigonometric mapping, our method derived excellent results [14–16].
148 S. Kar et al.

Gene Accession No D10914 (HAMHG5):

The gene is chosen due to its diverse nature of exons. The gene contains very short
exons (less than 100 bp) as well as long exons (close to 1000 bp). The gapping of
the exons in the gene is not symmetric. Hence, effectiveness of the implemented
method could be tested on this gene. The result obtained through this gene is very
satisfactory as all the exons in the gene captured successfully.

5 Conclusions

Finding coding regions of various genes using DSP requires more in-depth study
as there are still scope of improvement in every aspect. Numerical representation of
DNA sequence along with genomic signal processing algorithms can identify hidden
periodicities, nucleotide distribution and feature extraction which cannot be revealed
by conventional methods. Very recently machine learning, deep learning and neural
network models are gaining popularity in this domain of research. These techniques
are very much accurate and developing rapidly. Hidden features of DNA and genes
can be identified using these techniques when used simultaneously with digital signal
processing methods and thus can be considered in future studies.

References

1. Anastassiou D (2001) Genomic signal processing. IEEE Sig Process Mag 18(4):8–20
2. Yu N, Li Z, Yu Z (2018) Survey on encoding schemes for genomic data representation and
feature learning—from signal processing to machine learning. Big Data Mining Anal 1(3):191–
210
3. Kar S, Ganguly M, Ghosal S(2021) Prediction of coding region and mutations in human
DNA by effective numerical coding and DSP technique. In: 2021 international conference on
computing, communication, and intelligent systems (ICCCIS), 19 February 2021. IEEE, pp
180–185
4. Vaegae NK (2020) Walsh code based numerical mapping method for the identification of
protein coding regions in eukaryotes. Biomed Sig Process Control 58:101859
5. Voss RF (1992) Evolution of long-range fractal correlations and 1/f noise in DNA base
sequences. Phys Rev Lett 68(25):3805
6. Feng B, Sosa RP, Mårtensson AK, Jiang K, Tong A, Dorfman KD, Takahashi M, Lincoln
P, Bustamante CJ, Westerlund F, Nordén B (2019) Hydrophobic catalysis and a potential
biological role of DNA unstacking induced by environment effects. Proc Natl Acad Sci
116(35):17169–17174
7. Monajjemi M, Ketabi S, Zadeh MH, Amiri A (2006) Simulation of DNA bases in water:
comparison of the Monte Carlo algorithm with molecular mechanics force fields. Biochem
Mosc 71(1):S1-8
8. Proakis JG (2001) Digital signal processing: principles algorithms and applications. Pearson
Education India
Spectral Analysis of DNA on 1-D Hydration Enthalpy-Based … 149

9. El-Badawy IM, Gasser S, Aziz AM, Khedr ME (2015) On the use of pseudo-EIIP mapping
scheme for identifying exons locations in DNA sequences. In: 2015 IEEE international confer-
ence on signal and image processing applications (ICSIPA), 19 October 2015. IEEE, pp
244–247
10. Meher JK, Panigrahi MR, Dash GN, Meher PK (2012) Wavelet based lossless DNA sequence
compression for faster detection of eukaryotic protein coding regions. Int J Image Graph Sig
Process 4(7)
11. Akhtar M, Epps J, Ambikairajah E (2007) On DNA numerical representations for period-3
based exon prediction. In: 2007 IEEE international workshop on genomic signal processing
and statistics, 10 June 2007. IEEE, pp 1–4
12. Zhang R, Zhang CT (1994) Z curves, an intutive tool for visualizing and analyzing the DNA
sequences. J Biomol Struct Dyn 11(4):767–782
13. Mabrouk M (2017) Advanced genomic signal processing methods in DNA mapping schemes
for gene prediction using digital filters. Am J Sig Process 7(1):12–24
14. Singh AK, Srivastava VK (2020) A tri-nucleotide mapping scheme based on residual volume of
amino acids for short length exon prediction using sliding window DFT method. Netw Model
Anal Health Inform Bioinform 9(1):1–3
15. Das B, Turkoglu I (2018) A novel numerical mapping method based on entropy for digitizing
DNA sequences. Neural Comput Appl 29(8):207–215
16. Das L, Nanda S, Das JK (2019) An integrated approach for identification of exon locations
using recursive Gauss Newton tuned adaptive Kaiser window. Genomics 111(3):284–296
A Novel Ensemble Model for Breast
Cancer Diagnosis

Aalif Hamid

Abstract Breast cancer is one of the world’s second-leading causes of death. One out
of every eight women experiences this illness at some point in their lives. Patients
with early detection of the disease have a higher likelihood of survival and have
a better chance of recovery. There is a critical requirement to classify tumors for
malignancy. This study investigates automatic breast cancer prediction using a variety
of machine learning techniques. The most prevalent type of classification is binary
(benign cancer versus malignant cancer). Machine learning (ML) techniques are now
being employed to detect breast cancer. They have a high degree of classification
accuracy and a wide range of diagnostic capabilities. This research work presents a
novel ensemble model for binary classification of breast mass tumors on the WDBC
dataset. The outcomes of seven distinct machine learning individual models (logistic
regression, KNN, SVC, etc.) are compared to the hybrid ensemble model of the
above seven techniques. The evaluation of the model is done in terms of numerous
performance indicators, such as accuracy, precision and recall. Compared to stand-
alone models, the results demonstrate that the ensemble model performs remarkably
well, with an accuracy of 0.98%.

Keywords Machine learning · Ensemble · Breast cancer · Support vector

machine · Stacking classifier

1 Introduction

Cancer is a chronic condition caused by excessive growth and cell division that affects
the cellular structure. It causes accelerated cell growth in some cases, depending on
cancer type, and in others, cell growth is significantly decreased. This division abnor-
mality affects the body in numerous states, resulting in cancers, undermining the
immune system, and other significant complications. A record 15.5 million persons
with a history of cancer were residents of the United States alone, according to the

A. Hamid (B)
Department of Information Technology, Central University of Kashmir, Ganderbal, Kashmir, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 151
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_12
152 A. Hamid

2018 report. According to the World Health Organization (WHO) [19], the illness
will have a massive effect on the planet. About one in five individuals will face
cancer diagnosis once in their lifetime by 2020. 18.1 million persons worldwide had
cancer in 2018, and 9.6 million died of the disease. The figures will likely double
by the year 2040. Accelerating global cancer containment by early diagnosis, detec-
tion, care and observation is the need of the hour. Pulmonary cancer is the most
frequently diagnosed cancer (11.6%), followed by female breast cancer (11.6%) and
bowel cancer (10.2%). The study’s primary objective is to apply machine learning
to diagnose the disease early, potentially saving millions of lives worldwide. The
use of several machine learning techniques such as decision trees, artificial neural
networks (ANN) and support vector machines (SVM) in cancer diagnosis yields posi-
tive outcomes. The adoption of machine learning in the medical field will transform
disease diagnosis and help doctors get better insights into the disease.
Breast cancer is the most common type of cancer in women [2], affecting about
2.1 million women each year and resulting in the highest number of cancer-related
fatalities. Breast cancer claimed the lives of around 15% of cancer-related deaths in
2018 or 627,000 women. Research conducted by [1] shows that one woman dies every
13 min due to breast cancer, while one case of breast cancer gets diagnosed every
two minutes worldwide. Since the shift in clinical practice from cure-based evidence
medicine to care-based evidence medicine, disease detection and identification have
been given top importance while still in the early stages of development. Breast
cancer death rates can be reduced if early identification of the disease is possible.
An essential and often used diagnostic method is the triple-test; it consists of three
medical examinations, i.e., self-examination, ultrasonography and fine needle biopsy
(FNB) of the breast used to produce high accuracy in the disease diagnosis. FNB is
an invasive technique that consists of obtaining material from the tumor directly. The
examination of the content is performed under the microscope for the occurrence of
cancer cells.
Data mining is a discipline in which computer science and analytics converge to
detect correlations in the database. The data mining process’s fundamental goal is to
retrieve and shape valuable information from the data file into a coherent framework
for potential use. It employs a combination of strong analytical abilities, domain
knowledge and an explicit knowledge base to uncover hidden trends and patterns
in heterogeneous information [7]. Many forensic organizations, pharmacies, hospi-
tals and research institutions have many medical diagnosis data available. To make
the system computerized and rapid to identify diseases, it is barely necessary to
classify them [9]. Forecasting trends can aid in the development of new treatments
by assisting in the identification of high-risk individuals based on known patholog-
ical and experimental risk factors. Accurate breast cancer risk assessments can help
with lifestyle changes, immunotherapy, personalized screening and risk-stratified
follow-up therapy [2]. The suggested research project is a study to see if a breast
tumor is benign or malignant. In this regard, machine learning has been applied to
the dataset. Machine learning is a series of instruments that are used to build and
test algorithms that enable inference, identification of patterns and classification.
Machine learning focuses on four steps: data collection, a compilation of models,
A Novel Ensemble Model for Breast Cancer Diagnosis 153

model training and model verification [20]. The association between breast cancer
and machine learning is not new. It has been used for decades to diagnose cancers and
other autoimmune disorders, predict gene mutations that trigger cancer and evaluate
the prognostic. Data mining and machine learning techniques are widely used in the
medical industry since they are discovered to be quite useful in medical practitioners’
decision-making processes. The implementation of multiple machine learning algo-
rithms such as support vector machine (SVM), artificial neural network (ANN) and
decision trees has shown promising results in the field of cancer detection. In addition
to transforming illness identification, the adoption of artificial learning in medical
imaging can enable doctors to obtain more in-depth insights into the disease.
The triple-test is an essential and frequently used diagnostic method; it consists
of three medical tests, i.e., self-examination, ultrasonography and breast fine needle
biopsy (FNB) used to diagnose the disease to achieve high precision. FNB is an
invasive technique that consists of directly obtaining tumor material. The contents
are checked for the presence of cancer cells under the microscope. The study work
uses the WDBC dataset. The dataset consists of attributes that measure a digitized
description of a fine needle aspirate from the image of mammary mass (FNA). The
features were recorded by analyzing the characteristics of the cell nuclei under the
microscope.
The following is a description of the paper’s structure. The past work in breast
cancer diagnosis using machine learning approaches is briefly described in Sect. 2.
The materials and procedures employed in the research study are described in
Sect. 3. The work’s experimentation results are discussed in Sect. 4, while the paper’s
conclusion is discussed in Sect. 5.

2 Related Work

Large number of researches has been done in the past decade on the diagnosis of
breast cancer. Some researchers have used mammograms, histopathological images,
while others have worked on calculative data. The ability of the machine learning
algorithms to find new hidden patterns in a large amount of data has shown quite
great results in the healthcare sector. Cancer diagnosis at an early stage increases
the chances of recovery in patients. For the diagnosis of breast cancer, researchers
use a variety of machine learning methods such as artificial neural networks (ANN),
decision trees, support vector machines (SVM), convolutional neural network and
random forest. On the WDBC dataset, SVM and ANN were applied, and the results
of the applied machine learning (ML) techniques were compared using performance
measures.
In comparison with ANN, SVM demonstrated the best accuracy of 96% based on
the performance criteria of the employed ML approaches [8]. The dataset WDBC
was used to test a feed forward backpropagation neural network. The ANN in this
study comprises nine input and one output neuron, with 1, 2, 3, 4 and 5 hidden layers.
The results demonstrate that using the TANSIG transfer function, a 20-neuron feed
154 A. Hamid

forward backpropagation single hidden layer neural network using the TANSIG
transfer function achieves the highest classification accuracy (98.9% on training set
and 99% on test set) [10]. The methodology of adaptive feature selection was used,
and classification was enabled using KNN, random forest and multilayer perceptron
[22]. Results proved that before feature selection KNN, SVM and PNN obtained an
accuracy of 100%, 100%, 96%, respectively, and after selection of best attributes,
the accuracy got jumped to 97.19% for PNN. Breast Cancer Coimbra dataset was
used on ANN, ELM, KNN and SVM algorithms by [23]; ELM is same as that of
artificial neural network with one difference, and ELM consists of only one hidden
layer instead of many hidden layers in case of ANN. Results showed that ELM
outperformed the other two algorithms in terms of accuracy and training time and
is preferable for more samples. The research [11] used six different algorithms,
such as AdaBoost, random forest, extra tree classifier and the stacking classifier,
on basic learners on the WDBC dataset and obtained an accuracy of 92.9% on
voting classifier, with extra trees classifier receiving the highest score of 95.1739%,
followed by XGBoost 95.1691% and AdaBoost 94.7343%. AlexNet architecture was
used by [24] on the CBIS-DDSM dataset in which the CLAUDE algorithm for image
enhancement was done with image segmentation to increase the sample size of the
dataset. In the proposed work, linear SVM obtained the highest accuracy of 80.5%.
Table 2 shows the brief summary of various methodologies adapted by researchers
for breast cancer.

Table 1 Various breast

Database Database size (GB) Total patients
cancer datasets
MIAS 2.3 161
DDSM 6 2620
CBIS-DDSm 70.5 237
ISPY1 76.2 237
Breast-MRI-NACT-Pilot 19.5 64
QIN-Breast 11.286 67
Mouse-Mammary 8.6 32
QIN Breast DCE-MRI 15.8 10
BREAST-DIAGNOSIS 60.8 88
RIDER Breast MRI 0.401 5
BCDR 1.4 GB 1734
TCGA-BRCA 53.92 (TB) 1098
BreakHis 4 82
Inbreast 40 MB 115
Coimbra Breast Cancer 7 Kb 116
WDBC 122 Kb
A Novel Ensemble Model for Breast Cancer Diagnosis 155

There is a growing abundance of researchers drawing their interest attentions to

ensembles. Moreover, it has solid proof that they can substantially change the clas-
sification performance. Six machine learning techniques were employed to analyze
8942 patients with breast cancer using the local dataset [12]. All methods, including
decision trees, neural networks, extreme boost, logistic regression and support vector
machines, produced very close accuracy of 79.80%, 82.7%, 81.7%, 81.7% and
81.7%, respectively, with random forest outperforming the others with an accuracy
of 82.7%. Kadam et al. [25] suggested a feature ensemble model for categorizing
breast cancer tumors into benign and malignant tumors based on sparse autoen-
coders and soft-max regression. In the analysis, ten-fold cross-validation was used
that gave an impressive 98.60% result and showed that the proposed model performs
other state-of-the-art techniques. The research presented by [26] is focused on the
diagnosis of breast cancer using an ensemble learning algorithm based on SVM to
decrease the diagnosis variance and increase the diagnosis accuracy. Twelve separate
SVMs are hybridized using the weighted field under the receiver operating charac-
teristic curve ensemble (WAUCE) approach. Wisconsin breast cancer, Wisconsin
diagnostic breast cancer and the control, epidemiology and end outcomes (SEER)
program breast cancer datasets have been studied to determine the efficacy of the
suggested model. The results show that the WAUCE model achieves higher accu-
racy with a little reduced variance for breast cancer detection when compared to five
additional ensemble mechanisms and two conventional ensemble models, namely
adaptive boosting and bagging classification tree. The results show that the WAUCE
model achieves higher accuracy with a little reduced variance for breast cancer detec-
tion when compared to five additional ensemble mechanisms and two conventional
ensemble models, namely adaptive boosting and bagging classification tree. The
datasets for breast cancer research are listed in Table 1.

3 Materials and Methods

The Wisconsin diagnostic dataset (WDBC) of breast cancer is used to measure

the effectiveness of the proposed ensemble model for breast cancer diagnosis. This
dataset has been collected by the University of Wisconsin Hospitals, Madison [27].
The dataset is available for download from the UCI machine learning repository. It
is composed up of data derived from a digitized description of a fine needle aspirate
of breast mass (FNA). The properties of the nuclei of the cells seen in Table 3 are
identified by the characteristics. The dataset comprises a total of 32 attributes, of
which 30 were used in the experiment as independent variables. There are a total of
569 instances of the patients recorded in the dataset.
Preprocessing is the method to remove the unwanted components from the dataset
to fit the machine learning algorithms perfectly. WDBC dataset contains one indepen-
dent categorical variable under the attribute name, “diagnosis”. This column contains
values in the form of benign or malignant (B/M). But the machine learning algorithm
takes the input in the form of numbers (0/1); so, this column is preprocessed before
156 A. Hamid

Table 2 Brief overview of diverse techniques for cancer diagnosis

Author Technique Dataset Performance
[13] Extracted characteristics using deep CNN BreakHis Accuracy = 97.5%
architectures were proposed as a novel
framework for the categorization of breast
cancers
[14] VGG-16 as a feature extractor combined MIAS-DDSM Accuracy = 90.5%
with NN classifier for detection
[15] Implemented CNN along with other WDBC Accuracy = 99.67%
machine learning algorithms
[16] The (SVM) model with radial basis function Coimbra Breast Accuracy = 93.9%
(RBF) kernel appeared to be the most Cancer
successful classifier, according to the
performance metrics of multicollinearity
assessment on machine learning models
[17] Using a deep neural network with support Local Dataset Accuracy = 97.21
value (DNNS), a novel technique to identify
breast cancer was presented
[18] For the diagnosis of breast cancer, an WDBC AccuracyELM = 95.6
extreme learning machine model with a SVM = 88.9
radial basis function (RBF) kernel was used
[3] Used convolutional neural networks, as well DDSM Recall = 97.3%
as several preprocessing strategies such as
contrast scaling, dilation, cropping and
decision fusion using an ensemble of
networks, to tackle the challenge of breast
microcalcification classification

Table 3 Features of WDBC

Feature Feature name
dataset [5]
Feature 1 Patient ID
Feature 2 Diagnosis (B/M)
Feature 3 Nuclei radius
Feature 4 Nuclei texture
Feature 5 Nuclei perimeter
Feature 6 Nuclei area
Feature 7 Nuclei smoothness
Feature 8 Nuclei compactness
Feature 9 Nuclei concavity
Feature 10 Nuclei concave Points
Feature 11 Nuclei symmetry
Feature 12 Nuclei fractal dimension
A Novel Ensemble Model for Breast Cancer Diagnosis 157

fetching it into the algorithms. Label encoder which is available under Scikit-Learn
package is used for the processing of the dependent variable. The dataset is divided
using an 80:20 ratio, i.e., 80% of the data is used for algorithm training, while 20% of
the data is the unseen data used to assess the performance assessment algorithms as
suggested by [11]. Seven different machine learning algorithms like logistic regres-
sion, SVM extra trees classifier, Gaussian NB classifier, KNN, SGD classifier, etc.,
have been implemented in the research work. Combination of the said algorithms
proved very beneficial and gave outstanding results. Implementation of work is done
on the WDBC dataset which is preprocessed first and split into testing and training
data.

3.1 Implementation of Work

The capability of computers to train from experience without being instructed person-
ally is machine learning [21]. Both rules are coded in the standard programming style,
and the computer can produce output depending on the logical argument, i.e., the
rules. However, more rules need to be written as the structure gets complicated, so
it becomes unfit to manage. However, it is assumed that machine learning can fix
this dilemma. The computers here learn how the input and output data are associ-
ated and then write a law accordingly. Any time a new data or condition occurs, the
programmer does not need to write new rules. The main focus of machine learning
is learning and inference. Machine learning is graded into supervised, unsupervised
and reinforcement learning [6]. In this research work, supervised machine learning
algorithms such as logistic regression, random forest and K-nearest neighbor (KNN)
are implemented.
Logistic Regression
By incorporating data onto a straight line, linear regression algorithms predict the
values. It is said that these algorithms are unbounded since the value is not simply
between 0 and 1. So, they are not ideal for problems with grouping, giving rise to
“Logistic Regression”. Logistic regression is a statistical model that is better adapted
for problems of classification. “These algorithms use a mathematical function called
“Sigmoid” to squeeze a number between 0 and 1.Two logistic regression algorithms
were implemented in this research work, one with the solver equal to “saga” and the
other with the solver parameter equal to “lbfgs”.
Random Forest Classifier
The classifier averages the number of classifiers (decision-tree classifiers) on various
data sub-samples. To boost prediction accuracy, it uses averaging of all classifiers.
A parameter called "Criterion" is used in the random forest classifier. It is a feature
that tests the efficiency of the division. This split parameter performs the task of
splitting the data into clusters and should be so that a single class (0 or 1) outweighs
158 A. Hamid

each category. By applying variability, random forest models minimize the risk of
overfitting by:
Constructing many trees (n-estimators)
Drawing substitute findings (i.e., a bootstrapped sample)
The best split nodes are split between a random subset of features chosen for each
node.
A random forest classifier is implemented in the ensemble model with
n_estimators set to 10 and criterion parameter set to “gini”.
Extra Trees Classifier
This puts together a variety of randomized decision trees on various data sub-samples.
The average estimate of individual classifiers is the approximation of this ensemble.
We used two extra tree classifier models with distinct parameters in our ensemble
model, as described in the previous classifier. Extra trees is like random forest. It
creates several trees and breaks nodes using random subsets of characteristics, but
with two key differences: It does not bootstrap observations (meaning it tests without
replacement), and nodes are broken on random splits rather than best splits. An extra
tree classifier is implemented in the ensemble model with the same n_estimators as
the random forest classifier and criterion parameter set to “gini”.
KNN
It is among the straightforward and non-parametric techniques for classification that
stores and groups all data based on some similarity function (e.g., distance). The
plurality vote of neighbors controls the classification/grouping of any data point.
The number of neighbors eligible for voting is determined by the K. In this research
work, K is set to 5.
Support Vector Classifier
It is a supervised learning technique which is most commonly used to solve clas-
sification problems. Data points reflect the dataset in the SVM. SVM constructs
hyperplanes that have the highest margin in multi-dimensional space to categorize
the results. A margin represents the longest interval between the closest data points.
The linear kernel is proven to show promising results. SVC with the linear kernel is
used in the model.
Gaussian NB Classifier
This model utilizes the Gaussian Naïve Bayes algorithm to predict/forecast the result.
The constant values connected to each attribute are considered to have a Gaussian
distribution here, i.e., Gaussian is said to be the likelihood of attributes. All default
values are used for this algorithm in the model.
SGD Classifier
Stochastic gradient descent (SGD) is a simple but powerful optimization algorithm
used to find the parameter/feature coefficient values that minimize a cost function.
A Novel Ensemble Model for Breast Cancer Diagnosis 159

In other words, it is used under convex loss functions such as SVM and logistic
regression for discriminatory learning of linear classifiers. Since the update to the
coefficients is done for each training instance, it has been successfully extended to
large-scale datasets rather than at the end of cases. Each parameter is set to its default
values for this algorithm in the proposed work.
Ensembling Approach
The research work combines the eight different sets of algorithms to boost the perfor-
mance of the model. All the algorithms are merged and passed to the stacking classi-
fier (voting classifier). A voting classifier is a machine learning model that trains on
a range of diverse models and selects an output or class based on its most excellent
chance of determining the class as the output. It essentially aggregates the outcomes
of each classifier passed into the stacking classifier and, based on the most significant
majority of votes, forecasts the output class. Rather than creating individual-specific
models and finding each of them’ performance, we build a specific single model
that trains these models and forecasts output based on their cumulative plurality of
votes for each type of output. It supports two types of voting: hard voting and soft
voting. In hard voting, the predicted result class is a class with the most significant
majority of votes in hard voting, i.e., the class that was most likely to be predicted by
one of the classifiers. In soft voting, the output class is the forecast for soft voting,
based on the average likelihood given to that class. Hard voting is implemented in
the research work. The ensemble model consisting of eight algorithms surpasses the
individual model performance and achieves an accuracy of 98.2%. The category of
stand-alone models surpasses the individual models in cancer diagnosis, classifiers
and gain competitive outcomes (Fig. 1).

Fig. 1 Proposed ensemble model

160 A. Hamid

4 Results and Discussion

4.1 Results Using Individual Classifiers

Each algorithm’s confusion matrix is calculated. Following data preprocessing, the

performance of the classifiers is shown using several performance parameters like
accuracy, F1 score, recall, precision, etc.
The confusion matrix for the actual and forecast classes is expressed in true posi-
tive (TP), which indicates that the original and predicted classes were correctly clas-
sified [4]. True negative (TN) is incorrectly classified. False positive (FP) indicates
that something has been misclassified. FN stands for false negative, which indicates
that something has been misclassified. All classifiers’ performance is evaluated using
these formulas.
TP + TN
Accuracy = (1)
(TP + TN + FP + FN)
TP
Precision = (2)
(TP + FP)
TP
Recall = (3)
(TP + FN)
 
Precision ∗ Recall
F1-score = 2 ∗ (4)
Precision + Recall

Before testing the model on the hybrid ensemble, the dataset is passed to each of the
seven machine learning algorithms, i.e., logistic regression, random forest classifier,
etc., to test the performance in terms of different evaluation metrics. Hybrid ensemble
of the different algorithms is combined and passed to the stacking classifier the results
of the experimentation on various algorithms along with the stacking classifier are as
shown in Table 4. Ensemble model surpasses the individual models on the WDBC
dataset with the accuracy of 0.98% for the prediction of breast cancer on different
independent features of the dataset like radius, smoothness, concavity, etc.

5 Conclusion

Breast cancer is one of the most deadly diseases, impacting millions of women
worldwide. As a result, any advancement in cancer illness diagnosis and prediction
is vital for healthy living. Machine learning-based illness detection has proven to be
quite valuable for the early detection of a variety of fatal diseases. This study uses
the WDBC dataset for breast cancer diagnosis using a novel ensemble method. In
the research work, seven different machine learning models are hybridized and sent
A Novel Ensemble Model for Breast Cancer Diagnosis 161

Table 4 Comparison of individual model with the ensemble model

Algorithm Accuracy F1-score Recall Precision
Logistic regression 0.95 0.95 0.95 0.94
Solver = “lbfgs”
Logistic regression 0.89 0.88 0.86 0.91
Solver = “saga”
Random forest classifier 0.96 0.95 0.95 0.96
Extra tree classifier 0.95 0.95 0.94 0.95
KNN 0.94 0.94 0.94 0.94
SVC 0.95 0.95 0.95 0.94
Gaussian NB classifier 0.93 0.93 0.93 0.93
SGD classifier 0.90 0.90 0.90 0.90
Stacking classifier 0.98 0.98 0.98 0.96

to a stacking classifier for malignancy diagnosis. The proposed hybrid ensemble was
found to be more effective than the individual model like KNN, extra tree classifier,
SVC, etc.
In future work, feature selection can be utilized to determine the most relevant
features to provide to the algorithms and investigate their impact on the outcomes.
The findings can also be tested against a range of breast cancer databases. In addition
to breast cancer, the suggested ensemble model can be used to diagnose a variety of
diseases. Moreover, in a future study, numerous optimization and parallel computing
approaches can increase the model’s performance.

References

1. The official NVIDIA blog. https://blogs.nvidia.com/blog/category/deep-learning/. Last

accessed 2020/08/08
2. Ming C, Viassolo V, Probst-Hensch N, Chappuis PO, Dinov ID, Katapodi MC (2019) Machine
learning techniques for personalized breast cancer risk prediction: comparison with the BCRAT
and BOADICEA models. Breast Cancer Res 21(1):75
3. Sert E, Ertekin S, Halici U (2017) Ensemble of convolutional neural networks for classification
of breast microcalcification from mammograms. In: 39th annual international conference of
the IEEE engineering in medicine and biology society 2017, EMBC, IEEE, Korea
4. Mushtaq Z, Yaqub A, Hassan A, Su SF (2019) Performance analysis of supervised classifiers
using PCA based techniques on breast cancer. In: International conference on engineering and
emerging technologies 2019, ICEET, pp 1–6. IEEE, Pakistan
5. Osman A, Aljahdali H (2020) An effective of ensemble boosting learning method for breast
cancer virtual screening using neural network model. IEEE Access 8(39):165–39174
6. Abdullah AJ, Hasan TM, Waleed J (2019) An expanded vision of breast cancer diagnosis
approaches based on machine learning techniques. In: International engineering conference
2019, IEC. IEEE, Iraq, pp 177–181
162 A. Hamid

7. Prabadevi B, Deepa N, Krithika LB, Vinod V (2020) Analysis of machine learning algo-
rithms on cancer dataset. In: 2020 international conference on emerging trends in information
technology and engineering 2020, IC-ETITE, India, pp 1–10
8. Bayrak EA, Kırcı P, Ensari T (2019) Comparison of machine learning methods for breast cancer
diagnosis. In: 2019 scientific meeting on electrical-electronics and biomedical engineering and
computer science 2019, EBBT, Turkey, pp 1–3
9. Islam MM, Haque MR, Iqbal H et al (2020) Breast cancer prediction: a comparative study
using machine learning techniques. SN Comput Sci 1:290
10. Osmanović A, Halilović S, Ilah LA, Fojnica A, Gromilić Z (2019) Machine learning techniques
for classification of breast cancer. In: Lhotska L, Sukupova L, Lacković I, Ibbott G (eds) World
congress on medical physics and biomedical engineering 2018. IFMBE proceedings, vol 68/1.
Springer, Singapore
11. Chaurasia V, Pal S (2020) Applications of machine learning techniques to predict diagnostic
breast cancer. SN Comput Sci 1:270
12. Ganggayah MD, Taib NA, Har YC, Lio P, Dhillon SK (2019) Predicting factors for survival
of breast cancer patients using machine learning techniques. BMC Med Inform Decis Mak
9(1):48
13. Khan S, Islam N, Jan Z, Din IU, Rodrigues JJPC (2019) A novel deep learning based framework
for the detection and classification of breast cancer using transfer learning. Pattern Recogn Lett
125:1–6
14. Guan S, Loew M (2017) Breast cancer detection using transfer learning in convolutional neural
networks. In: 2017 IEEE applied imagery pattern recognition workshop, AIPR, USA
15. Khuriwal N, Mishra N (2018) Breast cancer diagnosis using deep learning algorithm. In:
International conference on advances in computing, communication control and networking,
ICACCCN
16. Rahman MM, Ghasemi Y, Suley E, Zhou Y, Wang S, Rogers J (2020) Machine learning based
computer aided diagnosis of breast cancer utilizing anthropometric and clinical features, IRBM
17. Vaka AR, Soni B, Sudheer Reddy K (2020) Breast cancer detection by leveraging machine
learning. ICT Exp 6(4):320–324
18. Mojrian S, Pinter G, Joloudari JH, Felde I, Szabo-Gali A, Nadai L, Mosavi A (2020) Hybrid
machine learning model of extreme learning machine radial basis function for breast cancer
detection and diagnosis; a multilayer fuzzy expert system. In: RIVF international conference
on computing and communication technologies 2020, RIVF. IEEE, Vietnam, pp 1–7
19. World Health Organization. https://www.who.int/cancer/prevention/diagnosisscreening/bre
ast-cancer/en/#:~:text=In 2018, it is estimated, in nearly every region globally. Last accessed
2020/08/08
20. Amrane M, Oukid S, Gagaoua I et al (2018) Breast cancer classification using machine learning.
In: Electric electronics, computer science, biomedical engineerings’ meeting 2018, EBBT, pp
1–4
21. Bhardwaj R, Nambiar AR, Dutta D (2017) A study of machine learning in healthcare, vol 2.
In: IEEE 41st annual computer software and applications conference 2017, COMPSAC, Italy,
pp 236–241
22. Shargabi B, Shami F (2019) An experimental study for breast cancer prediction algorithms,
vol 12. In: Proceedings of the Second international conference on data science, e-learning and
information systems 2019. Association for Computing Machinery, pp 1–6
23. Aslan M, Celik Y, Sabanci K et al (2018) Breast cancer diagnosis by different machine learning
methods using blood analysis data. Int J Intell Syst Appl Eng 6(4):289–293
24. Ragab DA, Sharkas M, Marshall S et al (2019) Breast cancer detection using deep convolutional
neural networks and support vector machines. PeerJ 7
25. Kadam V, Jadhav S, Vijayakumar K (2019) Breast cancer diagnosis using feature ensemble
learning based on stacked sparse autoencoders and softmax regression. J Med Syst 43(8)
A Novel Ensemble Model for Breast Cancer Diagnosis 163

26. Wang H, Zheng B, Yoon S et al (2018) A support vector machine-based ensemble algorithm
for breast cancer diagnosis. Eur J Oper Res 267(2):687–699
27. Mangasarian O, Street W, Wolberg W (1995) Breast cancer diagnosis and prognosis via linear
programming. Oper Res 43(4):570–577
A Novel Smart Aging Approach
for Monitor the Lifestyle of Elderlies
and Identifying Anomalies

Mohammed Shakil Malek, Pooja Gohil, Sharnil Pandya, Anand Shivam,

and Kuldeep Limbachiya

Abstract For the past two decades, there is a paradigm shift in the field of automation
(IoT), and it is affecting several aspects of life. “Activity Modeling” is a part of
such automation, and it has several applications. One such application of “activity
modeling” is in the field of distant well-being investigation and nursing. In proposed
study, modeling and analysis of everyday activities of elderly persons are done by
“recording of periodic behavior pattern of elderly people” on daily basis. Today, the
challenge is to develop such a device that requires least human touch and provides
automation in irregularity detection along with alarming arrangement in it. Presently,
the investigation focuses on the diagnosis of elderly people’s day-to-day behavioral
patterns by observing their daily basis routine tasks from time, context, and location
point of view. For the sake of precision, numeral sensing and actuator components are
installed at retirement homely places. Through literature review, an exclusive sensing
blend technique has been recommended in order to record physical, social, seasonal,
and well-being-related notes about humdrum activities for elderly persons. As a part
of research methodology, an innovative routine-based activity-learning system is
recommended that can monitor relative data annotations of countless positions at an
elderly home and can aware the custodians if any case of glitch recognition happens.
Observations for 20 activities have been recorded for one month in two of elderly
smart homes situated in the city having sample size of more than 5000. Analysis
leads to the results that legalize the precision and the proficiency of the projected
structure.

Keywords Health monitoring · Activity modeling · Behavior pattern · Sensor

networks

M. S. Malek (B)
F. D. (Mubin) Institute of Engineering and Technology, Gandhinagar, Gujarat, India
P. Gohil
Gujarat Technological University, Ahmedabad, Gujarat, India
S. Pandya · A. Shivam · K. Limbachiya
Symbiosis Institute of Technology, Symbiosis International (Deemed) University, Pune, India
e-mail: sharnil.pandya@sitpune.edu.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 165
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_13
166 M. S. Malek et al.

1 Introduction

Recently, the trend of increasing number of nifty shelters across the world and
computerization of everyday livelihood actions in people’s life drags attention to the
need for development of remote health monitoring systems. Amidst this rapidly rising
era of Internet, where reciprocal communication has become customary, allowance
for the involvement of users, far-flung well-being nursing is meant to aim a huge
crowd of spectators at old-aged homes, which are now facing poor quality of health
facilities [1]. In traditional far-flung well-being nursing activity, mainly, two tactics
were been used: First is the employment of touch body beams to quantify physique
constraints and second is the use of sensor systems in nifty homes for action modeling,
which conveys information via report to therapeutic specialists for their opinions
and verdicts [25–30]. Although the prime issue with initial tactic is that it subsidizes
toward troublesomeness of elderly inhabitants, while the second one generally tallies
with an outside healthiness professional, which in turn upsurges the repetitive well-
being outlays [31–33]. As a result, there is demand in the field to develop such a
device that requires least human touch and provides automation in glitch recognition
along with watchful arrangement in it [2, 34–37]. The projected structure largely
aims toward analysis of interactive patterns of aged individuals by observing their
daily routine tasks being monitored at regular intervals of day, week, month, and year
through an outline acknowledgment-based mechanism culture models and underpin-
ning erudition methods. In present study, an innovative tactic is used to carry our
daily well-being nursing of aging people and to distinguish, sense, and categorize
irregularity situations by giving instant notice to the well-being specialists [3, 38–
44]. The anticipated research work is further alienated into quatern segments: (i)
proposal and experimental setup (ii) activity modeling procedure (iii) outcomes and
debate (iv) impending augmentations.

2 Structure Planning

As portrayed below, entire well-being nursing structure planning comprises of five

different layers, individually having its own practical position [45–49]. The first layer
being physical layer consists of diverse hardware components, while next being
the edge layer where events are classified into different categories. Moving on to
processing layer through cloud layer, one can reach last and ultimate layer which
is layer of application. Here, the final interactions take place between the elderly
persons and the caretakers and medical specialists [50–56].
I. Physical Layer/Level—it is a lower most level in structure planning. It
contains the individual hardware setups such as beams, ESP8266, conditioning
circuits, and Raspberry Pi 3B+. Here, data are generated after detecting the
situation and transferring it to the next layer [45–49, 57–61].
A Novel Smart Aging Approach for Monitor the Lifestyle … 167

II. Edge Layer/Level—in this layer, concept of edge computing is utilized. The
information is received from the beams using Wi-Fi, and system accomplishes
the prior processing of information on micro-computer and then send it to the
cloud layer [62–66]. The sub-components over here are numerous perception
elements inside each room, and motion trials provided form base for motion
identification by a classifier of an event. Now, information is shifted to cloud
level using Internet [67–70] where MQTT is used rather than secure socket
layer/level (SSL) for encoding safely data among MQTT clients and Raspberry
Pi 3B+.
III. Cloud Layer/Level (CL)—the third level of health monitoring structure plan-
ning is cloud layer. This is made up of various segments for the function of
data verification, instantaneous catalogue, file loading, and ML Kit for coating
appliance culture on engendered C.S.V [71–74]. Here, firebase is used as
BAAS: back end as a service. Facts are initially collected in catalog, wherein
through cloud jobs, C.S.V files are generated and are put in a store unit. Later
on, appliance erudition prototypical is functional for activity gratitude [74–77].
IV. Processing Layer/Level—The C.S.V files generated in CL are put as an
input in this layer. Here, the established mechanism erudition prototypical
is executed on C.S.V files, leading toward activity gratitude, activity frame-
work peers, and also mapping results [78–80]. This activity pattern generated
by this is used in glitch recognition for instantaneous scenarios, as that is prime
objective of this entire arrangement. By the conclusion of daily work, report
is ready based on the pattern created through the arrangement. This report is
then utilized for far-flung well-being nursing by the caretakers.
V. Application Layer/Level—lastly, comes the practicality part. The last layer of
the system is application layer, wherein the entire far-flung well-being nursing
system relates thru caretakers, on basis of report engendered. Report creation at
regular interval aids corresponding caretakers to deeply observe elder people’s
well-being without any help from remedial staff. This indirectly reduces the
expense spent on elder person’s health. System also alarms caretakers if there
is detection of anomaly; thus, if required, doctor can be consulted well in time
by sharing the reports [80–82]. Figure 1 represents a structure planning of the
proposed system.

2.1 Action Health Nursing

The main motive of this study arrays from the delivery of low-altitude statistics to
sensors followed by addition of superior quality information which in turn transfers
information via combination of data-based and information-driven methodologies
[4]. Latest periodicals have to recognize the effort on its own, as a revenue of digging
high quality of data. However, general policy for all is as follows: They should be
seen through non-professional activities like “cooking food,” “bathing,” or “watching
television while sitting on a sofa” [5, 80–82]. The moment our tasks are appropriately
168 M. S. Malek et al.

Fig. 1 Structure planning

acknowledged and automatically fixed, a variety of services are accessible just like
medicinal emergencies, expert tips on overall lifestyle, pre-diagnosis, and physician
assistance. Few of the hands-on examples are as follows: drop recognition system,
mobile backup answer system and everyday well-being monitoring activities, and
endorse lively existence resources [6, 80–83]. The novel facts so generated from
beams, accommodates to a greater cause, serving mavens with treasured info to
detect irregularities and ease affected ones.

2.2 Irregularity Detection

Apart from social modeling, looking for behavior change “anomaly detection” is
likewise significant and challenging job too. Variations in regular living behavior
are stalked through irregularity detection, and the intricacy of anomalous data is not
considered to be the same as anomalous remoteness behavior [7, 84]. Recognition
variations include the variations in several facets of a state like work order, strategy,
fitness position, period overbearing, and so on [8, 85]. Fundamentally, communica-
tive finding is contest of gaging humanoid behavior from beam figures and irregularity
recognition to check the way to identify behavioral alterations that may contain best
pilot of a model to arrest a regular routine. The dual practices to feel behavioral
changes are as follows: (i) comment (ii) divider [9]. The comment practice empha-
sizes on overall behavior and reflects any new admissions that are erratic with the
traditional average by means of an irregularity. Discernment reads discrepancy info
on or afterward former figures and also examines meant at the similar strategy as
new input result records to replicate differences. The comment plan is supplementary
reachable by way of an anomaly facts are fewer observable trendy definite period,
giving sample learning instances.
A Novel Smart Aging Approach for Monitor the Lifestyle … 169

Though irregularity discovery is a novice among further approaches, it works in

a crafty household and is vastly reputable front-line supplementary areas like inter-
ference gratitude, fake credit, negotiator recognition, commercial feature gratitude,
picture dispensation, and manuscript data [10, 86]. A number of appliance knowledge
methodologies are being used for irregularity detection like segmentation (resultant
law, neuronal network, Bayesian, SVM), k-nearest neighbor, bunching, totaling, and
evidence philosophy [11, 87]. Now-a-days, an audacious homely base atmosphere,
variation recognition is been used broadly in safety awareness of senior citizens
and assistive expertise for brain damage as lunatic patients. Statistics can be deter-
mined by given methods: (i) irregularity (ii) relative anomaly (iii) cohesive anomaly
[12, 88].

2.3 Daily Living Activities

Daily living activities refer to belongings of our day-to-day life, say for an example,
grooming, household chores, feeding, bathing, dressing, etc. These aerobics describe
the capability near existing popular isolated families unconventionally.
ADL is crucial inhabited and vital humanoid workouts for people’s everyday
timetables including drinking, consumption, dressing, nodding, washing, and so on.
They are attentive in two time primary assemblies, basic chores of everyday living.
IADLs are diverse preps that are not substantial continually [13, 89]. Notwith-
standing, IADLs ease the elderly and hampered people, for stuff like, household
work, getting food ready, body toning, clearing, shopping, vocation, and pressing.
The greeting of human workouts in enclosed situations is bothersome undertaking to
achieve. ADL focuses on nursing the challenges and noticing reactions for thoughtful
human crusade in intense environments [14, 89]. Persistent circumstances like nifty
homes have stimulated everyday drills finding, facilitated associates with household
ventures just to help incapacitated persons and elderly people for leaving peacefully
and with liberty [15, 89].

2.3.1 Activity Beams for Daily Living

Basically, three kinds of beam are available along with the strategies for a person’s
act acknowledgment. They are as follows.

Physical Environment-Based Beams

Proximity, RFID, Zig-Bee, and Wi-Fi can be applied for spotting the overtone amid
the individuals and flora surrounding them. Environment-based beams utilize raw
data after detected objects to assume act-attempted past people [16, 90]. The carried
170 M. S. Malek et al.

plans detect act over constituents and their linking with substances. The informa-
tion which is collected with the help of universal beams are then mentioned to a
neighborhood server for added conduct.

Wearable Beams

Wearable beams like accelerometers, inertial beams, electro-mechanical switches,

pedometers gyroscopes, and goniometers are body-appended beams and are viewed
as broadly known devices aimed by individuals’ deed acknowledgment. They are
exploited to see actions of human being apart from the track of help for the
people salutation expansions [17, 91]. These devices are designed to continually
quantify physical and biomechanical information of the human body. Examining
such available information can help in distinctive humanoid exercises in usual
living besides making a construal of them in a central edifice employing enterprise
acknowledgment.

Remaining Activity Recognition Beams

Camera stranded gadgets go on broadly secondhand for social act acknowledg-

ment inside a restricted detection enclosure. These beams rely arranged the cameras
footage as well as cinematic provisions toward perceive hominoid movement by
the means of PC image calculations [18, 92]. Audiovisual instruments like RGB-D
video, diffusion pictures measurement gadget, are basic kind of chromatic gadgets
largely recognized in living being’s movement acknowledged by decent nod magni-
tudes. Though these devices are expensive, they do have high vivacity consumption
and furthermore need visit up keep and stand uncovered near shield linked alarms
[19, 93].

Intellectual Computation

Intellectual computation idiom is often accustomed to express technological philoso-

phies and podiums that are fundamentally built upon evolving arenas of artificial intel-
ligence together with sign dispensation. These daises encompass all latest mechanism
knowledge and in-depth learning zones varying from ordinary language dispensa-
tion, virtual reality, and cardinal signal processing for acoustic sorting to speech
appreciation [18, 85, 88, 92].
In general, cognitive computing means all the modern hardware and software
setups that are meant to thought-provoking an overall human mind functionality in
order to support the policymaking by individuals and eliminate the logic of uncer-
tainty in the structure [18, 59, 79, 85]. Cognitive computing systems integrate and
blend the evidence from several technical causes, keeping into account the circum-
stantial hefts, contradictory sources, and information in order to propose and acclaim
A Novel Smart Aging Approach for Monitor the Lifestyle … 171

Fig. 2 Architecture of intellectual computation [84]

a superlative key at the end of process. For accomplishing efficiency and precision,
intellectual computation integrates all possible kinds of self-education and evolving
skills based on diverse categories like “Natural Language Processing” (NLP), statis-
tics outline recognition, and fact mining with the purpose of successfully copying
the functioning of human mind as well as behavior. [19, 50, 62, 76] As a matter of
fact, the utilization of computers to resolve the issues, which are generally faced by
humans, gives rise to inevitable need of hefty amount of structured and non-structured
information which is fed to countless apparatus learning and in-depth learning set
of rules. Over the period of time, intellectual computation has evolved and progres-
sive methodology of categorizing forms so as to be proficient in anticipating and
giving corresponding feasible solutions [20, 31, 41, 53, 66, 84]. Figure 2 represents
the architecture of the intellectual computation. According to cognitive computing
consortium, in order to gain growing competences, intellectual systems must have
following five attributes:
Contextual: To have a thorough familiarity of background of a problem is necessary
for problem-solving and is crucial in thought progression. One has to understand and
categorize the contextual facts, then recognize and fetch elements such as processes,
rules and regulations, handler contour, respective domain, and assigned aims and
errands [21, 27, 35, 49]. These processes will extract and utilize various unstructured,
structured, sensory, imagery, and audio–visual data [58, 69, 71, 80].
Interactive: Human–computer interactions (HCIs) are considered as main compo-
nent of intellectual systems. Therefore, one of the vital necessities is to ensure that
end handlers must bestowed link with systems with ease and effectively both for
172 M. S. Malek et al.

input as well as a feedback based on growing environmental circumstances under

consideration [20, 47–55, 62–71, 81–85]. Diverse know-hows incorporated within
the arrangement must bestowed link with other external devices, micro-processors,
and far-off cloud podiums [21, 50, 60, 75].
Adaptive: Intellectual schemes are targeted to assist policymaking systems. Hence-
forth, schemes must be assembled in a manner that permits policymakers to iteratively
gather material and have elasticity unified into them, allowing end users to learn as
an informational data vagaries and goalmouths of schemes as a whole evolves. Thus,
schemes must slog mainly on instantaneous data and make alterations to previously
defined schemes consequently [21, 91–93].
Stateful and Reiterative: Intellectual computation systems and know-hows can be
competent too and altered to distinguish and tactic complications by collecting excess
data by enquiries and sensory instantaneous inputs incase accessible state evidence
is deficient or vague in nature to reach any final verdict. In such circumstances,
structures can address problem by use of any alike incident that has occurred in past
[21, 84, 91–93].

3 Proposal and Experimental Setup

In an experimental setup, all the sensors are positioned at different places in the
house, with an ESP8266 along with Wi-Fi connected. Here, basic data dispensation
occurs, and then, the administered data are transported and charted into Raspberry Pi
B+ server. To dodge doubling and flawed data, sensor information is pre-processed at
ESP8266 level; thereby, safeguarding that Raspberry Pi is solitarily used as resident
server, wherein fresh data are stored in overall setup. Charted data are then sent to
cloud storage at regular intervals so as to ensure that backup of data is saved in case
of any system shutdown, along with the logs of entire system’s working, which can
also be used at future stage for restoring the system.
PIR Sensor—it is utilized to detect human motion in a chamber. It chiefly senses
the movement of a human being and sends data to cloud database (CDB) via local
network.
IR-Sensor—it is alike individual’s perceptive senses. It recognizes the restraints and
send information to CDB where some hitches are classified like when an entrance is
closed it identifies a hindrance.
Hall Magnetic Sensor—this sensor perceives a magnetic field in surroundings. When
the sensor notices a magnetic field, it gets charged and sends data to CDB.
Pressure Mat Sensor—it is used to notice the pressure on any area particularly for
the sitting places like chair, bed, and sofa. If it detects some pressure, it becomes
lively and sends data to CDB.
A Novel Smart Aging Approach for Monitor the Lifestyle … 173

Temperature–Humidity Sensor (dht-11)—it senses temperature and humidity level

in a chamber. Generally, it takes average data of 1 min and send it to the CDB.

4 Outcomes and Debate

In this segment, we debate about the two datasets that represent the outcomes for
various happenings being logged in a house through fitted sensory items for ADL of
elder people, for cross-validation of an anticipated methodology.

4.1 DataSet for Activities of Different House Sensors

Data are collected for a house with single elderly occupant for a duration of about a
week. Sensors installed in the architecture are encompassing sensors that are discreet
in nature, with an added benefit of cheaply available in market. In the conducted
experiments we have placed various temperature and humidity sensing units such
as DHT11 in different corners of the smart home. Accordingly, doors and cupboard
sensors are placed for corresponding data assortment.
Table 1 shows different types of activities mapped for a house with solo elderly
occupant. All the given activities have their discrete sensors that map an individual’s
data for a duration of 1 week.
Figure 3 portrays data for humidity in atmosphere outside a house. It can be
observed that due to rainfall at regular interval, by the end of July month, the humidity
level reaches up to the level of 95%.
Figure 4 illustrates the usage of AC with respect to duration. As one can see,
AC usage is maximum on 13th and 16th day of July month, while its minimum
on 15th and 17th day of July month. This is due to effect of various factors like
hall temperature, climate temperature, humidity in atmosphere, time of attendance
in house and health state on corresponding week days.
Figure 5 represents frequency of opening a cupboard in a particular day. Actually,
this data help in governing if any activity is ongoing inside bedroom, or person is
simply sleeping or reading.

Table 1 Data of mapped

Type of activity Start date Finish date
activities
Hall temperature 13/07/2019 19/07/2019
Climate temperature 13/07/2019 19/07/2019
Internal temperature of a house 26/07/2019 01/08/2019
Humidity in atmosphere outside a 26/07/2019 01/08/2019
house
Bedroom 13/07/2019 19/07/2019
174 M. S. Malek et al.

Fig. 3 Measure of humidity in atmosphere

Fig. 4 Usage of AC

4.2 ADL Measurement

Data produced by the installed sensors are both cardinal as well as referend in nature.
For instance, information created by hall magnetic sensor is cardinal and duple in
nature and is in form of 0 or 1 only which indicates non-active and active states,
A Novel Smart Aging Approach for Monitor the Lifestyle … 175

Fig. 5 Frequency of cupboard opening in a day

Table 2 ADL mapping

Type of activity Start date Finish date
Dining 13/07/2019 19/07/2019
Relaxing 13/07/2019 19/07/2019
Sleeping 13/07/2019 19/07/2019
Bathing 13/07/2019 19/07/2019

respectively. In the same way, the data produced by pressure mat is also cardinal
and duple in nature. Activities done by aged persons in a house are read from a
sensor reading only by applying apt logic which differs for each and every activity.
In all, total 14 activities are recorded by the physical grid setup which includes
relaxing, bathing, sleeping, dinning, reading, cooking, opening, and closing of doors
and cupboards of a house, watching television, lavatory usage, and sitting on chairs
and sofas. Each recorded activity has start date, finish date, duration, and location
committed to it as depicted in Table 2. Figure 6 represents daily activities of elderly
persons.

4.3 Comparison of Older and Younger People’s Data

Equated to younger ones, elder person’s health when gets deteriorate, their toilet
visits reflect a significant rise from a usual pragmatic pattern. As recorded for about
one week, during sickness in elders, i.e., on 17th and 18th July, there is noticeable
176 M. S. Malek et al.

Fig. 6 Daily activities of elderly persons

change in their sum of toilet visits. Figure 7 represents a comparison of elder and
younger person visits.

Fig. 7 Comparison of elder and younger person’s lavatory visits

A Novel Smart Aging Approach for Monitor the Lifestyle … 177

5 Impending Augmentations

The present work reflects the prevailing state of research for IOT by reviewing
presently relevant and accessible masses of literature and highlighting recent drifts
in the ongoing study field. The motive of the work is to intricate the current issues
faced by the researchers in development of remote well-being nursing systems useful
for prognostic well-being nursing, and how anxieties of confidentiality assault have
disappointed systems based on pictorial systems grounded on concurrent computer
visualization applications and thus motivate scholars to move forward to well-being
nursing using non interfering sensor grids. The current study proposes a distant well-
being nursing system reposed on a wireless sensor grid installed at an elderly resi-
dent’s house aiming to monitor health of aging people staying inside house and predict
in case of any irregularity based on analytical investigation of behavioral pattern of
concerned individual which can otherwise be done by using several machine learning
models such as advanced hidden Markov models and predictive time-series analysis.
Proposed arrangements can be successfully arrayed at various places such as
old-age homes where there are a smaller number of caretakers compared to the
elderly people residing over there; houses consisting of senior citizens staying alone,
hospitals dealing with health monitoring, etc. Over here in all the cases, distant and
prophetic health nursing can be a bonus as it is based on concept of edge computing
and no touch body radars henceforth no requirement for a person under deliberation
to put on sensor regularly. With the help of these sensors, medicinal personnel can
also endlessly observe the behavioral reports generated by them on his/her mobiles
or computers and can prescribe medicines based on past history of a patient under
observation.
Some of the pitfalls of the present device are flawed data created by falsely recti-
fied sensors, low prognostic power of current prevailing models, and prerequisite of
huge quantity of data before model begins to bounce its analytical results. However,
these all faults can be efficiently settled by using nearly advanced machine learning
models and making copies skilled of developing it selves based on instantaneous
situations and applying data collected at similar location for a group to be exploited
for analysis for same group till substantial quantity of data is obtainable for current
setup. Also, further exploration of the field of remote health monitoring is required
for upgrading existing system and further progress the distant well-being nursing
systems by integrating prevailing technologies in better ways.

References

1. Cadavid H, Garson W (2005) Towards a smart farming platform: from IoT-based crop sensing
to data analytics. In: Radford T (ed) The guardian. Springer. https://doi.org/10.1007/978-3-
319-98998-3_19. [Internet]. 2005 Available from https://www.theguardian.com/science/2005/
mar/30/environment.research
178 M. S. Malek et al.

2. Nandurkar S, Thool V, Thool R (2014) Design and development of precision agriculture system
using wireless sensor network. In: International conference on automation, control, energy and
systems (ACES), Hooghly
3. Andrew R, Malekian R, Bogatinoska D (2018) IoT solutions for precision agriculture. In:
MIPRO, Opatija
4. Benyezza H, Bouhedda M (2018) Smart irrigation system based ThingSpeak and Arduino. In:
International conference on applied smart systems, ICASS, Médéa
5. Mat I, Kassim M, Harun A (2015) Precision agriculture applications using wireless moisture
sensor network. In: IEEE 12th Malaysia international conference on communications, Kuching
6. Fountas S, Aggelopoulou K, Gemtos T (2016) Precision agriculture: crop management for
improved productivity and reduced environmental impact or improved sustainability. Supply
Chain Manag Sustain Food Netw
7. Miles C (2019) The combine will tell the truth: on precision agriculture and algorithmic
rationality. Big Data Soc 1–12
8. Malek S, Akalkotkar P (2014, December) Need of PPP in Indian road sector. Int J Emerg
Technol Appl Eng Technol Sci. In: Proceedings of the international conference on informa-
tion, knowledge and research in engineering management and sciences (IC-IKR-EMS), Kalol,
Gujarat, India. KITRC, pp 300–305
9. Saiyed F, Pitroda J, Malek S (2016, April) Assessment of very critical risk factors in real estate
sector by criticality index. In: Proceedings of the international conference on engineering:
issues, opportunities and challenges for development, Umrakh, Gujarat, India, SNPITRC, pp
193–205
10. Ghayvat H, Mukhopadhyay S, Gui X, Suryadevara N (2015) WSN-and IoT-based smart homes
and their extension to smart buildings. Sensors 15:10350–10379
11. Nautiyal V, Shah Y, Malek S (2019) Application of pavement management system in HDM-4:
a critical review. Int J Manag Technol Eng 9(4):1850–1859. https://doi.org/16.10089.IJMTE.
2019.V9I4.19.27237
12. Macwan J, Bhavsar A, Malek M (2019) Smart project scheduling for construction projects
using AI. Int J Tech Innov Mod Eng Sci 5(5):14–19
13. Agrawal A, Malek M (2018) Principles of sustainability in construction project management.
Int J Adv Innov Res 5(4):45–52
14. Gajera D, Malek M (2018) A Study on consumer behaviour in real estate for Vadodara city.
Univ Rev 7(12):956–969
15. Kumbhani H, Malek M (2018) Cost management process in construction industry. Suraj Punj
J Multi Res 8(12):151–158. https://doi.org/16.10089.SPJMR.2018.V8I11.16.3145
16. Upadhyay U, Malek M (2018) Identification of factors responsible for organizational transfor-
mation of GSRDC. Pramana Res J 8(12):189–201. https://doi.org/16.10089.PRJ.2018.V8I12.
18.2063
17. Jariwala K, Malek M (2018) Total quality management in construction of textile market in
Surat city. J Appl Sci Comput 5(12):699–706. https://doi.org/16.10089.JASC.2018.V5I12.453
459.1500186
18. Bachwani D, Malek M (2018) Parameters indicating the significance of investment options in
real estate sector. Int J Sci Res Rev 7(12):301–311. https://doi.org/16.10089.IJSRR.2018.V7I
12.287391.2950
19. Rathod S, Malek S (2018) Six sigma in construction industry: literature review. Int J Manag
Technol Eng 8(12):913–920. https://doi.org/16.10089.IJMTE.2018.V8I12.17.2102
20. Kochai Z, Malek M (2018) Causes of schedule delay and cost overrun in Afghanistan
infrastructure projects. Int J Adv Innov Res 5(3):18–25
21. Araya SY, Malek S (2018) Application of building information modelling in Indian construction
projects—a critical review. Int J Creative Res Thoughts 6(1):979–987
22. Tadewos SG, Malek S, Patel D (2018) Challenges of building construction in Mekelle city,
Ethopia: from contractor’s perspective. J Emerg Technol Innov Res 5(3):420–424. https://doi.
org/10.6084/m9.jetir.JETIR1803080
A Novel Smart Aging Approach for Monitor the Lifestyle … 179

23. Aynalem YF, Malek S (2018) Developing an effective project concept in 5D environment. J
Emerg Technol Innov Res 5(5):372–380. https://doi.org/10.6084/m9.jetir.JETIR1805660
24. Pandya A, Malek S (2018) A review for the causes of delay in Indian construction projects. Int
J Adv Eng Res Dev 5(3):1–6
25. Shah B, Malek S (2018) A recapitulation on exigency of smart villages in Indian ambience.
Int J Adv Eng Res Dev 5(3):1–8
26. Patel V, Malek S (2018) A review on identification of safety factors in construction projects.
Int J Adv Eng Res Dev 5(3):1–9
27. Parekh P, Malek S (2018) A critical literature review on implementing lean construction
concepts in infrastructure projects. Int J Adv Eng Res Dev 5(3):1–7
28. Chaudhari K, Malek S (2018) Effectiveness of public private partnership in infrastructure
development in India: a review. Int J Adv Eng Res Dev 5(3):1–12
29. Hada N, Malek S (2018) Identification of risk factors in government housing project: a review.
Int J Adv Eng Res Dev 5(3):1–6
30. Jethwa B, Bhavsar A, Malek S (2017) An overview of factors affecting risk management in
PPP based infrastructure projects. Int J Eng Dev Res 5(2):1737–1742
31. Jethwa B, Bhavsar A, Malek S (2017) Critical review on risk management in PPP based
infrastructure projects. Int J Eng Dev Res 5(1):357–361
32. Parmar D, Pitroda J, Malek S (2017) Use of BIM (Building Information Modelling) as an
integrated tool to plan, design and manage critical construction projects. Int J Adv Res Innov
Ideas Edu 3(1):1145–1153
33. Malek S, Saiyed F (2017) A study on incipient trends: the real estate sector of Ahmedabad city
(Gujarat). Int J Res Granthalayah 5(1):304–315
34. Patel D, Malek S (2016) Application of performance management system in infrastructure
sector of construction industry. Int J Eng Dev Res 4(2):1998–2000
35. Kalkani K, Malek S (2016) Analyzing causes of delay in construction projects. Int J Innov Res
Sci Technol 2(12):257–262
36. Upadhyay S, Malek S (2016) Adoption and implementation of total quality management in
construction industry. Int J Eng Dev Res 4(2):1256–1260
37. Patel K, Malek S (2016) Planning, scheduling and resource management for residential real
estate projects in Ahmedabad by using microsoft project. Int J Eng Dev Res 4(2):189–194
38. Malek S, Akalkotkar P (2016) Driving forces leading to the adoption of PPP: perspective from
Gujarat (India), Hong Kong and Australian practitioners. Int J Innov Res Sci Technol 2(9):6–11
39. Malek S, Akalkotkar P (2016) Factors contributing to successful public private partnership
projects for highway projects in Indian context. Int J Sci Res Dev 3(12):25–29
40. Malek S, Akalkotkar P (2016) Negative factors adopting public private partnership for highway
projects: a case study of Gujarat. Int J Sci Technol Eng 2(8):30–35
41. Malek S, Akalkotkar P, Zala L (2016) Gujarat PPP: attractive factors for adopting public private
partnership for highway projects. Int J Futuristic Trends Eng Technol 3(2):1–6
42. Saiyed F, Pitroda J, Malek S (2015) Reaping the returns: review on risk management in real
estate. Int J Futuristic Trends Eng Technol 3(1):1–8. Malek M, Mohibali S, Bachwani D (2021)
Identification, evaluation, and allotment of critical risk factors (CRFs) in real estate projects:
India as a case study. J Project Manag 6(2):83–92. https://doi.org/10.5267/j.jpm.2021.1.002
43. Malek MS, Gundaliya PJ (2020) Negative factors in implementing public–private partnership
in Indian road projects. Int J Constr Manag. https://doi.org/10.1080/15623599.2020.1857672
44. Malek M, Gundaliya P (2021) Value for money factors in Indian public-private partnership
road projects: an exploratory approach. J Project Manag 6(1):23–32. https://doi.org/10.5267/
j.jpm.2020.10.002
45. Malek M, Bhatt V, Patel A (2020) Global, local and national growth of road projects through
PPP. Test Eng Manag 83:25837–25860
46. Memon A, Malek M, Patel A (2020) A review on risk and risk management technique in
construction projects. Int J Disaster Recovery Bus Continuity 11(1):2695–2701
47. Patel R, Bachwani D, Malek M (2020) Human resource management, a vital tool for labour
management in construction industry. Stud Indian Place Names 40(9):201–206
180 M. S. Malek et al.

48. Parikh R, Bachwani D, Malek M (2020) An analysis of earn value management and
sustainability in project management in construction industry. Stud Indian Place Names
40(9):195–200
49. Shah K, Malek M, Phugat S (2020) Alternate building materials used in construction industry.
Stud Indian Place Names 40(9):184–189
50. Khan BR, Malek M, Tidke A (2020) Risk management in construction industry. Stud Indian
Place Names 40(9):165–171
51. Khan S, Malek M, Phugat S (2019) Factors influencing time and cost overrun in construction
of government buildings: Indore as a case study. CIKITUSI J Multi Res 6(4):288–295. https://
doi.org/16.10089.CJMR.2019.V6I4.19.2857
52. Pandya S, Sur A, Kotecha K (2020) Smart epidemic tunnel—IoT based sensor-fusion assistive
technology for COVID19 disinfection. Emerald
53. Patel NR, Kumar S (2017) Enhanced clear channel assessment for slotted CSMA/CA in IEEE
802.15.4. Wirel Pers Commun 95:4063–4081
54. Patel NR, Kumar S (2018) Wireless sensor networks’ challenges and future prospects. In: 2018
international conference on system modeling and advancement in research trends (SMART),
Moradabad, India, pp 60–65
55. Ghayvat H, Awais M, Pandya S, Ren H, Akbarzadeh S, Chandra Mukhopadhyay S, Chen C,
Gope P, Chouhan A, Chen W (2019) Smart aging system: uncovering the hidden wellness
parameter for well-being monitoring and anomaly detection. Sensors 19:766
56. Saket S, Pandya S (2016) An overview of partitioning algorithms in clustering techniques
57. Shah, JM, Kotecha K, Pandya S, Choksi DB, Joshi N (2017) Load balancing in cloud computing:
methodological survey on different types of algorithm. In: 2017 international conference on
trends in electronics and informatics (ICEI). https://doi.org/10.1109/ICOEI.2017.8300865
58. Ghayvat H, Pandya S, Shah S, Mukhopadhyay SC, Yap MH, Wandra KH (2016) Advanced
AODV approach for efficient detection and mitigation of wormhole attack in MANET. In: 2016
10th international conference on sensing technology (ICST)
59. Pandya S, Shah J, Joshi N, GhayvatH, Mukhopadhyay SC, Yap MH (2016) A novel hybrid
based recommendation system based on clustering and association mining. In: 2016 10th
international conference on sensing technology (ICST)
60. Patel S, Singh N, Pandya S (2017) IoT based smart hospital for secure healthcare system. Int
J Recent Innov Trends Comput Commun
61. Pandya SP, Prajapati MR, Thakar KP Assessment of training needs of farm women. Guj J Ext
Edu 25(2):169–171
62. Pandya S, Ghayvat H, Sur A, Awais M, Kotecha K, Saxena S, Jassal N, Pingale G (2020)
Pollution weather prediction system: smart outdoor pollution monitoring and prediction for
healthy breathing and living. Sensors 20:5448
63. Pandya S, Ghayvat H, Kotecha K, Awais M, Akbarzadeh S, Gope P Smart home anti-theft
system: a novel approach for near real-time monitoring and smart home security for wellness
protocol. Appl Syst Innov 1(4):42
64. Patel RR, Pandya SP, Patel PK Characterization of farming system in north west Agro climatic
zone of Gujarat state. Guj J Ext Edu 27(2):206–208
65. Pandya S, Ghayvat H, Kotecha K, Yep MH, Gope P (2018) Smart home anti-theft system:
a novel approach for near real-time monitoring, smart home security and large video data
handling for wellness protocol
66. Joshi N, Kotecha K, Choksi DB, Pandya S (2018) Implementation of novel load balancing
technique in cloud computing environment. In: International conference on computer commu-
nication and informatics (ICCCI)
67. Patel W, Pandya S, Mistry V (2016) i-MsRTRM: developing an IoT based intelligent medi-
care system for real-time remote health monitoring. In: 2016 8th international conference on
computational
68. Wandra KH, Pandya S (2012) A survey on various issues in wireless sensor networks. Int J Sci
Eng
A Novel Smart Aging Approach for Monitor the Lifestyle … 181

69. Saket JS, Pandya S Implementation of extended K-medoids algorithms to increase efficiency
and scalability using large dataset. Int J Comput Appl
70. Bhola YO, Socha BN, Pandya SB, Dubey RP, Patel MK (2019) Molecular structure, DFT
studies, Hirshfeld surface analysis, energy frameworks, and molecular docking studies of
novel (E)-1-(4-chlorophenyl)-5-methyl-N -((3-methyl-5-phenoxy-1-phenyl-1H-pyrazol-4-yl)
methylene)-1H-1, 2, 3-triazole-4-carbohydrazide. Molecular Cryst Liq Cryst
71. Patel WD, Pandya S, Koyuncu B, Ramani B, Bhaskar S (2019) NXTGeUH: LoRaWAN based
next generation ubiquitous healthcare system for vital signs monitoring and falls detection. In:
2018 IEEE Punecon
72. Dandvate HS, Pandya S (2016) New approach for frequent item set generation based on Mirabit
hashing algorithm. In: 2016 international conference on inventive
73. Swarndeep SJ, Pandya S (2016) Implementation of extended K-medoids algorithm to increase
efficiency and scalability using large datasets. Int J Comput Appl
74. Wandra K, Pandya S (2014) Centralized timestamp based approach for wireless sensor
networks. Int J Comput Appl
75. Garg D, Goel P, Pandya S, Ganatra A, Kotecha K (2002) A deep learning approach for face
detection using YOLO. In: 2018 IEEE Punecon
76. Sur A, Pandya A, Sah RP, Kotecha K, Narkhede S (2020) Influence of bed temperature on
performance of silica gel/methanol adsorption refrigeration system at adsorption equilibrium.
Part Sci Technol
77. Sur A, Sah RP, Pandya S (2020) Milk storage system for remote areas using solar thermal
energy and adsorption cooling. Mater Today Proc
78. Cohen JM, Pandya S, Tangirala K, Krasenbaum LJ (2020) Treatment patterns and character-
istics of patients prescribed AJOVY, Emgality, or Aimovig. In: HEADACHE
79. Cohen JM, Pandya S, Krasenbaum LJ, Thompson SF (2020) A real-world perspective of
patients with episodic migraine or chronic migraine prescribed AJOVY in the United States.
In: HEADACHE
80. Barot V, Kapadia V, Pandya S (2020) QoS enabled IoT based low cost air quality monitoring
system with power consumption optimization. Cybern Inform Technol
81. Ghayvat H, Pandya S, Patel A (2019) Proposal and preliminary fall-related activities recognition
in indoor environment. In: 2019 IEEE 19th international conference on
82. Akbarzadeh S, Ren H, Pandya S, Chouhan A, Awais M (2019) Smart aging system
83. Ghayvat H, Pandya S (2018) Wellness sensor network for modeling activity of daily livings—
proposal and off-line preliminary analysis. In: 2018 4th international conference on computing
84. Awais M, Kotecha K, Akbarzadeh S, Pandya S (2018) Smart home anti-theft system
85. Patel M, Pandya S, Patel S (2017) Hand gesture based home control device using IoT. Int J
Adv Rese
86. Pandya S, Yadav AK, Dalsaniya N, Mandir V Conceptual study of agile software development
87. Samani MD, Karamta M, Bhatia J, Potdar MB (2016) Intrusion detection system for DoS attack
in cloud. Int J Appl Inform Syst
88. Ayanelam YF, Malek S (2018, 16–17 February) A critical review of effective project manage-
ment using 5D BIM. In: Smart construction. 1st international conference on engineering and
technology: smart engineering, Parul University, Baroda, India. McGraw Hill Education, pp
155–158
89. Kochai Z, Malek S (2018, 16–17 February) An assessment of time and cost overrun in
Afghanistan’s construction. In: Smart construction. 1st international conference on engineering
and technology: smart engineering, Parul University, Baroda, India. McGraw Hill Education,
pp 159–166
90. Chaudhari K, Malek S (2018, 16–17 February) Critical review on delay and cost overrun in
construction projects. In: Smart construction. 1st international conference on engineering and
technology: smart engineering, Parul University, Baroda, India. McGraw Hill Education, pp
167–172
182 M. S. Malek et al.

91. Shah B, Malek S (2018, 28–30 June) Identification of significant factors affecting the design of
smart village. In: Proceedings of the 2nd international conference on women in science and tech-
nology: creating sustainable career, Birla Vishvakarma Mahavidyalaya, Vallabh Vidyanagar,
India
92. Saiyed F, Malek S (2017, 10–11 November) The realty growth trajectory of india and china-
inexorable comparison. In: Proceedings of the 2nd international conference on construction,
real estate, infrastructure and project management, NICMAR, Pune, India
93. Olatinwo S, Joubert T (2019) Enabling communication networks for water quality monitoring
applications: a survey. IEEE 7(100332)
Spatial Analysis of Violent Crime Dataset
Using Machine Learning

Falade Adesola, Ambrose Azeta, Sanjay Misra, Aderonke Oni, Ravin Ahuja,
and Ademola Omolola

Abstract The monster called crime has been living with us from the beginning of
human existence and impacts negatively on the general health of a nation. Different
approaches were employed in the past studies for predicting occurrence of violent
crime to aid predictive policing, which makes conventional policing more efficient
and proactive. This paper investigates the accuracy of Machine Learning-based crime
prediction approaches, which were used previously by other researchers. This study
presents Machine Learning approaches to violent crime prediction. Five years’ histor-
ical dataset between July 2014 and July 2019 were collected from Nigerian Police
Lagos, analyzed and used for training the models built. Two different Machine
Learning predictive models, Decision Tree and K-Nearest Neighbor, were imple-
mented using IBM Watson Studio and violent crime prediction accuracy of 79.65%,
and 81.45% were obtained, respectively, with the real-life dataset collected from
Nigerian Police Obalende Lagos and online crime reported portal during violent
crime prediction in Lagos. This could be used to enhance crime prevention and
control strategies in curbing the worrisome crime rate in the country.

Keywords Machine learning · Crime · Prediction

F. Adesola · A. Azeta · A. Oni · A. Omolola

Covenant University, Ota, Nigeria
e-mail: adesola.falade@covenantuniversity.edu.ng
A. Azeta
e-mail: ambrose.azeta@covenantuniversity.edu.ng
A. Oni
e-mail: aderonke.oni@covenantuniversity.edu.ng
S. Misra (B)
Ostfold University College, Halden, Norway
e-mail: sanjay.misra@covenantuniversity.edu.ng
R. Ahuja
Shri Viswakarma Skill University, Gurgaon, Hariyana, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 183
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_14
184 F. Adesola et al.

1 Introduction

Crime remains human security challenges across the globe. Different countries of
the world have to grapple to contain the rising wave of violent crime such as banditry,
kidnapping, rape, murder, and manslaughter which in some cases already assuming
a different dimension. According to [1, 2], Nigeria is currently found in the web of
crime manifesting in an upsurge of violent and non-violent crime. This worrisome
trend has made Nigeria notable with high crime rate in the world [3].
Presently, the crime is now ravaging the country like tsunami and blowing a
climate of anxiety and fear about our public safety. Negative effects of crime include
and not limited to the following: increased poverty among citizenry, unemployment,
and sudden death just to mention a few.
Following a report by National Bureau of Statistics, 2018, crime rate in Nigeria
is increasing on a yearly of 3.4% over the last thirty years. This should be a thing
of concern to all and sundry and demand urgent attention. Due to the increase in
violent crime such as armed robbery, kidnapping, rape, murder, manslaughter, and
aggravated assault across the length and breadth of the nation, anxiety of people in
the community have become intensified. Predictive policing is the way to go as it
is already being implemented in some countries of the world for example USA and
China [3, 4].
In the year 2010 and 2011, there were more reports of armed robbery incidents
in Nigeria compared to other years as shown in Fig. 1. Additionally, cult killings
were also reported in various parts of Nigeria between year 2009 and 2018, and it is
observed that year 2010, 2018, and 2019 had the deadliest cult killings in the country
[4].
Notable authors in crime prediction domain [2–4] have done a lot of research
work to improve on the techniques of crime prediction. Machine Learning and data

Fig. 1 Violent crime dataset summary for this study

Spatial Analysis of Violent Crime Dataset Using Machine Learning 185

mining approaches have been reportedly used by these authors. Their works have
received diverse attention looking at the attendance benefits of crime prediction which
is both beneficial to the police patrol team and the citizens at large. Occurrence of
violent crime brings about death and very serious injuries in some cases. One useful
application of crime prediction is that police patrols are undertaken by the police
depending on the predicted crime hotspots in the vicinity.
The paper is organized as follows. Section 2 provides the literature review.
Section 3 provides materials and methods. Section 4 provides the results and
discussion and finally conclusion drawn in Sect. 5.

2 Related Work

In [5], an effort to make useful contribution to crime prediction, divided the entire
region in consideration into grids and proposed a novel spatio-temporal crime
network prediction model that was based on Convolutional Neural Network for auto-
matic crime referenced features extraction. The model developed can forecast crime
risk in each region of the urban areas for the next day using high volume of dimen-
sional data. The model was evaluated and discovered to outperform other models in
the previous studies. However, the gap observed is it requires high dimensionality of
crime data to be able to make accurate prediction, which is the major shortcoming
of deep neural network.
Authors in [6] developed a linear regression model use for forecasting crime trends
in Bangladesh. Real crime dataset was collected from Bangladesh police website,
and this data contained a collection of different crime types. The linear regression
model developed was trained on the dataset, after which the model was used in
forecasting future crime trends. In the empirical evaluation carried out, the linear
regression model outperformed other three selected algorithms. The gap observed
here is it takes time to make crime prediction.
Authors [7], in a research work titled “Crime Prediction Using Twitter Senti-
ment and Weather” developed a crime prediction model that was able to predict the
location and time of occurrence of a particular type of crime by using sentiment anal-
ysis based on lexicon methods and understanding of categorized weather data. The
obtained crime prediction by combining sentiment analysis based on lexicon method
and Kernel Density Estimation using historical crime dataset via linear modeling
approach. The result of the experiment showed that the accuracy of the model devel-
oped surpasses the benchmark model, which makes prediction using Kernel Density
Estimation. The limitation of this approach is that it could not predict time slot and
crime location for the occurrence of crime.
186 F. Adesola et al.

2.1 K-Nearest Neighbor Classifier

KNN usually stores the available objects during classification and then use similarity
measure to classify new objects by seeking the nearest neighbor of the input values
[8]. It is a classifier that makes a prediction based on the majority vote of the k-nearest
samples on the feature vector space. KNN performs the following basic steps during
training:
• The distance is calculated
• The closest neighbor is sought after
• The labels are voted for.

2.2 Decision Tree Classifier

In building the model for Decision Tree, the datasets are normally divided into
smaller pieces to form the tree structure. During every singular stage of the process,
the feature that best fit and splits the dataset is chosen with the assistance of two the
important functions, namely information gain and Gini index. Gini index determines
the likelihood of a random sample being classified incorrectly using the equation in
(1)


k
IG ( p) = pi (1 − pi ) (1)
i=1

Information gain assists in deciding which the features in the dataset will split
next. Equation (2) is used in calculating the information gain using entropy.


k
H (T ) = I E = − pi log( pi ) (2)
i=1

3 Materials and Methods

Datasets used for this study were derived from the historical crime dataset of Nige-
rian Police Obalende Lagos and online crime reported portal. The violent crime
datasets between July 2014 and July 2019 were pulled out from the general crime
data made available for this research and this amounting to a total of 8234 turples
with 14 attributes. After data cleaning and diligent application of principal compo-
nent analysis, nine attributes which are relevant for violent crime prediction were
made available and then applied. The various important attributes in the dataset are:
Spatial Analysis of Violent Crime Dataset Using Machine Learning 187

Table 1 Violent crime

Crime type Sum of data collected Percentage (%)
dataset summary used for the
experiment Arm robbery 4984 60.5
Murder 2549 31.0
Willful manslaughter 82 1.0
Aggravated assault 146 1.8
Rape 273 3.3
Kidnapping 200 2.4
Total 8234

crime description, crime id, time of crime, date crime was committed, number of
deaths involved, type of crime, crime location, as well as crime hotspot. The first
seven attributes are the predictor or dependent variables and the last attributes repre-
sent the target variable or the class label. Presented in Table 1 is the violent crime
dataset summary pulled out between July 2014 and July 2019.
The violent crime dataset summary used for the study is also presented in Fig. 1.
Decision Tree and K-Nearest Neighbor algorithms were chosen for this study
because of the outstanding performances of the duo in literature and quick adaptation
to changes in dataset.
Watson Studio was used to realize the model implementation. Watson studio is
an integrated and collaborative platform for data scientists and business analysts
specifically built on open-source components and IBM added value tools [9]. It is
available both in the cloud and on premise. It is built around open-source components
such as Python, Scala, R, SQL, Apache Spark, and Notebooks (Jupyter and Zeppelin).
IBM added value tools such as Watson Machine Learning, Watson Natural Language
Processing, Watson Natural Language Understanding, SPSS Predictive analytics
algorithms, Chatbots, and many more are provided in Watson Studio. It is endowed
with set of tools to enable data scientists and business analysts to collaboratively and
work easily.
Therefore, the Machine Learning models for both K-Nearest Neighbor and Deci-
sion Tree were built in IBM Watson Studio [10, 11]. Python programming language
in Watson Studio was then used for violent crime prediction. Confusion matrix was
used to evaluate the performances of the models developed with evaluation results
presented in Table 2.

4 Results

During the empirical study, Decision Tree and K-Nearest Neighbor models or
predicting the target column were built. This was done after the dataset was split
into test and training set [12].
188 F. Adesola et al.

Table 2 Evaluation results of K-Nearest Neighbor classifier for different violent crime types
S/N Evaluation results for K-Nearest Neighbor for different violent crime types
Violent TP FP Accuracy Precision Recall F1 MSE
crimes rates rates (%) (%) (%) score
(%)
1 Armed 0.64 0.425 83.72 81.62 80.53 79.21 0.0179
robbery
2 Kidnapping 0.69 0.389 81.95 78.75 81.14 78.78 0.0626
3 Rape 0.68 0.578 81.82 80.02 78.99 79.75 0.0709
4 Aggravated 0.65 0.546 80.20 78.24 79.35 75.92 0.0478
Assault
5 Murder 0.67 0,255 79.45 77.75 76.55 81.89 0.0187
6 Ritual killing 0.67 0,555 81.55 71.15 80.15 78.39 0.0587

Evaluation Results of K-Nearest Neighbor Classifier for different violent crimes

is presented in Table 2, and the accuracy was found to be 81.45%.
Figure 2 presents K-Nearest Neighbor prediction results comparison graph for
different violent crime types.
Evaluation Results of Decision Tree Classifier for the violent crimes is presented
in Table 3 and the accuracy was found to be 76.69%.
Decision Tree classifier delivers 76.69% accuracy. Figure 3 shows Decision Tree
prediction results comparison for different violent crime types.
The Decision Tree classification has an accuracy of 76.69% in its prediction and
59% of the instances correctly classified, while 41% of the instances were incorrectly
classified.
The summary of the evaluation results based on accuracy and training time as the
standard evaluation techniques from literature [13, 14] is presented in Table 4.

Fig. 2 Prediction performance comparison of KNN for violent crimes

Spatial Analysis of Violent Crime Dataset Using Machine Learning 189

Table 3 Evaluation results of Decision Tree classifier for different violent crime types
S/N Evaluation results for Decision Tree
Violent TP FP Accuracy Precision Recall F1 MSE
crimes rates rates (%) (%) (%) score
(%)
1 Armed 0.78 0.425 78.98 77.62 81.53 81.22 0.0179
robbery
2 Kidnapping 0.77 0.389 77.25 81.75 83.24 83.78 0.0126
3 Rape 0.79 0.478 79.83 84.02 83.99 83.75 0.0109
4 Aggravated 0.76 0.546 76.20 73.24 76.35 74.90 0.0178
Assault
5 Murder 0.72 0,345 72.45 73.75 76.50 79.89 0.0543
6 Ritual 0.75 0.653 75.40 81.25 84.46 85.82 0.0645
killings

Fig. 3 DT prediction performance comparison between different violent crime types

Table 4 Evaluation results

S/N Metrics Decision Tree KNN
summary of the two models
1 Accuracy (%) 76.69 81.452
2 Training time (s) 217 209

Discussion
Several other Machine Learning algorithms were taken into consideration before
chosen Decision Tree and K-Nearest neighbor models for this study. This decision
was informed by the outstanding performances of both Decision Tree and K-Nearest
Neighbor models from literature and their quick adaptation to changes in dataset. The
models for both algorithms were developed using IBM Watson studio with Python
as the programming language. The preprocessed dataset was trained on the models
190 F. Adesola et al.

developed using the split of the training and the test sets. The results of the empirical
study reveal that Decision Tree delivers 76.69% predictive accuracy of the unknown
class labels. Also K-Nearest Neighbor model returns 81.45% predictive accuracy
of the unknown class labels. These two results are good enough for any predictive
system to rely on.
The dastard effects of violent crime include anxiety and sudden death among
citizenry, bad economy, poor quality of life, and increase in poverty rates among
others. During the experiment, over-fitting was avoided during training and testing
by engaging in cross validation [8–10, 15–17] Testing of hypothesis is not included
in the formulation of this model, rather Decision Tree and KNN Machine Learning
techniques were engaged in the formulation of the prediction model.

5 Conclusion

The study has further confirmed the efficacy and more outstanding performance of K-
Nearest Neighbor algorithm over Decision model with 81.45% predictive accuracy
and Decision Tree with 76.69% prediction accuracy in a new context. By drastically
reducing false positives and high true positives rates, a high accuracy was attained.
The use of modern method for violent crime prediction has again being displayed in
this empirical study. Combining multiple Machine Learning methods to see if they
could deliver better results in violent crime prediction may be considered as future
work.

Acknowledgements We appreciate the sponsorship and support from Covenant University for the
overall success of this study.

References

1. NSW Criminal Law Handbook (2017). Available online https://www.booktopia.com.au/nsw-

criminal-law-handbook-2017/book/9780455500058.html
2. Chen P, Yuan H, Shu X (2008) Forecasting crime using the ARIMA model. In: Proceedings
of the 5th IEEE international conference on fuzzy systems and knowledge discovery, vol 5, pp
627–630
3. Liao R, Wang X, Li L, Qin Z (2010) A novel serial crime prediction model based on Bayesian
learning theory. In: Proceedings of the 2010 IEEE international conference on machine learning
and cybernetics, vol 4, pp 1757–1762
4. Wang P, Mathieu R, Ke J, Cai HJ (2010) Predicting criminal recidivism with support vector
machine. In: Proceedings of the 2010 IEEE international conference on management and
service science, pp 1–9
5. Lian D, Tao H, En C, Jianfeng Z, Chao G (2017) Deep convolutional neural networks for
spatiotemporal crime prediction. In: International conference information and knowledge
engineering, IKE’17
Spatial Analysis of Violent Crime Dataset Using Machine Learning 191

6. Mohler G, Short M, Brantingham P, Schoenberg F, Tita G (2011) Self-exciting point process

modeling of crime. J Am Stat Assoc 106(493):100–108
7. Xinyu C, Youngwoon C, Suk Y (2015) Crime prediction using Twitter sentiment and weather.
In: 2015 systems and information engineering design symposium, 24–24 April 2015. https://
doi.org/10.1109/SIEDS.2015.7117012
8. Barnadas MV (2016) Machine learning applied to crime prediction. Thesis, Universitat
Politècnica de Catalunya, Barcelona, Spain, September 2016
9. Cortes C, Vapnik V (1995) Support vector network. Mach Learn 20(3):273–297. https://doi.
org/10.1023/A:1022627411411
10. Alves L, Ribeiro H, Lenzi E, Mendes R (2013) Distance to the scaling law: a useful approach
for unveiling relationships between crime and urban metrics. PLoS One 8(8):1–8. https://doi.
org/10.1371/journal.pone.0069580. PMID: 23940525
11. Ristea A, Leitner M (2018) Integration of social media in spatial crime analysis and prediction
models for events. In: AGILE 2018, Lund, 12–15 June 2018
12. IBM Watson studio documentation and resources. Available online https://dataplatform.cloud.
ibm.com/docs/content/wsj/getting-started/overview-ws.html
13. Adesola F, Azeta A, Oni A, Chidozie F (2019) Forecasting violent crime hotspots using a
theory-driven algorithm. Int J Eng Res Technol (IJERT). ISSN 09743154
14. Adesola F, Misra S, Omoregbe N, Damasevicius R, Maskeliunas R (2019) An IOT-based
architecture for crime management in Nigeria. In: Shukla R, Agrawal J, Sharma S, Singh TG
(eds) Data, engineering and applications. Springer, Singapore, pp 245–254
15. Nicholas-Omoregbe OS, Azeta AA, Chiazor IA, Omoregbe N (2017) Predicting the adoption of
e-learning management system: a case of selected private universities in Nigeria. Turk Online
J Distance Edu-TOJDE 18(2):106–121
16. Azeta AA, Misra S, Azeta VI, Osamor VC (2019) Determining suitability of speech-enabled
examination result management system. Wirel Netw 1–8
17. Isah AO, Alhassan JK, Misra S, Idris I, Crawford B, Soto R (2016, July) Network system design
for combating cybercrime in Nigeria. In: International conference on computational science
and its applications, pp 497–512
Analysis of Violent Crime Dataset Using
Support Vector Machine Model

Falade Adesola, Ambrose Azeta, Sanjay Misra, Aderonke Oni, Ravin Ahuja,
and Ademola Omolola

Abstract It is usually a challenging task predicting violent crime occurrences in

space and time. Huge dataset are needed for accurate prediction of future violent
crime occurrence, which in most cases were subjected to artificial intelligence or
statistical methods. Most studies from literature adopted data mining techniques
for violent crime prediction with some inherent limitation of accuracy as one of
the gaps that needed to be filled. The study applied support vector machine model
on the six different historical violent crime dataset gathered between July 2016
and July 2019 from Nigeria Police Lagos headquarter to predict spatio-temporal
occurrences of violent crime in the state. The six different violent crime dataset used
for the study are: armed robbery, rape, kidnapping, assault, murder and manslaughter.
The dataset was preprocessed and fed into the support vector machine model built
in Watson machine learning studio using python as a programming language. The
model returned 82.12% prediction accuracy, which is assumed to be good enough for
any prediction system. This result was evaluated using confusion matric, and tested
against some results from literature, and was found to out-perform some machine
learning models used in the previous studies. Based on this empirical study, the police
authority could adopt this model to strengthen violent crime prevention strategies in
order to mitigate violent crime occurrences in Lagos state, Nigeria.

F. Adesola · A. Azeta · A. Oni · A. Omolola

Covenant University, Ota, Nigeria
e-mail: adesola.falade@covenantuniversity.edu.ng
A. Azeta
e-mail: ambrose.azeta@covenantuniversity.edu.ng
A. Oni
e-mail: aderonke.oni@covenantuniversity.edu.ng
S. Misra (B)
Ostfold University College, Halden, Norway
e-mail: sanjay.misra@covenantuniversity.edu.ng
R. Ahuja
Shri Viswakarma Skill University, Gurgaon, Hariyana, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 193
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_15
194 F. Adesola et al.

Keywords Machine learning · Support vector machine · Violent crime · Prediction

1 Introduction

Crime has been in existence from time immemorial and remains the number one
enemy of humanity. Any act committed or omitted deliberately or otherwise that is
punishable under the law of the land is regarded as crime [1]. Crime impacts dastardly
on the economy of any nation and equally affect the quality of life of citizens.
Crime is more pronounced in the urban areas because of rapid population growth
and social infrastructural development. In other words, rapid population growth in
the urban cities has culminated into crime rate increase [2]. In addition, movement
of inhabitants from villages to cities has also lead to the increase in crime rate. It is
noteworthy that crime that is violent in nature has negative effect on the economic
development of any nation. It is therefore pertinent at this juncture to create a safe
and secured urban environment for everyone to live and benefit from. Any research
or efforts toward reducing crime rate are always a welcome development. All of
these are geared toward implementation of effective policing and police patrols to
reduce crime rate. Following a report by National Bureau of Statistics, 2018, crime
rate in Nigeria is increasing on a yearly average of 3.4% over the last thirty years.
This should be a thing of concern to all and sundry and demand urgent attention. Due
to the increase in violent crime such as armed robbery, kidnapping, rape, murder,
manslaughter and aggravated assault across the length and breadth of the nation,
anxiety of people in the community have become intensified. Predictive policing is
the way to go as it is already being implemented in some countries of the world for
example USA and China [3, 4].
In the year 2010 and 2011, there were more reports of armed robbery incidents
in Nigeria compared to other years as shown in Fig. 1. Additionally, cult killings

Fig. 1 Violent crime dataset summary for this study

Analysis of Violent Crime Dataset Using Support … 195

were also reported in various parts of Nigeria between year 2009 and 2018. It can
be observed that year 2010, 2018 and 2019 had the most deadly cult killings in the
country [5].
The paper is organized as follows. The second section provide the literature review.
The third section provide materials and methods. The forth section provides the
results and discussion and finally conclusion drawn in Sect. 5.

2 Literature Review

Previous research works by Chen et al. [2], Liao et al. [3], Wang et al. [4], Alves
et al. [6] on violent crime prediction attracted several attentions from literature by
considering the potential benefits of crime free society to the country. Different
machine learning algorithms were used by these authors to predict occurrence of
violent crime with inherent limitation of accuracy to be improved upon. Data mining
approached were equally engaged for dataset preparation, analysis and prediction.
In [7], authors developed a crime prediction model that was able to predict the
location and time of occurrence of a particular type of crime by using sentiment anal-
ysis based on lexicon methods and understanding of categorized weather data. They
obtained crime prediction by combining sentiment analysis based on lexicon method
and Kernel Density estimation using historical crime dataset via linear modeling
approach. The result of the experiment showed that the accuracy of the model devel-
oped surpasses the benchmark model which makes prediction using Kernel Density
Estimation a good choice. The limitation of this approach is that it could not predict
time slot and crime location jointly for violent crime occurrence.
In [8], an effort to make useful contribution to crime prediction, divided the
entire region in consideration into grids and proposed a novel spatio-temporal crime
network prediction model that was based on convolutional neural network for auto-
matic crime referenced features extraction. The model developed can forecast crime
risk in each region of the urban areas for the next day using high volume of dimen-
sional data. The model was evaluated and discovered to outperform other models in
the previous studies. However, the gap observed is it requires high dimensionality of
crime data to be able to make accurate prediction, which is the major shortcoming
of deep neural network.
Authors in [9] developed a linear regression model use for forecasting crime trends
in Bangladesh. Real crime dataset was collected from Bangladesh police website,
and this data contained a collection of different crime types. The linear regression
model developed was trained on the dataset, after which the model was used in
forecasting future crime trends. In the empirical evaluation carried out, the linear
regression model out-performed other three selected algorithms. The gap observed
here is it takes time to make crime prediction.
196 F. Adesola et al.

2.1 The Support Vector Machine (SVM)

Authors [10] discovered support vector machine model and was premised on struc-
tural risk management theory according to [11]. The author used decision boundaries
to define decision planes. Group of objects with diverse class membership were sepa-
rated. It works using constructed hyper-plane that uses linear models implementing
class boundaries that has non-linear attributes which engages input vectors with non-
linear mapping in high dimensional feature space [12]. The SVM has been engaged
in different domains for prediction ranging from system intruder’s detection, pattern
recognition, estimation of age, face recognition, as well as telecommunications
according to [13–16].

2.1.1 Methodology

Eight thousand, two hundred and thirty-four total dataset for violent crime
were extracted from the huge collection of overall crime dataset made avail-
able by the statistics department of Nigerian Police, Obalende Lagos. The rele-
vant attributes in the dataset are: crime description, crime_id, crimetimecommitted,
crimedate_was committed, numberofdeaths_involved, crimelocation as well as type-
ofcrime. Presented in Table 1 is the violent crime dataset summary made available
between July 2016 and July 2019 used for this study.
Summary of dataset used during the study is presented in Fig. 1.
Watson Studio was used to realize the model implementation. Watson studio is
an integrated and collaborative platform for data scientists and business analysts
specifically built on open source components and IBM added value tools [17]. It is
available both in the cloud and on premise. It is built around open source components
such as Python, Scala, R, SQL, Apache Spark and Notebooks (Jupyter and Zeppelin).
IBM added value tools such as Watson machine learning, Watson natural language
processing, Watson natural language understanding, SPSS predictive analytics algo-
rithms, Chatbots and many more are provided in Watson studio. It is endowed with
set of tools to enable data scientists and business analysts to collaboratively and
easily work to produce data analysis reports.

Table 1 Total violent crime

Crime type Sum of data collected Percentage (%)
dataset summary
Arm robbery 4984 60.5
Murder 2549 31.0
Wilful manslaughter 82 1.0
Aggravated assault 146 1.8
Rape 273 3.3
Kidnapping 200 2.4
Total 8234
Analysis of Violent Crime Dataset Using Support … 197

A total of eight thousand, two hundred and thirty-four (8234) violent

crime dataset with seven important attributes were used for this study. These
important attributes are: crime_description, crime_id, timeofcrime, dateofcrime,
numberofdeaths_involved, crimelocation as well as typeofcrime. The dataset was
preprocessed with independent variables and target columns identified. Then the
SVM model was built in Watson studio [17] with python as a programming language
for the purpose of training and testing with confusion matrix use for evaluation. The
model returned an accuracy of 82.12% across all the violent crime dataset used.
In addition, after engaging the stratified cross validation [4], 89.04% was classified
rightly and 10.96% was classified wrongly.

3 Results and Discussion

During the empirical study, the dataset was split into 75% training and 25% test set
according to [4], the target columns of violent crime hot spot location and time of
violent crime occurrence were predicted using support vector machine.
The prediction results for the support vector machine model on the preprocessed
dataset is presented in Table 2 and found to return an accuracy of 82.12%. This result
out-performed the results of some machine learning approaches on crime dataset
reported in literature.
Table 2 presents the empirical results of the support vector machine model trained
using Watson studio with Python as a programming language. The attribute crime-
location represents the violent crime hot spots area in the state, HotSpot and Predict-
edtime columns are the support vector machine prediction results. Hot_Spot value
0 represents happening event not possible of violent crime, while HotSpot value 1
is to predict the occurrence of violent crime at a particular time reported that must
warrant police intervention. The last column Predictedtime shows the likely time of
violent crime occurring. The evaluation results using confusion matrix of different
violent crime types is also presented in Table 3.
The support vector machine model used during the empirical study returned
82.12% accuracy. The results summary on different violent crime dataset is presented
in Fig. 2.

4 Discussion

During the machine learning comparative analysis stage, support vector machine
became a choice for this study because of its unique performances in different
domains from literature and its quick adaptation to new dataset. Watson studio was
engaged to develop the model using Python as a programming language. Both the
split training and test dataset were applied on the SVM model. Consequently, the
198 F. Adesola et al.

Table 2 The model sample

CrimeLocation HotSpot PredictedTime
results
10 0 0
11 0 0
12 0 0
13 1 8.101
14 1 16.101
30 1 11.101
31 1 8.101
32 1 14.101
51 0 0
52 1 6.101
53 0 0
54 0 0
32 0 0
34 0 0
35 1 19.122
60 1 9.122
61 1 10.122
62 1 6.251
55 1 9.122
63 1 11.201
64 1 9.201
81 1 15.201
82 1 9.201
83 0 0
90 1 9.201
91 1 9.201
57 0 0
92 0 0
93 0 0
84 1 10.001

experimented results showed the prediction of unseen class labels with 82.12% accu-
racy. This could represents a fairly good system that can premise on. This result
out-performed the work of [17] with 79.2% accuracy for support vector machine
using the same dataset.
The dastard effects of violent crime include anxiety and sudden death among
citizenry, bad economy, poor quality of life and increase in poverty rates among
others. During the experiment, over-fitting was avoided during training and testing by
Analysis of Violent Crime Dataset Using Support … 199

Table 3 Evaluation results of SVM classifier

Evaluation results for support vector machine
S/N Violent TP FP Accuracy Precision Recall F1 MSE
crimes rates rates (%) (%) (%) score
(%)
1 Armed 0.82 0.315 82.92 87.62 81.53 86.22 0.0227
robbery
2 Kidnapping 0.81 0.301 81.25 81.70 83.24 85.71 0.0326
3 Rape 0.79 0.478 82.43 83.12 82.92 85.75 0.0309
4 Aggravated 0.76 0.546 84.26 83.24 86.05 84.09 0.0278
Assault
5 Murder 0.82 0.345 82.45 83.75 86.50 81.89 0.0245
6 Ritual 0.78 0.655 79.41 81.25 83.66 83.82 0.0445
killings

Fig. 2 SVM prediction performance comparison between violent crime types

engaging in cross validation [18–21]. Consequently, the training and testing produces
an accuracy of 82.12%.

5 Conclusion

In the context of this study, support vector machine technique has demonstrated its
efficacy as a prediction method among other machine learning techniques. In this
study, support vector machine delivers 82.12% predictive accuracy across the six
violent crime dataset used. The use of SVM as a modern method for violent crime
prediction has again being proven. Combination of different artificial intelligence
200 F. Adesola et al.

techniques may be considered to see if they could deliver a better predictive accuracy
as a future work.

Acknowledgements We appreciate the sponsorship and support from Covenant University for the
support received during this study.

References

1. NSW Criminal Law Handbook (2017) Available online: https://www.booktopia.com.au/nsw-

criminal-law-handbook-2017/book/9780455500058.html
2. Chen P, Yuan H, Shu X (2008) Forecasting crime using the ARIMA Model. In: Proceedings
of the 5th IEEE international conference on fuzzy systems and knowledge discovery. vol 5, pp
627–630
3. Liao R, Wang X, Li L, Qin Z (2010) A novel serial crime prediction model based on Bayesian
learning theory. In: Proceedings of the 2010 IEEE international conference on machine learning
and cybernetics, vol 4, pp 1757–1762
4. Wang P, Mathieu R, Ke J, Cai HJ (2010) Predicting criminal recidivism with support vector
machine. In: Proceedings of the 2010 IEEE international conference on management and
service science, pp 1–9
5. Nigeria Watch Project (2019) Fourth report on violence in Nigeria (2009–2018). Retrived rom
www.nigeriawatch.org/media/html/NGA-Watch-Report14.pdf
6. Alves L, Ribeiro H, Lenzi E, Mendes R (2013) Distance to the scaling law: a useful approach
for unveiling relationships between crime and urban metrics. PLoS One 8(8):1–8. https://doi.
org/10.1371/journal.pone.0069580PMID:23940525
7. Xinyu C, Youngwoon C, Suk Y (2015) Crime prediction using twitter sentiment and weather,
2015 systems and information engineering design symposium, 24–24 April 2015. https://doi.
org/10.1109/SIEDS.2015.7117012
8. Lian D, Tao H, En C, Jianfeng Z, Chao G (2017) Deep convolutional neural networks
for spatiotemporal crime prediction. International conference information and knowledge
engineering | IKE’17
9. Cortes C, Vapnik V (1995) Support vector network. Mach Learn 20(3):273–297. https://doi.
org/10.1023/A:1022627411411
10. Chiu NH, Guao YY (2008) State classification of CBN grinding with support vector machine.
J Mater Proc Technol 201:601–605
11. Awal A, Rabbi J, Rana I (2017) Using data mining technique to analyze crime of Bangladesh.
IJCSN - Int J Comput Sci Netw 6(4):2277–5420
12. Elmi HE, Sallehuddin R, Ibrahim S, Zain AM (2014) Classification of sim box fraud detection
using support vector machine and artificial neural network. Int J Innovat Comput 4(2):19–27
13. Guo G, Fu Y, Dyer CR, Huang TS (2008) Image-based human age estimation by manifold
learning and locally adjusted robust regression. Trans Image Proc IEEE 17(7):1178–1188
14. Kumar M, Ghani R, Mei ZS (2010) Data mining to predict and prevent errors in health insurance
claims processing. In: ACM 16th international conference on knowledge discovery and data
mining, pp 65–74. Available from: https://doi.org/10.1145/1835804.1835816
15. Kirlidog M, Asuk C (2012) A fraud detection approach with data mining in health insurance.
Proc Soc Behav Sci 62:989–994. https://doi.org/10.1016/j.sbspro.2012.09.168
16. Anwar S, Zain JM, Zolkipli MF, Inayat Z, Khan S, Anthony B, Chang V (2017) From intrusion
detection to an intrusion response system: fundamentals, requirements, and future directions.
MDPI Algor 10(2):1–24. https://doi.org/10.3390/a10020039
17. Keivan K, Reda A (2008) Effectiveness of support vector machine for crime hotspots prediction.
Appl Artif Intell Int J 22(5)
Analysis of Violent Crime Dataset Using Support … 201

18. Adesola F, Misra S, Omoregbe N, Damasevicius R, Maskeliunas R (2019) An IOT-based

architecture for crime management in Nigeria. In: Shukla R, Agrawal J, Sharma S, Singh TG
(eds) Data, engineering and applications. Springer, Singapore, pp 245–254
19. Nicholas-Omoregbe OS, Azeta AA, Chiazor IA, Omoregbe N (2017) Predicting the adoption
of e-learning management system: a case of selected private universities in Nigeria. Turkish
Online J Dist Educ-TOJDE 18(2):106–121
20. Azeta AA, Misra S, Azeta VI, Osamor VC (2019) Determining suitability of speech-enabled
examination result management system. Wireless Networks, pp 1–8
21. Isah AO, Alhassan JK, Misra S, Idris I, Crawford B, Soto R (2016) Network system design for
combating cybercrime in Nigeria. In: International conference on computational science and
its applications, pp 497–512
Nigeria Human Population Management
Using Genetic Algorithm Double
Optimized Fuzzy Analytics Engine
Approach

Abraham Ayegba Alfa, Sanjay Misra, Blessing Iganya Attah,

Kharimah Bimbola Ahmed, Jonathan Oluranti, Ravin Ahuja,
and Robertas Damasevicius

Abstract Governments and informal sectors require accurate perception concerning

the impending size of various entities such as population, resources, demands,
consumptions, and failures for purpose of planning activities. In order to acquire
this information, the behavior of the associated variables is analyzed on the basis of
past; thereafter utilize the outcomes to make imminent predictions of the targeted
variable. Presently, statistical methods such as conventional and Bayesian has been
deployed for population data analytic about regions and countries. More recently,
heuristics and metaheuristic are being exploited for the purposes of forecasting popu-
lation growth rate and optimization problems. The concept of staggered optimization
of fuzzy analytic engine rules list have been undertaken in the benchmark studies
to reduce redundancy in the rules lists and increase forecasting accuracy. There are
still problems of redundancy and low accuracy, which were explored with evolu-
tionary algorithm (that is, genetic algorithm (GA) and fuzzy analytics) approaches
to manage Nigeria’s human population changes. This proposed approach combines
the outcomes of staggard GA optimizations of fuzzy analytics engines whose rules

A. A. Alfa · K. B. Ahmed
Kogi State College of Education, Ankpa, Nigeria
e-mail: abraham.alfa@kscoeankpa.edu.ng
S. Misra (B)
Department of Computer Science and Communication, Ostfold University College, Halden,
Norway
B. I. Attah
Federal University of Technology, Minna, Nigeria
J. Oluranti
Centre for ICT/ICE Research, Covenant University of Technology, Otta, Nigeria
e-mail: jonathan.oluranti@covenantuniversity.edu.ng
R. Ahuja
Shri Viswakarma Skill University, Gurgaon, Hariyana, India
R. Damasevicius
Kaunas University of Technology, Kaunas, Lithuania
e-mail: robertas.damasevicius@ktu.lt

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 203
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_16
204 A. A. Alfa et al.

lists were filtered to generate the finest fuzzy rules list. The outcomes showed that
12.92% error rate was observed against 17.82, 26.95%, and 42.32% errors observed
in the benchmark works. This model developed offers useful insights to govern-
ment agencies, development partners, and economic planners on the potent ways
to manage population, birth, and death rates for improved resources allocation and
well-being of populace throughout the country.

Keywords Population growth · Error rate · Heuristic · Rules-lists · Genetic

algorithm · Fuzzy analytics · Birth rate · Death rate · Forecasts · SDG 2

1 Introduction

Several nations of the world are facing the problem of diminishing population and
demographic dynamics, which have brought about economic and social interplays.
The projections about population have implications on workface, economic activi-
ties, care-takers, aging and growth rates, and industrialization of nations. The need
to comprehensively understudy population dynamics cannot be over-emphasized
from management to finding answers to numerous problems of both under-and-over-
population growths. Specially, the concept of graying crisis was utilized to explain
rising life expectancy and falling birth rate over a period of time [1].
In the developing world (such as the African continent), the population changes
are influenced greatly by factors such as birth rate, mortality rate, and migration rate.
In general, these factors are capable of propelling or retarding economic growth.
Aside that, the population figures of certain countries impact the global economy
directly through disease spread or migration of people; or indirectly by geopolitics
or environmental factors [2].
Geometric mean, logistic equations, and arithmetic mean are classical tech-
niques for ascertaining future population trends using births, deaths, annexations,
and migration for the purpose of investigating the pattern of growth [3]. Monte Carlo
algorithm, Markov chain, and probability are stochastic-based techniques that have
been deployed for estimating population growth [4]. Population forecasts are less-
desirable for a diverse group of benefactors (including users, planners, policy-makers,
and regulatory agencies) due to its inaccuracies, crude-approach, and uncertainties
[5]. Population changes have diverse variations in the structure of age and spatial
considerations, which impact the demography, environment, and economy [6, 7].
In the case of Nigeria, there is no reliable census data available caused by the
failure to enumerate all citizens, incomplete documentation, and false estimates for
people across age groups [8]. Again, the population figures are outdated or lopsided
population data updates leading to inaccurate assumptions and inappropriate policies.
Therefore, there is the need to deploy more accurate methodologies to ascertain the
dynamics of population in future times in order to facilitate appropriate decision-
making and national prosperity. This paper develops double optimization approach
Nigeria Human Population Management Using … 205

of the Genetic Algorithm on Fuzzy analytics to accurately and effectively forecast

future population changes in Nigeria.
The remaining sections of the paper include: literature review in section two;
research methodology in section three; discussion of results in section four; section
five is the conclusion.

2 Literature Review

2.1 Theories of Human Population Changes

The Malthusian model noted that the growth in population is at a geometric propor-
tion; production of food was rising at an arithmetic proportion. Also, Malthus in [9]
opined that there is the need to put in place measures to regulate the sharp increases
in population otherwise, the population tends to control itself naturally though in a
positive perspective. Accordingly, the earth and its available resources are limited or
scarce, while production of food cannot outlast population growth proportion, then,
factors such as hunger, unrests, and natural occurrences were capable of impacting
the growth in population [2]. Malthus proposed that since population growth is detri-
mental to economic prosperity of a state, serious and effective must be adapted to
minimize unwarranted increases through quality education, moral suasion, marriage
delays, birth controls, and family planning.
In the later part of the 1960s, Neo-Malthusians experts and proponents stretched
the original idea of the Malthus by alluding that institutions within the framework of
the society are incapable of tilting their economic situations to accommodate shocks
triggered by uncontrollable growth in population and high concentration of humans,
which are capable of causing havoc to the environment and social fabric. In contrast,
these institutions make attempts to deal with the obvious changes whenever they
are allowed to fester uncontrollably which is in agreement with two previously held
views: the Distributionists, and the Neoliberals.
Simon in [10], a Neoliberal, pushed the fact that growth in population is never
a serious challenge reason being that the institutions are capable of outlasting it,
through technological innovations. These inventions could offer a new perspective to
overcoming and reshaping the trends caused overpopulation situations. Again, Distri-
butionists think that growth in population meant deliberate shifts from the pressing
issues because institutions of the state have the ability to rollout poverty alleviation
and equity programmes for the purpose of adjusting the resource distribution gaps.
In the late 1980s, the Revisionist opined that to achieve population growth adjust-
ments, countries having high populations are expected to establish and operate effi-
cient contemporary institutions in order to withstand debilitating negative impact
of supposed population explosion. Nevertheless, the school of thought warned that
ineffectiveness of these institutions could inevitably result in catastrophic economic
conditions. Majority of the population policy arguments favored the revisionist
206 A. A. Alfa et al.

ideas that the prospects in population explosion on the nation’s economy were
inconsequential according to Birdsall, Kelley, and Sinding in [11].
During the study of developing economies of 1986, National Academy of Sciences
reported that on-balance and sluggish population growth were considered to be
advantageous to the economic progress at large according to Landau and Rosenberg
[12]. Significantly, this idea contributed to a weakening in the policy and political
priority aimed at controlling growth in population but concentrated efforts towards
revamping economies in these nations.

2.2 Population Forecasting Techniques

Heuristic Based Algorithms. These algorithms are built on the concept of several
natural processes, which offer sub-optimal solutions from obtainable huge piles of
data within a satisfactory frame of time. In particular, Evolutionary Algorithms (EAs)
are a class of metaheuristic algorithms that imitate biological evolution and steered by
the Darwinian principle of the fittest survives. The Genetic Algorithm (GA) is in the
class of EAs introduced by John Henry Holland in 1975 [13]. It makes use of binary
vectors for the purpose of encoding the solution. It is an artificial intelligence system
reproducing the evolutionary principle of survival-of-the-fittest process in order to
produce progressively better solutions to a problem by Khmeleva [14]. Generally,
EAs is an optimizing system because it finds input arrangement that produce the
best outputs. They are most suitable for decision-making situations where thousands
or even millions of solutions are probable. More importantly, it is easy to find and
evaluate possibilities in a faster and more thoughtfully style rather than humans
[15]. However, the algorithm’s parameters relied on outcomes from independent
experiments which may fall short in optimal for a particular problem. In future, the
parameters might be retested in order to verify their suitability for the given problems
[14].
Fuzzy Logic Analytics. This is utilized for assessing, decision-making, and
mechanical control systems such as automobile controls, air conditioning, smart
houses, and industrial process controllers. The notable real-world adoption of fuzzy
logic was in the numerous applications, in Japan as process controllers, even though
the foremost fuzzy control developments were carried out in Europe. Though, the
British engineer Ebrahim Mamdani was the foremost person to practicalize fuzzy
sets in a control system unpremeditatedly. In the early 1970s, Mamdani was working
on an automated control system for stream engine by means of human operator’s
expertise with Bayesian decision theory (that is, techniques of eliciting possibilities
in uncertain situations) that focused on events after the fact to modify prediction
concerning imminent outcomes [16].
Consequent upon the poor performance of the work that artificial intelligence
method of enhancing the learning controller known as rule-based expert system
through the combination of human proficiency with sequence of logical rules for
Nigeria Human Population Management Using … 207

utilizing the knowledge. Mamdani put forward five kinds of systems capable of
profiting from fuzziness concepts. These include [17]:
1. Complex system that is impossible or difficult to model;
2. Human experts-controlled systems;
3. Complex and continuous input and output systems;
4. Systems making use of human observation for inputs or basis for rules creation;
and
5. Naturally vague systems similar to behavioral and social sciences.
The input and output fuzzy variables are specified during fuzzy rule-based
modeling. For each fuzzy variable, its units are defined; the scope of discourse and
a set of membership functions are specified in order to describe the specific fuzzy
concepts linked with the fuzzy variable. The challenge of the model is the selection
of technical indicators required to form the inputs to the decision support system
[18].

2.3 Related Studies

ARIMA model was applied to 145 nations of the world to ascertain the population of
COVID-19 infections across 6 continents using ARIMA parameters such as popu-
lation (per million), cases of infection, and polynomial functions. Other sub-models
were developed to forecast behavior of COVID-19 such as climate, culture, and
humidity [19].
The geographically weighted regression (GWR) mechanism was utilized in deter-
mining the interplay between change in population and various driving factors for
the purpose of understanding the spatial variations in the relationships with small-
area population forecasting in Wisconsin, USA. The outcomes could be improved in
terms of accuracy of climate change and disasters [20].
The role of correct population estimation with stochastic methods in the design
and planning of public sector and critical infrastructure was highlighted in the study
by Patel and Gundalia [5]. It is established that incorrect population forecasts tamper
future services developments such as transportation, water resources, sewage, and
water treatment plant. Modeling of prospective extremes by means of deterministic
technique for measuring population growth in order to account for total equalization
and current differentiation. The main idea is to prepare for inequalities of the future
geographical decision-making process across European countries according to Bleha
[21].
Fuzzy controller has been deployed to control movement of robots in farm envi-
ronment during the application of herbicide by Bala et al. [22]. The outcomes revealed
that the superiority of the model over existing approaches, and could improve the
production of food crops and agribusiness. An SIR model for forecasting COVID-19
infection was built using parameters such as infection rate, death rate, and recovery
rate by Abdy et al. [23]. The member functions were constructed and used to train
208 A. A. Alfa et al.

the model with high accuracy in revealing the possibilities of spread of COVID-19
in Indonesia.
A hybrid correction model composed of Fuzzy logic and ARIMA was proposed
by Xie et al. [24] for forecasting transportation traffic volume. The results indi-
cated the suitability of the model for long-term forecasts of waterway volume of
traffic in more accurate and reliable manner. A logic-related system for assessing the
learning achievements of students, detection of faulty teaching methods timely by
their teachers was developed by Rojas et al. [25]. The results offered the possibility
of early remediation to teaching practice and educational support space for schools
in Colombia.

3 Methodology

3.1 The Proposed Model

This paper proposed the double optimization approach composed of GA-based opti-
mizations of the antecedents [26] and the consequents [27] for fuzzy analytics engine.
These best solutions are further combined to produce the finest rules list for fore-
casting population changes as discussed in the next subsection. The events and
processes engaged by the proposed model are illustrated in Fig. 1.
From Fig. 1, the complete description of the proposed model for population
management is as follows:
Event 1. The raw dataset is collected and formatted properly by removing
redundancy and unnecessary information to fit it for analytics.

Fuzzy logic
rules

Input Optimized fuzzy Output

Raw dataset 1 2 5
variables analytics engine variable

Genetic
algorithm
procedure

Fig. 1 A double GA-optimized fuzzy analytic engine model

Nigeria Human Population Management Using … 209

Event 2. The input variables and corresponding values are used for model training.
Event 3. The original fuzzy rules list generated are passed to genetic algorithm
procedure block for optimization processes.
Event 4. The rules lists are obtained for optimized antecedents, consequents, and
combined outcomes known as double optimized rules list.
Event 5. The test dataset is used to evaluate the performance and effectiveness of
the distinct models built with optimized rules lists using standard metrics in Sect. 3.3.

3.2 Fuzzy Rule List Construction

The inputs variables are Birth rate and Death rate (crude per 1000 people) and the
population rate served as output variable, which was obtained for the period (1960–
2021) from United Nations—World Population Prospects dataset. The structural
relationships between the factors for constructing the fuzzy analytics engine rules-list
are illustrated in Fig. 2.
The degrees of input conditions, combination, and their outcomes using structural
relationships represented in Fig. 2 are shown in Table 1.

Birth_rate
(Input 1)

FIS Population_rate
(Process) (Output)

Death_rate
(Input 2)

Fig. 2 The structure of the fuzzy analytics rules-list

Table 1 Fuzzy decision table

Antecedents Condition indices Maximum-minimum values
What is today’s birth rate per thousand people? High = 3 [36.86–47.24]
Medium = 2
Low = 1
And, what is today’s death rate per thousand High = 3 [11.38–31.15]
people? Medium = 2
Low =
Consequents
Therefore, population rate per billion for today Over = 3 [37.86–211.40]
will be? Stable = 2
Under = 1
210 A. A. Alfa et al.

Table 2 Unoptimized
Rule number Birth_rate Death_rate Population_growth
rules-lists indices
1 3 3 3
2 3 2 2
3 3 1 3
4 2 3 3
5 2 2 2
6 2 1 2
7 1 3 3
8 1 2 2
9 1 1 1

From Table 1, the first or unoptimized rules-list for the fuzzy analytics engine is
presented as shown in Table 2 [26, 27].
From Table 2, the double GA-optimization procedures are performed on the
antecedents, and consequents separately as represented in Table 3.
From Table 3, the values of antecedents and consequents, which are [3332312322
21131211] and [323322321] form the chromosomes for the first stage of the GA-
optimizations whose outcomes are shown accordingly.
The second stage of GA-optimization (known as the double GA-optimization
of rules-list) proposed in this paper used the following optimized fuzzy rules-list
derivable with mean square error (MSE) fitness function:
Antecedents optimized = [233 132 133 313 112 332 233 312 221].
Consequent optimized = [331 322 313 232 222 213 133 122 113].
The antecedents and consequent are combined to construct the required rules-lists
for fuzzy analytics, which is [231 132 133 312 112 333 233 312 223].
After inspection and further refinement of the rules-lists, these rules: 1, 2, 3, 7, and
9 are excluded due to redundancy and inconsistency in order to realize 5 rules-lists

Table 3 GA-optimizations of rules-list indices

S/N Antecedents Consequent (Before) Antecedents (After) Consequent (After)
(Before) [26] [27]
1 33 3 23 1
2 32 2 13 2
3 31 3 13 3
4 23 3 31 2
5 22 2 11 2
6 21 2 33 3
7 13 3 23 3
8 12 2 31 2
9 11 1 22 3
Nigeria Human Population Management Using … 211

as final double-GA-optimization outcomes. The new rules-lists, that is [312 112 333
312 223], is used to build the fuzzy analytics for determining the population growth
based on the birth rate and death rates.
The datasets depict the raw values of variables or factors influencing population
changes, which are used to create the membership functions and subsequently trained
with the double-optimized rules lists of antecedents and consequents in Tables 1 and
3.

3.3 Experimental Setting

The simulation is carried out using MATLAB (R2019b) with system specifications
for hardware and software as follows: Hardware (×64-based processor, AMD E1-
1200 APU with Radeom™, HD Graphics, 1.40 GHz, 4 GB RAM, and 160 HDD).
Software (Windows 8, and 64-bit Operating System).

3.4 Performance Evaluation Metrics

This paper adopted a number of performance evaluation parameters including the

Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute
Percentage Error (MAPE), and Relative Absolute Error (RAE) (Alfa et al. in [26,
27]). The mathematical representations are provided as follows:
The mean square error is given by Eq. 1.

1 
m
MSE = (A − B)2 . (1)
m k=1

The root mean square error is depicted by Eq. 2.



1  m
RMSE =  (A − B)2 . (2)
m k=1

The mean absolute percentage error is represented by Eq. 3.

m  
1  A − B 
MAPE = × 100 (3)
m k=1  A 

The relative absolute error is represented by Eq. 4.

212 A. A. Alfa et al.
 
A − B
RAE =   × 100 (4)
A 

where,
A is real values of training dataset.
B is observed values of training dataset.
m, k are the first and last items of the training dataset.

4 Discussion of Results

The outcomes of the double-optimization procedures of genetic algorithm on the

fuzzy analytics engine are presented in Table 4.
From Table 4, the performance of proposed double GA-optimization of
antecedents and consequents carried out together outperformed benchmark studies
by 12.92% to 42.32%, 17.82% respectively. This implies that there is a minimal error
rate for the proposed model when used to forecast population changes.
Similarly, the proposed approach outperformed the remaining metrics such as
RAE, MAPE, MSE, and RMSE. The distribution of forecasts from initial forecasts,
double GA-fuzzy optimization, final GA-fuzzy, and the target datasets are plotted in
Fig. 3.
From Fig. 3, the proposed approach provided the nearest-best solution to the
problem of population management in the future time due to smallest errors. It is
caused by minimal rules lists in the fuzzy analytics engine offered by the optimization
procedures of GA. Therefore, the birth_rate and death_rate are valuable factors in
determining changes in population of countries and regions of interest. It follows
that a growing population must keep birth rate on the increase while making efforts
to keep citizens safe from untimely deaths such as diseases, wars, and low-living
standards.

Table 4 The outcomes of GA-fuzzy analytics engine optimizations

Metrics Double GA-fuzzy Final double GA-fuzzy [26] GA-fuzzy [27]
(Unrefined) GA-fuzzy
(Proposed)
RAE 0.3451 0.2389 3.768072451 2.365752510
MAPE 0.0173 0.0119 0.091904206 0.057701281
MSE 0.00066309 0.00031780 0.001041135 0.000438377
RMSE 257.50 178.27 0.032266628 0.020937461
No. of fuzzy rules 8 5 7 5
GA operation Both Both Consequent Antecedents
Error rate (%) 26.95 12.92 42.32 17.82
Nigeria Human Population Management Using … 213

Fig. 3 The performance of GA-fuzzy analytics compared

5 Conclusion

The population size management impacts numerous factors such as possible supply
of inexpensive unskilled and skilled labor, capability to consume and raise aggregate
demand, market demand, and the Gross Domestic Product (GDP) growth, geopo-
litical status, tax and revenue, and infrastructural provisions. There is a need to
control the rate of population growth in order to match the pace of economic growth.
When this is achieved, many crisis situations can be avoided including balance of
payment, and high youth unemployment caused by surplus net imports, inflation,
corruption, food insecurity, pollution, deforestation, social vices, and environmental
degradations (such as dwindling natural resources and global warming).
This study adopted GA optimized fuzzy rules lists for both antecedents and conse-
quents, which serve as basis for generating new double optimized fuzzy rules-list.
The outcomes showed superior performance of the proposed model against compa-
rable approaches [26, 27] in case of Nigeria’s population forecasts. However, fuzzy
logic analytics-based models cannot be easily modeled, aside high error rates. Several
simulations and finetunings are needed to make it more suitable for or crisp system,
or mathematical precision, and linear modeling. There is the need to consider other
optimization techniques for the purpose of enhancing the effectiveness of fuzzy
analytics especially in the construction of its rules-lists for forecasting tasks.

Acknowledgements The authors appreciate the Covenant University through its Centre for
Research, Innovation and Discovery for Financial assistance.
214 A. A. Alfa et al.

References

1. Sigurðsson EÖ (2017) Impacts of population aging in modern Japan and possible solutions for
the future. Universitatis Islandiae, Sigillum
2. Anudjo EY (2015) The population growth-economic growth nexus: new evidence from Ghana.
University of Ghana, Legon
3. Kumar V (2020) Application of mathematical modeling in population forecasting. Uttaranchal
Univ J Appl Life Sci 3(1):50–57
4. Zhu D, Jiang K (2018) Population forecasting model based on monte carlo algorithm. ACM
ICCDE 2018:67–72
5. Patel N, Gundalia M (2020) Comparative study of population forecasting methods for Surat
City (Gujarat-India). J Interdisc Cycle Res 12(10):633–636
6. Uddin GA (2016) Population changes and implications for economic growth and the envi-
ronment in Australia. A Ph.D. Thesis, Department of Commerce, University of Southern
Queensland, Australia, pp 1–201
7. Pettersen AM (2017) Youth employment in Uganda: challenges and survival strategies for
young women in Kampala. A Master’s Thesis, Centre for Peace Studies, The Arctic University
of Norway, pp 1–66
8. Kusumasindra F (2017) Comparing sources of population data in health information systems:
a case study of Zambian health sector. Master’s Thesis, Department of Informatics, University
of Oslo, Norway, pp 1–113
9. Malthus TR (1798) An essay on the principle of population (Printed for J. Johnson)
10. Simon JL (1977) The economics of population growth
11. Birdsall N, Kelley AC, Sinding S (2001) Population matters: demographic change, economic
growth, and poverty in the developing world: demographic change, Economic Growth, and
Poverty in the Developing World, Oxford University Press
12. Landau R, Rosenberg N (Eds) (1986) The positive sum strategy: harnessing technology for
economic growth. National Academies Press
13. Holland JH (1975) Adaptation in natural and artificial systems. MIT Press
14. Khmeleva E (2016) Evolutionary algorithms for scheduling operations, A Ph.D. Thesis,
Sheffield Hallam University, pp 1–337
15. Baltzan P (2015) Business driven information systems, 5th edn. McGraw-Hill, New York
16. Kayacan E (2011) Interval type-2 fuzzy logic systems: theory and design. A Ph.D. Thesis,
Department of Electrical and Electronic Engineering, Bogazici University, Instanbul, Turkey,
pp 1–149
17. Kotta A (2015) Condition monitoring using computational intelligence methods. unpublished
M.Sc. Thesis, Department of Computer Engineering, University of Technology Chemnitz,
Germany, pp 1–92
18. Delnavaz B (2014) Forecasting of the stock price index by using fuzzy-neural network and
genetic algorithms. J Appl Sci Agric Am-Eurasian Netw Sci Inf J 9(9):109–117
19. Hernandez-matamoros A, Fujita H, Hayashi T, Perez-Meana H (2020) Forecasting of COVID19
per regions using ARIMA models and polynomial functions. Appl Soft Comput J 96:106610.
https://doi.org/10.1016/j.asoc.2020.106610
20. Chi G, Wang D (2017) Small-area population forecasting : a geographically weighted regression
approach. In: Swanson DA (ed) The frontiers of applied demography, applied demography serie,
pp 449–471
21. Bleha B (2020) Future population developments in Europe. Is the concept of convergence
indisputable? Towards the role of geographical thinking in population forecasting. Appl Spat
Anal Pol. https://doi.org/10.1007/s12061-019-09330-6
22. Bala JA, Folorunso TA, Olaniyi OM, Daniya E (2021) Design of an agribusiness innovative
and autonomous robot system for chemical weed control for staple food crops production in
sub-saharan Africa. In: Che FN, Strang KD, Vajjhala NR (eds) Opportunities and strategic use
of agribusiness information systems, IGI Global, pp 234–262
Nigeria Human Population Management Using … 215

23. Abdy M, Side S, Annas S, Nur W, Sanusi W (23) An SIR epidemic model for COVID-19
Spread with fuzzy parameter: the case of Indonesia. Adv Diff Equ 105, https://doi.org/10.
1186/s13662-021-03263-6. BioMed Central Ltd Part of Springer Nature
24. Xie Y, Zhang P, Chen Y (2021) A Fuzzy ARIMA correction model for transport volume
forecast. Math Prob Eng Article ID 6655102. Hindawi. https://doi.org/10.1155/2021/6655102
25. Rojas JA, Espitia HE, Bejarano LA (2021) Design and optimization of a fuzzy logic system
for academic performance prediction. Symmetry 13(133):1–20. https://doi.org/10.3390/sym
13010133.MDPI
26. Alfa AA, Yusuf IO, Misra S, Ahuja R (2020) Enhancing stock prices forecasting system
outputs through genetic algorithms refinement of rules-lists. In: 1st IC4S 2019, Lecture Notes
in Networks and Systems, vol 12. Springer Nature Singapore Pte Ltd, pp 669–680
27. Alfa AA, Misra S, Bumojo A, Ahmed KB, Oluranti J, Ahuja R (2020) Comparative analysis of
optimisations of antecedents and consequents of fuzzy inference system rules lists using genetic
algorithm operations. In: Chillarige RR (ed) ICACII 2019, Lecture Notes in Networks and
Systems, Advances in Computational Intelligence and Informatics, vol 119. Springer Nature
Singapore Pte Ltd, pp 373–379
Flower Species Detection System Using
Deep Convolutional Neural Networks

Arun Solanki and Tarana Singh

Abstract A system that correctly identifies the name of a flower species may be
beneficial for botanists, camping enthusiasts, and researchers. Previously, classifica-
tion was only done based on a flower’s shape, geometry, and texture, which is not
enough for an efficient system. Some significant challenges in this classification task
include inter-species similarity, intra-class variation, and the same objects such as
leaves or grass around a flower, making this task a research topic. This research has
developed an efficient and robust deep learning flower classifier to overcome these
problems and limitations based on the current state of the art convolutional neural
networks and transfer learning. This research has utilized the Oxford-102 flower
dataset having 8189 images of 102 flower species. The proposed method is divided
into two different steps. Firstly, the flower images are segmented, and secondly,
these segmented images are fed as an input to a convolutional neural network for
classifying the species of the flowers. This work has used the PyTorch library for
recognition purposes. The flower’s dataset uses various pre-trained models on the
ImageNet dataset such as AlexNet, VGG, DenseNet, Inception v3, and GoogLeNet.
Out of these, DenseNet achieved the highest classification accuracy of 97.92% when
trained on GPU provided by Google Collaboratory. This classifier can be integrated
with a mobile application to provide an accurate real-time flower species prediction.

Keywords Convolution neural network (CNN) · Segmentation · Cropping ·

Augmentation · Transfer learning · ReLU activation function

1 Introduction

Plant species recognition based on flower recognition remains a challenge in the

field of image processing and computer vision, primarily due to their widespread
presence, complex structures, and unpredictable species in nature. Due to this natural
complexity, it is highly undesirable to segment or extract regular features or combine

A. Solanki (B) · T. Singh

Department of Computer Science and Engineering, Gautam Buddha University, Greater Noida,
U.P., India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 217
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_17
218 A. Solanki and T. Singh

shape, texture, and color features, resulting in moderate accuracy in benchmark

datasets increase. Several feature extraction techniques that combine global and
local feature descriptors have achieved the highest accuracy in flower classification,
but automatically identify and recognize large flower species in complex environ-
ments [1]. Still needs a powerful and efficient system. This paper also uses mecha-
nism of transfer learning to save our time and resources. For this, we have utilized
Oxford-102 dataset of images having 8189 flower images belonging to 102 kinds
of different flower species. The proposed method is divided into two major steps
[2]. Firstly, the flower images are segmented, and secondly, these segmented images
which act as input afterward go into a convolutional neural network for classifying
the species belonging to different flower categories. We have also pre-processed
the flower images which we will discuss in the later section of this paper. To facil-
itate our proposed work, we have employed PyTorch library for the recognition
purposes and various pre-trained models on ImageNet dataset such as AlexNet,
VGG, DenseNet, Inception v3, and GoogLeNet were used on the flower’s dataset.
Out of these, DenseNet achieved the highest classification accuracy when trained on
GPU provided by Google Collaboratory [3]. This classifier can be built into a mobile
application so that it can provide an accurate flower species prediction in real-time
environment. For the implementation of the flower species recognition system, we
have used python 3.6.6 version. We have employed PyTorch library for the devel-
opment of the code based on transfer learning. The whole training of the model is
done on Google Collaboratory which is free GPU provided by Google. We have
obtained the dataset which is used in our proposed methodology is from Science
and Engineering department of University of Oxford. This dataset is known as 102
flower category datasets [4]. This flower’s dataset is having 102 types of different
flower species. Each category of flower contains the images ranging from 40 to 258
images.
CNN comes under the picture of an artificial neural network that has wide recog-
nition in image classification. Several layers in a CNN includes convolutional layers,
pooling layers, and fully connected layers [5]. Operation in the series format is applied
to the data, which acts as input to the CNN network to find a particular pattern in
the image. This network does the processing of the data with a grid-like topology
[6]. A CNN model uses the image pixel in the form of an array as input. The input
data is processed through a hidden layer, and the final output is shown or given by
the output layer. The primary function of the hidden layer is feature extraction that
deals with the calculation and manipulation of the data [7]. This work is based on the
convolution layer, which filters the matrix and performs the convolution operations
to help pattern finding in the image. Hidden layers can be many depending upon
the architecture of the network like we can have a ReLU layer or a pooling layer or
convolution layer [8]. At the end of each process, we get feature maps convolved,
rectified, and pooled. The modified pixel values have to get passed through the fully
connected layer where the real object detection occurs [9]. Figure 1 illustrates the
overall architecture of a convolutional neural network. This also has the depiction of
all the layers contained in the network [10].
Flower Species Detection System Using Deep … 219

Fig. 1 Architecture of convolutional neural network [11]

When reviewing previous studies, several flower identification methods have been
suggested [1, 6, 12]. These methods usually consist of four steps: pre-processing,
segmentation, manual design feature extraction, and classification [13]. Due to the
complex background of flower images, this task can be very time consuming, and
for many types, in particular, the accuracy obtained is still low. Recently, learning
feature representations using convolutional neural networks (CNNs) has been very
successful in various areas of computer vision, including object detection, segmenta-
tion, and visual image classification [9]. Feature learning methods provide a natural
way to capture clues using many codewords (sparse coding) or neurons (deep
networks) [14, 15]. All of these are useful clues because you can capture the natural
features of the object. Therefore, this article examines and presents the efficiency
of deep convolutional neural networks, which may more effectively identify plant
species based on flowers [16, 17].
This whole paper is organized into eight sections. In section one, introduction
and the motivation of the work are presented. In the second section of this paper, a
literature survey of the related domain is presented. In section three, the proposed
architecture of the system is discussed. The process flow chart of the proposed system
is discussed in section four. Then, in section five of this chapter, the pseudo-code
of the proposed system is presented, followed by the step-by-step discussion of the
implementation of the proposed algorithm in section six. Section seven discusses the
results of the proposed system, followed by the comparison of the proposed work
with the existing systems. At the end of the paper, a conclusion is given, followed
by future work.

2 Literature Survey

Krizhevsky [18] brings out the phenomenal results on the ILSVRC2012 through
developing a deep convolutional neural network. The top-1 error rate was 37.5% and
the top-5 error rate was 17%. This method was certainly better than other methods
in the past for the same domain. A system based on a convolutional neural network
was a build-up of deep layers ensemble with the structural network containing eight
layers. To avoid overfitting problem [19], there is the incorporation of the essential
220 A. Solanki and T. Singh

features. These features are pooling layers along with normalizing layers with the
functionality of dropout. According to Sermanet [20], using CNN for object loca-
tion and object detection in images will boost classification accuracy. It will also
increase the accuracy of detection and location tasks. This method is the winner
of the localization task on the challenge of ILSVRC2013 through the developed
integrated approach used for detection, localization, and recognition [21, 22]. This
algorithm gave brilliant results through classification accuracy. Szegedy [10] devel-
oped and designed the architecture of a deep convolutional neural network which
is called inception and there is seen great classification and detection results for the
challenge ILSVRC2014 [23]. The author in [20] states that for the benefit of classifi-
cation depth representation is essential. With the substantial increase in the intensity,
good results can be achieved on the ImageNet dataset using a conventional CNN.
We can use a convolutional neural network for the segmentation of the images
and can be employed to detect the objects in the images. Segmentation through CNN
has been achieved through the paper’s fully convolutional networks (FCN) concept
[20]. Several methods extend the concept of CNN to allow object detection tasks
with good accuracy on benchmark datasets. These methods are R-CNN [24] which
is region proposals with CNN. Another advancement is fast R-CNN explained in
[25]. Later on, there is the development of the architecture of Faster R-CNN [26] and
YOLO [27]. The results are similar if we compare FCN with these methods when
using CNN’s architectures, including AlexNet [28] and VGG-16 [10].

3 The Architecture of the Proposed System

Figure 2 shows the framework, which is the designed architecture for our proposed
method employed to deal with flower species recognition.

Fig. 2 Architecture of the proposed work

Flower Species Detection System Using Deep … 221

The efficient and robust system we have developed to classify different types
of flower species is depicted in Fig. 2. This figure shows the overall framework,
which is also the architecture of the proposed method. Architecture is composed of
modules, blocks, and sub-modules [29]. Architecture describes the flow of the code
right from data collection to the prediction of an unknown set of flower examples by
the trained classifier. Here, we have utilized the architecture type of convolutional
neural network, DenseNet, a pre-trained model on the ImageNet dataset [30]. This
is called a transfer learning mechanism. There are two significant modules in the
architecture are:
• Training Module: The training module of the architecture proposed contains
three blocks and five modules which are described below. The input to this
module is the raw flower images one by one. This module has three blocks:
Image Processing Block, which is mainly responsible for the data preparation for
the training and contains three modules: segmentation module, cropping module,
and data augmentation module. The second is the Transfer Learning Block, which
focuses on the transfer learning mechanism and comprises two major modules:
loading pre-trained model module and parameter tuning module. And the third
one is the ImageNet Weight Block, which has the weights of the ImageNet
dataset used in our flower classification problem.
• Testing Module: The testing module of the architecture proposed contains two
blocks and three modules which are described below. The output from this module
is the class label, the predicted species of an unknown flower image, which is
the input to this testing module. This module has three further modules: Predic-
tions with the Training Module, Segmentation and Cropping Module, and Image
Augmentation Module.

4 Process Flow Chart of the Proposed System

The process flow carried out in the proposed work is systematically explained by the
below flowchart, which contains all the steps of execution to accomplish the research
(Fig. 3).

4.1 Oxford 102 flower’s Dataset

We have the oxford 102 flower’s dataset at our disposal, which has to go into our clas-
sification model for flower species prediction [31]. Figures 4 and 5 are the depiction
of variability between flower species and variability within flower types.
222 A. Solanki and T. Singh

Fig. 3 Process flow of the proposed method

Fig. 4 Variability between classes

Flower Species Detection System Using Deep … 223

Fig. 5 Variability within classes

4.2 Flower Images Segmentation Using BiCoS

To improve classification accuracy, there is a need to separate foreground (the flower

structure) from the complex background due to leaves and grass, which is useless
for the classification task. So, we do the flower images segmentation to obtain the
desired segmented images of the flowers.

4.3 Cropped and Segmented Flower Images

After segmentation, the background becomes black. We need to remove that, so we

crop the segmented images using a python script to obtain the cropped and segmented
images of flowers.

4.4 Pre-trained Models on ImageNet Dataset

There are many pre-trained models like DenseNet, VGG-16, or AlexNet that can be
loaded from the Torchvision module of PyTorch.

4.5 Transfer Learning

This mechanism of transfer learning is gaining huge popularity in deep learning. We

load the pre-trained models on the ImageNet dataset into our code. This mechanism
is called transfer learning which is reusing things that have very high standards. Pre-
trained networks generally contain two things. One is feature detectors, and the other
is a classifier. Feature detectors extract from each image the information [32, 33]. Our
classifier will learn the input given by feature layers, and therefore we will freeze the
feature layers to avoid any modification. If we talk about the most commonly used
224 A. Solanki and T. Singh

pre-trained models on the ImageNet dataset, these are AlexNet, DenseNet, VGG-16
and many more that have gained popularity over recent years.

5 Pseudo-Code of the Proposed System

Step 1: Loading Oxford 102 Dataset.

Step 2: Essential Libraries Loading.
Step 3: Data Segmentation and Cropping of Flower Images.
Step 4: Data Augmentation.
Step 5: Loading pre-trained model.
Step 6: Classifier Building.
Step 7: Model Training.
Step 8: Model Testing.
Step 9: Save Model Checkpoint.
Step 10: Load Model Checkpoint.
Step 11: Processing Images.
Step 12: Class Prediction.
Step 13: Sanity Check.

6 Implementation of the Algorithm

Step 1: Loading Oxford 102 Dataset—In this step, we load the Oxford 102 dataset of
flowers images into our code to apply our model for the prediction of flower species.
This dataset contains 102 species of flower images and is divided into training and
test sets.
Step 2: Essential Libraries Loading—This process involves the loading of essen-
tial libraries and packages to make use of the functions in the modules of these
packages.
Step 3: Data Segmentation and Cropping—We have to remove the complex back-
ground, which contains leaves and grass and these things create significant confusion
for the flower classification task. Therefore, we have segmented the flower images
using a technique called BiCoS segmentation for image classification. Then, this
segmented image is cropped with the help of a python script to improve the accuracy
of the network. Figure 6 shows the conversion of the original image of the flower
picked up from the Oxford 102 species to the segmented image and the conversion
from segmented image to cropped image.
Step 4: Data Augmentation—As our dataset is not very large, we need to augment
the dataset of flower images. This is because we want our program to learn as much as
it can. So, we must apply some random transformations to build a robust and efficient
flower classifier. For this, we have to train our model on various variations of the
Flower Species Detection System Using Deep … 225

a) Original Image b) Segmented Image c) Cropped Image

Fig. 6 Cropped image obtained from the segmented image

original image. These variations include rotation, translation, scaling, cropping, or

flipping of the original image.
The transformations are done on each image and these images are passed through
a neural network in each epoch. This will allow the network to train on a more signifi-
cant number of images. Therefore, we have increased the variety of our training data
which will further reduce overfitting. This also improves the capability of gener-
alization by the classifier, so there is a sure shot increase in the model’s accuracy.
To normalize the image values before entering them into the network, we have to
provide the network with mean and standard deviation. If we look at the dimensions
of image tensors, we can have the values of mean and standard deviation. PyTorch
library allows us to do the data transformation through its torchvision package.
Inside torchvision package, we have the module named transform, which has several
functions helpful in transforming the images.
Step 5: Loading pre-trained model—We load the pre-trained models on the
ImageNet dataset into our code. This mechanism is called transfer learning which is
reusing things that have very high standards. This mechanism of transfer learning is
gaining huge popularity in the field of deep learning. Pre-trained networks generally
contain two things. One is a feature detector and the other is a classifier. Feature
detectors extract from each image the information. Our classifier will learn the input
given by feature layers and therefore, we will freeze the feature layers to avoid any
modification. If we talk about the most commonly used pre-trained models on the
ImageNet dataset, these are AlexNet, DenseNet, VGG-16, and many more that have
gained popularity over recent years. All we save through the power of these trained
feature models on huge datasets is our precious time and the computer’s resources.
These models provide cutting edge results on the smaller datasets when reused.
Step 6: Classifier Building—We will build our classifier and this classifier has to
be replaced by the model pre-trained on ImageNet. We will freeze the feature layers
of the network. This is because we will provide the input according to our dataset.
The feature layer will learn these inputs and we don’t want these layers to learn the
same inputs as of ImageNet dataset. We will set the output size of the classifier as
102 as we have these many species in our dataset. We will achieve this task with the
help of defining a precise function.
Step 7: Model Training—In this step, it is time to train our classifier on our
flower’s 102 category dataset. We will train the final layers of the network. We only
226 A. Solanki and T. Singh

train the classifier parameter while the feature parameters are kept frozen. We can
change our optimizer as well as a scheduler in the piece of our code.
Step 8: Model Testing—In this step, our trained model is evaluated to measure
the performance of the test images of our dataset. At the end of this step, we obtain
the percentage accuracy, which means how many flower test images are correctly
classified.
Step 9: Save Model Checkpoint—We will save our model in the directory created.
This is done to ensure the backup of our created and trained model. This will come in
handy when we have to use this trained model on some unknown images of flowers.
Step 10: Load Model Checkpoint—We load our trained model to use this on the
unknown flower images to predict their species name.
Step 11: Processing Images—We will carry out the processing of the images
because we will take this image as the unknown image for which we need to predict
the class label. So, there is a need of related data transformations.
Step 12: Class Prediction—We will predict the class of flower species of the given
image unknown to the model. This whole process is carried out in the probability
that a particular flower type belongs to that class.
Step 13: Sanity Check—All the earlier parts of this code are combined in a
function. This function performs the plotting, or we can say just graphing where the
models predict with uncertainty.

7 Discussion of Result of the Proposed System

and Comparison of the Results with the Existing Work

7.1 Number of Epochs Versus Classification Accuracy

In Graph 1a, we plotted 60 values of classification accuracies measured at epochs

ranging from 1 to 60. This graph depicts the dependency of several epochs on the
classification accuracy of our deep learning trained classifier. As the number of
epochs increases, the accuracy also increases and reaches the maximum of 97.92 at
epoch 25. From then onwards, accuracy declined and then increased again. Then, the
graph follows a pattern of increase and decreases until epochs 60. But we observed
that the accuracy could not cross the value of the maximum, which is 97.92% at any
epochs after 25. So, we conclude that increasing the epochs after 25 do not affect
or increase the classification value of the model. Graph 4 shows the results of the
Proposed System.
In Graph 1, we have plotted the number of epochs on the x-axis and training
accuracy on the y-axis. We plotted 100 values of training accuracies measured at
the epochs, ranging from 1 to 100 in the training phase. This graph illustrates that
training accuracy shows a sudden increase till epoch 15. Still, from there onwards,
the accuracy doesn’t show a desirable increase and there is also not drastic decrease
in the training accuracy. This means that after epochs 15 the accuracy curve seems to
Flower Species Detection System Using Deep … 227

Graph 1 a and b Number of epochs versus classification accuracy and training accuracy

be constant as it keeps on increasing and decreasing only by a very small or negligible

value.

7.2 Number of Epoch Versus Training Loss

In Graph 2, we have plotted the number of epochs on the x-axis and training loss
on the y-axis. We have plotted 100 values of training loss measured at the epochs,
which range from 1 to 100 in the training phase. This graph illustrates that training
loss shows a sudden decrease till epochs 10, but the loss doesn’t show a desirable
decrease and there is also no drastic increase in the training loss. This means that
after epochs 10 the loss curve seems to be constant as it keeps on decreasing and
increasing but only by a very or negligible small value.

Graph 2 Number of epochs

versus training loss
228 A. Solanki and T. Singh

7.3 Validation Curve

Number of Epochs versus validation accuracy.

Graph 3a shows the dependency of the number of epochs on the validation accu-
racy. We have plotted 100 values of epochs on the x-axis and 100 values of validation
accuracy on the y-axis. As the number of epochs increases, the accuracy of the vali-
dation points also increases. Still, this accuracy value increases only until the epoch
value of 8 at a fast rate. Then, from epoch 8 the validation accuracy fluctuates. That
is it keeps on increasing and decreasing till epochs 100. Overall, we can say that after
epoch 8 the validation accuracy doesn’t show a significant increase or also does not
display a drastic decrease.
Graph 3b shows the dependency of the number of epochs on the validation loss.
We have plotted 100 values of epochs on the x-axis and 100 values of validation loss
on the y-axis. As the number of epochs increases, the loss of the validation points
decreases, but this value of loss decreases only till the epoch value of 8 at a very fast
rate. Then from epoch 8 the validation loss fluctuates. That is, it keeps on decreasing
and increasing till epochs 100. Overall, we can say that after epoch 8 the validation
loss doesn’t show a significant decrease or also does not display a drastic increase.

Graph 3 a and b Number of epochs versus validation accuracy and validation loss

Graph 4 Existing work

versus proposed work
Flower Species Detection System Using Deep … 229

7.4 Comparison of the Results of the Proposed System

with the Exiting System

Graph 4 compares the existing work [34] in flower classification with the proposed
method, which is robust and efficient. Both of the work is performed on the Oxford-
102 dataset. The existing work achieved a classification accuracy of 84.02% with
the use of five convolutional layers. Our developed deep learning flower classifier
system has set the really high standards in this domain by achieving a very high
recognition rate of 97.92%.

8 Conclusion and Future Work

This work has developed an efficient and robust flower species recognition classifier
based on deep learning. We have used the dataset from the University of Oxford,
which is the Oxford-102 flower dataset with a total of 8189 images of different
categories of flower species [35, 36]. We have divided our dataset into training
sets and validation set for evaluation purposes. We have employed PyTorch library
by Facebook to code our research work. DenseNet161, a pre-trained model of the
ImageNet dataset, was loaded to use its weights and later applied to our flower’s
dataset [11, 37]. This all result was achieved through transfer learning mechanism,
which is gaining popularity in deep learning. We have developed a four-step novel
approach for the classification of the 102 categories of flower species which is below;
1. The data Augmentation for better training of flower classifier.
2. The Flower Image Segmentation using the BiCoS method for removing the
complex background.
3. The cropping of segmented flower images using python script.
4. The model training using the pre-trained model—DenseNet.
5. For training purposes, we have used Jupyter Colab Notebook, a free graphics
processing unit (GPU) provided by Google Collaboratory. Our proposed method
achieved very high accuracy on the flower’s dataset, which is 97.92% classifi-
cation accuracy. This is one of the best results obtained in the domain of flower
species classification.
A deep learning-based CNN classifier is being developed in this work, one of
the most robust and efficient with 97.92% classification accuracy on the benchmark
dataset. But still, there exists some more work in this domain that can be done
in future to use the system in the real world with high accuracy. Some future work
points are an extension in the dataset having more categories, integration with mobile
applications, and increase in the training data.
230 A. Solanki and T. Singh

References

1. Singh SP, Solanki A, Singh T, Tayal A (2021) Internet of intelligent things: injection of intel-
ligence into IoT devices. In: Artificial intelligence to solve pervasive internet of things issues.
Academic Press, pp 85–102
2. Issa MB, Daraghmeh M, Jararweh Y, Al-Ayyoub M, Alsmirat M, Benkhelifa E (2017) Using
logistic regression to improve virtual machines management in cloud computing systems. In:
2017 IEEE 14th international conference on Mobile Ad Hoc and Sensor Systems (MASS),
22–25 Oct. 2017
3. Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. In: Fleet
DJ, Pajdla T, Schiele B, Tuytelaars T (eds) ECCV, volume 8689 of Lecture Notes in Computer
Science. Spring, pp 818–833
4. Redmon J, Divvala S, Girshick R et al (2016) You only look once: unified, real-time object
detection. In: Proceeding IEEE conference computer vision and pattern recognition, Las Vegas,
NV, June 2016, pp 779–788
5. Singh T, Nayyar A, Solanki A (2020) Multilingual opinion mining movie recommenda-
tion system using RNN. In: Proceedings of first international conference on computing,
communications, and cyber-security (IC4S 2019). Springer, Singapore, pp 589–605
6. Luong DTA, Chandola V (2017) A K-means approach to clustering disease progressions. In:
IEEE Conference on 14 September 2017
7. Solanki A, Singh T (2021) COVID-19 epidemic analysis and prediction using machine learning
algorithms. Emerging technologies for battling Covid-19: applications and innovations, pp
57–78
8. Rong F (2017) Audio classification method based on machine learning. In: IEEE Conference
on 21 September 2017
9. Singh T, Mishra J (2021) Learning with artificial intelligence systems: application, challenges,
and opportunities. In: Impact of AI technologies on teaching, learning, and research in higher
education. IGI Global, pp 236–253
10. Simonyan K, Zisserman A (2015) Very deep convolutional networks for large-scale image
recognition. In: Proceeding of international conference learning representations, San Diego,
CA, May 2015, arXiv preprint arXiv:1409.1556
11. Hu W, Hu R, Xie N et al (2014) Image classification using multi-scale information fusion based
on saliency driven nonlinear diffusion filtering. IEEE Trans Image Process 23(4):1513–1526
12. Nitta T (1993) A back-propagation algorithm for complex numbered neuralnetworks. In:
Proceedings of 1993 International Joint Conference on Neural networks, IJCNN’ 93-Nagoya.
25–29 Oct. 1993
13. Nilsback M, Zisserman A (2008) Automated flower classification over a large number of
classes. In: Proceeding sixth indian conference computer vision, graphics and image processing,
Bhubaneswar, India, December 2008, pp 722–729
14. Nilsback M, Zisserman A (2006) A visual vocabulary for flower classification. In: Proceeding
IEEE conference computer vision and pattern recognition, New York, NY, June 2006, vol 2,
pp 1447–1454
15. Pandey S, Solanki A (2019) Music instrument recognition using deep convolutional neural
networks. Int J Inf Technolhttps://doi.org/10.1007/s41870-019-00285-y (SpringerPublication)
16. Singh T, Solanki A, Sharma SK (2021) Role of smart buildings in smart city—components,
technology, indicators, challenges, future research opportunities. Digital Cities Roadmap: IoT-
Based Architecture and Sustainable Buildings, pp 449–476
17. Bhardwaj N, Solanki A (2016) An efficient algorithm for color image segmentation.
Selforganizology 3(3):87–99
18. Krizhevsky A, Sutskever I, Hinton G (2012) ImageNet classification with deep convolutional
neural networks. In: Pereira F, Burges C, Bottou L et al (eds) Advances in neural information
processing systems (Curran Associates, Inc., Red Hook, NY, USA, 2012), pp 1097–1105
19. Hinton GE, Srivastava N, Krizhevsky A, Sutskever I, Salakhutdinov R (2012) Improving neural
networks by preventing co-adaptation of feature detectors, CoRR, abs/1207.0580
Flower Species Detection System Using Deep … 231

20. Szegedy C, Liu W, Jia Y et al (2014) Going deeper with convolutions. arXiv preprint arXiv:
1409.4842
21. Yang M, Zhang L, Feng X et al (2014) Sparse representation based Fisher discrimination
dictionary learning for image classification. Int J Comput Vis 109(3):209–232
22. Priyadarshni V, Nayyar A, Solanki A, Anuragi A (2019) Human age classification system using
K-NN classifier. In: Luhach A, Jat D, Hawari K, Goa XZ, Lingras P (eds) Advanced informatics
for computing research. ICAICR 2019. Communications in computer and information science,
vol 1075. Springer, Singapore
23. Khan F, van de Weijer J, Vanrell M (2012) Modulating shape features by color attention for
object recognition. Int J Comput Vis 98(1):49–64
24. Shelhamer E, Long J, Darrell T (2017) Fully convolutional networks for semantic segmentation.
IEEE Trans Pattern Anal Mach Intell 39(4):640–651
25. Girshick R, Donahue J, Darrell T et al (2014) Rich feature hierarchies for accurate object
detection and semantic segmentation. In: Proceeding IEEE conference computer vision and
pattern recognition, Columbus, OH, June 2014, pp 580–587
26. Ren S, He K, Girshick R et al (2017) Faster R-CNN: towards real-time object detection with
region proposal networks. IEEE Trans Pattern Anal Mach Intell 39(6):1137–1149
27. Girshick R (2015) Fast R-CNN. In: Proceeding of IEEE international conference computer
vision, Santiago, Chile, December 2015, pp 1440–1448
28. Zhou H, Zheng J, Wei L (2013) Texture aware image segmentation using graph cuts and
active contours. Pattern Recog-nition 46(6):1719–1733. https://doi.org/10.1016/j.patcog.2012.
12.005
29. Xie L, Wang J, Lin W et al (2017) Towards reversal-invariant image representation. Int J
Comput Vis 123(2):226–250
30. Hsu T, Lee C, Chen L (2011) An interactive flower image recognition system. Multimedia
Tools Appl. 53(1):53–73
31. Mottos A, Feris R (2014) Fusing well-crafted feature descriptors for efficient fine-grained
classification. In: Proceeding IEEE international conference image processing, Paris, France,
October 2014, pp 5197–5201
32. Chai Y, Rahtu E, Lempitsky V et al (2012) TriCoS: a tri-level class discriminative co-
segmentation method for image classification. In: Proceeding of European conference computer
vision, Florence, Italy, October 2012, vol I, pp 794– 807
33. Chen Q, Song Z, Hua Y et al (2012) Hierarchical matching with side information for image
classification. In: Proceeding IEEE Conference. Computer Vision and Pattern Recognition,
Providence, RI, June 2012, pp 3426–3433
34. Liu Y, Tang F, Zhou D et al (2016) Flower classification via convolutional neural network.
In: Proceeding of IEEE international conference functional-structural plant growth modeling,
Simulation, Visualization and Applications, Qingdao, China, November 2016, pp 110–116
35. Chai Y, Lempitsky V, Zisserman A (2011) BiCoS: a Bi-level co-segmentation method for image
classification. In: Proceeding of international conference computer vision, Barcelona, Spain,
November 2011, pp 2579–2586
36. Qi X, Xiao R, Li C et al (2014) Pairwise rotation invariant co-occurrence local binary pattern.
IEEE Trans Pattern Anal Mach Intell 36(11):2199–2213
37. Ciresan DC, Meier U, Schmidhuber J (2012) Multi-column deep neural networks for image
classification. In: CVPR, pp 3642–3649
Data Localization
and Privacy-Preserving Healthcare
for Big Data Applications: Architecture
and Future Directions

Ashwin Verma, Pronaya Bhattacharya, Yogesh Patel, Komal Shah,

Sudeep Tanwar, and Baseem Khan

Abstract With the exponential rise of generated big data in Healthcare 4.0 ecosys-
tems, the data is shared among multiple stakeholders for analysis and data collec-
tion process. During the data collection process, a public data set is released from
privately collected data through anonymizing critical and sensitive attribute fields.
However, existing privacy preservation techniques are often susceptible to linkage
and statistical-based attacks that releases sensitive attributes of the patient. Thus,
researchers worldwide have shifted attention towards privacy preservation of health-
care data. Owing to the risk of release of critical attributes, well-defined privacy-
preservation techniques and ethical data collection process are much-sought issues.
Thus, in this paper, we present a systematic survey of data localization issues coupled
with privacy preservation mechanisms, specifically in the context of the healthcare
industry. In privacy preservation, we present key techniques like K -anonymity, I -
diversity, t-closeness, aggregation, removal of linkage-based records, homomorphic
encryption, and noise-based schemes. The survey intends to help existing healthcare
practitioners, stakeholders, and researchers design effective data localization and
privacy preservation mechanisms to safeguard critical and sensitive patient data.

A. Verma · P. Bhattacharya (B) · Y. Patel · K. Shah · S. Tanwar

Institute of Technology, Nirma University, Ahmedabad, Gujarat, India
e-mail: pronoya.bhattacharya@nirmauni.ac.in
A. Verma
e-mail: ashwin.verma@nirmauni.ac.in
Y. Patel
e-mail: 20mcec13@nirmauni.ac.in
K. Shah
e-mail: 20mcei18@nirmauni.ac.in
S. Tanwar
e-mail: sudeep.tanwar@nirmauni.ac.in
B. Khan
Hawassa University, Institute of Technology, Awasa, Ethiopia
e-mail: khanbaseem@hu.edu.et

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 233
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_18
234 A. Verma et al.

Keywords Anonymization · Big-data · Data localization · Healthcare · Privacy

preservation

1 Introduction

Privacy preservation in healthcare and big data has received much attention from
industry as well as from academic researchers. The problem has become more preva-
lent due to the widespread storage of data on the network. Electronic medical care has
improved a lot from a manual paper-based framework, where every detail of patient
is stored in files towards digital electronic health record (EHRs) stored in databases
[1]. This provides remarkable benefits such as high performance, better precision,
and extensive accessibility. The hospital chain or set of hospitals have their data set,
and sharing medical records among the chain of hospitals exploits the privacy of
patients personal data [2]. Thus, data sharing among all those entities should be done
securely without record linkages and violation of security concerns. For the same,
the data is required to be localized properly, and at the same time, privacy needs to be
addressed [3]. There are multiple entities such as doctors, patient family members,
nurses, data providers, patients, pharmaceuticals, and many others. Each one of them
has access privileges to shared data, and only a few entities are authorized to view,
update and delete the EHR, as EHR information is sensitive, and access should be
in consensus to the patient only.
EHR contains patients’ medical records as well as private information such as
name, gender, age, guardian detail, and address, and loss and modification of such
data might create a severe impact on patient life and hospital management system.
In today’s technological era, everything is stored on the cloud or Web server, or cen-
tralized database. The attacker can exploit the loopholes of the system and manipu-
late the data if proper security and privacy preservation techniques are not adopted.
Thus, EHR needs a system that ensures the confidentiality, integrity, and availability
of information. In recent times, blockchain-based ledgers are considered a poten-
tial solution to managing distributed healthcare records [4]. Thus, blockchain-based
healthcare solutions are increasingly applied to monitor even the recent COVID-19
pandemic situations, where chronological contact-tracing ledgers are proposed [5].
So a chain of the hospital need to provide authorized access to their system to ensure
security, as well as no other entity in the system, can view the identity of the patient
from EHR. The existing system mostly uses a machine learning model in clouds to
preserve privacy. The privacy of EHR is very crucial in the current novel coronavirus
(COVID-19) pandemic, where different state governments require statistical data to
estimate the population of the region and uses the national population register (NPR)
[6]. It requires data sharing among research centres, hospitals, pharmacy, and govern-
ment entities, and localizing the data require privacy, confidentiality, and security on
patients data [7]. Thus, EHR must not be disclosed with other entities while sharing
its network.
Data Localization and Privacy-Preserving Healthcare … 235

Randomization

Data Attribute
Homomorphic
Encryption

Data Type
Aggeration
Medical
Data-set
Privacy Removal of
Preserving record linkage

Constraint on
Data Analysis Database

Anonymization
Classification SVM

Re-
identification

Fig. 1 A systematic classification of privacy-preserving techniques in healthcare

There is a huge amount of data generated every second from different sensors to
monitor a patient’s health, several medical test reports generated after diagnosis, and
we store that information along with EHR and HIS. The data may be structured or
unstructured, and currently, the traditional database cannot handle this voluminous
data, so we need a sophisticated system to store this data; such data are referred
to as big data. Healthcare data are one of the driving forces of big data, as one
single genome of human deoxyribonucleic acid (DNA), that is, double helix genetic
structure, takes 100–160 gigabyte (GB) of disk space. The expected size of big data
in healthcare in 2022 is around 100 zettabytes (ZB). Figure 1 presents a systematic
classification of privacy preservation techniques in healthcare and highlights the
underlying issues and challenges that healthcare stakeholders face while ensuring
accurate analytics (more requirement of data) and privacy preservation.

1.1 Research Contributions

Following are the research contribution of the survey

• We present the overview of data localization in generic healthcare architecture and
highlight the importance of privacy preservation for healthcare users.
236 A. Verma et al.

• We present the overview and solution taxonomy of privacy preservation techniques

in healthcare.
• An overview of data anonymization (DA) techniques in privacy preservation and
supporting reference architecture are presented.

1.2 Article Layout

The survey is divided into four sections. Section 2 presents the generic architecture
of healthcare systems that includes sensor-based networks and decision analytics on
EHR data sets. Section 3 presents the overview of the big-data-oriented healthcare
scheme and presents the privacy preservation mechanisms in big data healthcare.
Finally, Sect. 4 concludes the paper.

2 Architecture of Healthcare System

In this section, we present the architecture of the healthcare ecosystem. Depending

on the way healthcare stakeholders process the EHRs, many reference architectures
are followed. In some enterprise solutions, EHRs are shared and managed at the dis-
cretion of the patient’s end, and other stakeholders monitor and manage it in a central
fashion [8]. Figure 2 depicts the major components of any healthcare enterprise and
can be considered as a generic solution to handle EHR at different levels. The major
components of the system are as follows.
• Communication Network: The main network that connects all other parts in the
system is considered as the communication network.
• Sensor units: Patients’ body sensors that monitor different body part movements
and their day-to-day activities, and information about the different parameters to
be considered as wireless body area networks.
• Healthcare entities: Different entities of the healthcare system (patients, doctor,
insurance provider, pharmacist), who generally access the EHR and interact with
other entities to provide life-saving services to the patient.
• Underlying protocols: The main communication channel and underlying protocols
that connect all the entities to create one single and unified ecosystem, through
which information is shared to remote entities or centralized systems, where pre-
dictive analysis is done.
The main concern of any healthcare system is the privacy and security of hospital
information systems (HIS) where data are categorized into three different views,
namely view-1, view-2, and view-3. In view-1, we consider the patient with the
sensor units strategically placed at different body locations, which also consists of
the source nodes in the sensor environment and the sink nodes. Overall, view-1
presents the body area network view. In view-2, we consider the communication
Data Localization and Privacy-Preserving Healthcare … 237

Body
Sensors
Humidity Smart Gateway HIS
Temprature Phone
REMOTE
PATIENT
MONITORING Gateway
Tier-3 Doctor Tablet
Storage HOSPITAL
Server

Tier-2
Cellular EMERGENCY SERVICE
INTERNET Policy Gateway
Server

Gateway Prescription system

Motion
ECG Authentica PHARMACY
EEG tion Server
N
F
C
BP Access Gateway Research &
Smart
Proximity phone Point Development Global
Disease Analysis
WIRELESS BAN Tier-1

Fig. 2 Basic architecture of the healthcare system [8]

channel and the cloud service providers (CSP), or edge service providers (ESP), that
communicate through low-powered communication networks like Zigbee, z-wave,
and Bluetooth in a personal range. To address the issues of network management,
researchers have presented a solution that combines fifth-generation (5G) services
for low-latency communication like ultra-reliable low-latency service (uRLLC), and
in the case of real-time response, services like tactile Internet (TI) is considered with
the near-real-time response and high availability of 99.99999% [9]. In view-3, we
consider the healthcare stakeholders like doctors, nurses, drug research labs, and
pharmaceutical companies. The collected data is shared among the different entities
in the system with different access rights, so we cannot impose a single security
mechanism on the entire enterprise. In such cases, the data view remains dynamic,
and every view is materialized with an association [10].
Each component view is different from others and requires different handling of
EHR records. The complete healthcare system and EHRs are required to be protected
from different threats at different attack points in the communication channels, pro-
tocols, and the core network to safeguard the IoT analytics over the ingested big
data. The body area network and communication channel to mobile devices have
a threat environment. The mobile devices collect the data and prepossess it before
sending it to centralized or edge systems. Thus, strong and efficient encryption algo-
rithms are required to enable secure data transfer. Moreover, authorization of data
238 A. Verma et al.

requires signing from different stakeholders in the healthcare chain [11]. Once the
data arrives securely to the main network, privacy preservation, users anonymity, and
access control are required to be established so that exclusive access control rules
are designed.

3 Privacy Preservation in Big-Data-Oriented Healthcare:

A Solution Taxonomy

Big data analytics has provided promising aspects in recent Healthcare 4.0 solu-
tions, where a high amount of data is ingested, cost-effective confidentiality and
privacy-based big data healthcare schemes are required, that can improve and sim-
plify the machine learning and deep learning-based fusion model designed for pre-
dictive analysis of healthcare big data and at the same time manage the privacy of
users. Moreover, in big data ecosystems, the data is constantly transformed owing
to the high velocity and veracity of data [12]. Thus, solution providers are working
towards services that fetch and process data for inside learning to the entities so
that they can take important decisions with a more integrated and preventive model
[13]. Intelligent use of data can speed up decision-making capabilities and provide
better engagement to patients. The data can provide important drivers of healthcare
in different segments of the population with solid proof. Gender, ethnic, and cultural
prejudices, whether inherent or deliberate, can be eliminated via the proper use of
healthcare data. Federal law regulates the information and is not available for all.
Lack of digitization and trained employees to handle the big data creates a barrier to
systematic solutions. The different issues related to this are discussed below.

3.1 Privacy Preservation Techniques

In case EHRs are communicated among different components and entities, then
it is important to ensure security and privacy among the different entities; only
a defined set of stack-holders are authorized to read, write, and manipulate the
data. There are different ways we can preserve privacy in big data such as K -
anonymity, L-diversity, T -closeness, Randomization, Data distribution, and crypto-
graphic techniques like homomorphic encryption and multidimensional sensitivity-
based anonymization (MDSBA). Each technique has its pros and cons as per sys-
tem requirement and usage. The data set can be centralized or decentralized, and
based on that environment, we identify the technology which is most suitable to the
requirement. Figure 3 presents a solution taxonomy of different privacy preservation
techniques in healthcare-oriented big data ecosystems. The details of the techniques
are discussed in the following subsections in detail.
Data Localization and Privacy-Preserving Healthcare … 239

Big Data Privavy

Rules & Regulation

Design Based Based

Encryption Noise Based Anonymization K-Anonymity

L-Diversity
Differencial
Privacy
T-Clossness

Fig. 3 A solution taxonomy of privacy preservation in big-data-oriented healthcare

Data Anonymization: Data anonymization (DA) refers to the process of hiding

sensitive data either by deleting or encrypting the information that connects to the
explicit identification of the patient through his public viewable EHR. Through DA,
the linkage and cross information of patient identity are not revealed, and thus,
other stakeholders can effectively share the EHRs among themselves. Moreover, the
publicly available EHRs collected as healthcare data sets can be used for research
even by third parties. Thus, DA is a critical step that is required before the data is
shared with some untrusted third party [14]. Figure 4 presents the overview of DA
in healthcare ecosystems.
K -anonymity: K -anonymity is a security model that normally applies to protect
the information in data-sharing scenarios where data is collected in large groups by
masking the identity of the user [15]. In various privacy preservation frameworks, the
ultimate objective is anonymity for the unlabelled data subjects [16]. Elimination of
explicit user identifiers is not sufficient to assure DA. Anonymity information can be
re-distinguished by connecting information with another publicly available data set.
The information might incorporate snippets of data that are not themselves excep-
tional identifiers but rather can become distinguishable when joined with other data
sets; these are known as semi-identifiers. Thus, K -anonymity ensures the following
conditions:
• The data of interest should not uncover data that was redacted in the unspecified
partitions or columns. There is a scenario of medical data set where similar kind
of diseases has been caught to men or women, then we can redact the gender
parameter from there [16].
• The sensitive data represented in columns are not different either way for a specific
classification of K . When sensitive parameters are not different in either way for
240 A. Verma et al.

Yes Is Prone to No
Attack?

Reporting Testing of Privacy

Isuues Attack
Design of Enhanced
Anonymization

Data Anonymization

EHS Data Medical

Owner Research

Result of Data
Analysis

Fig. 4 Overview of DA in healthcare ecosystems

a bunch of K records that share the semi distinguishing credits, at that point, the
data set is defenceless against a supposed homogeneity attack. In a homogeneity
assault, the assailant utilizes how it is sufficient to discover the gathering of records
that the individual has placed with if every one of them has a similar worth. For
example, if we consider that all men more than 60 in our data set have malignancy,
and if some person X age, denoted as Age(X ) > 60, in the data set, then a user
might infer that X has a particular disease D explicitly.
• The dimensionality of the information should be adequately low. There are some
high-dimensional attributes like time-series information; it turns out to be very
difficult to give similar protection like low-dimensional information. It is feasible
to recognize an individual by focussing on different sensitive parameters. Addi-
tionally, the dimensionality of information increases frequently along with the
focussed information are scantily circulated. This makes it hard to bunch records
without intensely contorting the information to accomplish K -anonymity.
Data Localization and Privacy-Preserving Healthcare … 241

I -diversity: I -diversity, or l-diversity, refers to the security conservation technique

where the homogeneity assaults are covered. l-diversity is preposterous to carry out
on each type of different data set [3]. The l-diversity model is an augmentation over
the K -anonymity model that diminishes the granularity of information portrayal and
utilizes strategies that include speculation and concealment. The end goal is that a
given record should map onto at least K − 1 other records in the information [15].
The l-diversity model handles the shortcomings of the K -anonymity model where
ensured personalities to the level of K -people is not comparable [17]. A K -anonymity
result is supposed to be l-diverse with the assumption that each of the identity classes
in the table has event ‘l’ that represents values for every quality [18].
T -closeness: An improvement of the l-diversity model is the T -closeness strategy
that diminishes the unprocessed deciphered information [19]. The spectator’s degree
of information on explicit information is restricted whereas the information is not
restricted to the general table containing the data sets. This lessens the connection
between the semi-identifier parameter and the sensitive properties [15]. The distance
between the propagation is estimated through the earth mover’s distance (EMD). For
a categorical parameter, EMD is utilized to scale the distance between the qualities
in it as per the base degree of speculation of qualities in the area of importance.
Randomization: Randomization presents the user random reaction models, termed
as k-mix, instead of utilization of the standard activities i.e. speculation, conceal-
ment, or added substance clamour. In k-mix, conceivable deniability for utilizing
a combination of attributes by the data reconstruct system is achieved [20]. k-mix
component reports the value that draws from the diffusion of the genuine perceptions
with a likelihood p, and it also reports value inspected from other pre-decided (mix)
allocations, which all cover the whole range of the genuine appropriation with a
likelihood of 1 − p altogether.
Homomrophic Encryption—The cryptographic solution to data privacy: Homo-
morphic encryption (HE) can perform significant computations on encoded infor-
mation without decoding the information, and the HE framework can be utilized
to encode the information to address the data protection issue [21]. HE provides
an answer of extraordinary encryption that can settle these issues permitting any
outsider to work on scrambled information without earlier decoding [7].
HE assures proper security for sharing of clinical EHRs. It scrambles clinical pic-
tures and performs valuable procedures on them without breaking the classification
picture. There is a protected and high-vision clinical picture system to ensure the
protection and security. In HE, we perform reversible data hiding (RDH) technique
that implants private information into clinical pictures.
We assume that an encryption scheme G is represented as (P, C, K s , E, D),
where P is the plaintext space, C is the ciphertext space; K s is the key space, E is the
encryption algorithm, and D is the decryption algorithm. ⊕ represents the C − P
operator. Based on the notations, different HE types are presented as follows.
242 A. Verma et al.

Definition 1 Normal HE: Assuming P and L are operations on plain text of data set
P = { p1 , p2 , p3 , . . . , pn } , ks ∈ K s , it satisfies P(E ks ( p1 ), E ks ( p2 ), . . . , E ks ( pn )) =
E ks = (L( p1 , p2 , p3 , . . . , pn )).

Definition 2 Addictive HE: For plaintext pi , p j ∈ P , the corresponding ciphertext

is ci = E( pi ), c j and ci , c j ∈ C. If E( pi + p j ) = E( pei ) ⊕ E( p j ) or D(E( pi ) ⊕
E( p j )) = pi + p j , then the condition is termed as additive HE in G.

Definition 3 Multiplicative HE: For any plaintext pi , p j ∈ P,the correspond-

ing ciphertext is ci = E( pi ), c j and ci , c j ∈ C. If E( pi . p j ) = E( pi ) ⊕ E( p j ) or
D(E( pi ) ⊕ E( p j )) = pi . p j is satisfied then, it is known to be an multiplicative HE
in G.

Multidimensional Sensitivity-Based Anonymization: In the multidimensional

sensitivity-based anonymization (MDSBA) approach, we ensemble the K -anonymity
model with l-diversity and (X, Y ) anonymity. With basic MDSBA, one-dimensional
sets are trained and validated to represent the multi-dimensional plots. It combines
the approach of top-down anonymization and presents considerable authority in
information based on collected user scores. It applies to the big data MapReduce
scheme. In a two-stage multi-dimensional top-down approach, it separates the enor-
mous information into small-sized lumps and provides anonymized data. Every piece
of information requires n seasons of emphasis to reflect the best score.
In MDSBA, it provides base anonymization conveyed in equal cycles, and it dimin-
ishes the MapReduce to reduce the overall complexity. The algorithm of MDSBA
includes a data proprietor, that decides on the identity of item-sets, denoted as
Q(I Ds), the item delicate characteristics S, and provides the anonymized Q(I Ds).
Noise-based techniques: The name itself suggests that privacy is preserved by
adding some amount of noise to the data sets. The noise added must be so much
that it hides the sensitive information of the patient. One of the popular techniques
is differential privacy. In terms of privacy preservation in healthcare, differential
privacy works very well. It improves the systems’ speed, and it does not allow any
person to interfere even if he knows the data of other members [22]. It proposes
the method that makes sure that the output of the two different data sets should be
approximately the same. Thus, the attacker is not able to intercept the differences
between two information sets, and thus, explicit user identifiers are not linked to a
particular user.

4 Conclusion

In this paper, we present a systematic overview of DA and privacy preservation tech-

niques in big-data-oriented healthcare ecosystems. The generic reference architecture
of the healthcare system is highlighted, and major components are discussed. Then,
big-data-oriented healthcare and privacy preservation are discussed, and a solution
Data Localization and Privacy-Preserving Healthcare … 243

taxonomy is presented. Next, we present the overview of DA and presented a ref-

erence to DA architecture in healthcare. Finally, issues and challenges of ensuring
privacy and security are discussed. Thus, the survey assures that effective healthcare
ecosystems are designed with an assurance of secured privacy matrices for patient
EHRs as well as assure correct and responsive analytics.

References

1. Bhattacharya P, Tanwar S, Bodkhe U, Tyagi S, Kumar N (2021) Bindaas: blockchain-based

deep-learning as-a-service in healthcare 4.0 applications. IEEE Trans Netw Sci Eng 8(2):1242–
1255. https://doi.org/10.1109/TNSE.2019.2961932
2. Chen L, Yang JJ, Wang Q, Niu Y (2012) A framework for privacy-preserving healthcare data
sharing. In: 2012 IEEE 14th international conference on e-health networking, applications and
services (Healthcom). pp 341–346. https://doi.org/10.1109/HealthCom.2012.6379433
3. Jin H, Luo Y, Li P, Mathew J (2019) A review of secure and privacy-preserving medical data
sharing. IEEE Access 7:61656–61669. https://doi.org/10.1109/ACCESS.2019.2916503
4. Bodkhe U, Tanwar S, Bhattacharya P, Verma A (2021) Blockchain adoption for trusted medical
records in healthcare 4.0 applications: a survey. In: Singh, PK, Wierzchoń ST, Tanwar S,
Ganzha M, Rodrigues JJPC (eds) Proceedings of second international conference on computing,
communications, and cyber-security. Springer Singapore, Singapore, pp 759–774
5. Zuhair M, Patel F, Navapara D, Bhattacharya P, Saraswat D (2021) Blocov6: a blockchain-
based 6g-assisted uav contact tracing scheme for covid-19 pandemic. In: 2021 2nd International
Conference on Intelligent Engineering and Management (ICIEM), London, United Kingdom,
pp 271–276. https://doi.org/10.1109/ICIEM51511.2021.9445332
6. Verma A, Bhattacharya P, Zuhair M, Tanwar S, Kumar N (2021) Vacochain: blockchain-based
5g-assisted uav vaccine distribution scheme for future pandemics. IEEE J Biomed Health Inf
1–1. https://doi.org/10.1109/JBHI.2021.3103404
7. Sinha K, Majumder P, Ghosh SK (2020) Fully homomorphic encryption based privacy-
preserving data acquisition and computation for contact tracing. In: 2020 IEEE international
conference on advanced networks and telecommunications systems (ANTS). pp 1–6. https://
doi.org/10.1109/ANTS50601.2020.9342834
8. Sahi MA, Abbas H, Saleem K, Yang X, Derhab A, Orgun MA, Iqbal W, Rashid I, Yaseen A
(2018) Privacy preservation in e-healthcare environments: state of the art and future directions.
IEEE Access 6:464–478. https://doi.org/10.1109/ACCESS.2017.2767561
9. Bhattacharya P, Tanwar S, Shah R, Ladha A (2020) Mobile edge computing-enabled blockchain
framework—a survey. In: Singh PK, Kar AK, Singh Y, Kolekar MH, Tanwar S (eds) Proceed-
ings of ICRIC 2019. Springer International Publishing, Cham, pp 797–809
10. Verma A, Bhattacharya P, Bodkhe U, Ladha A, Tanwar S (2021) Dams: Dynamic association
for view materialization based on rule mining scheme. In: Singh PK, Singh Y, Kolekar MH, Kar
AK, Chhabra JK, Sen A (eds) Recent innovations in computing. Springer Singapore, Singapore,
pp 529–544
11. Bhattacharya P, Mehta P, Tanwar S, Obaidat MS, Hsiao KF (2020) Heal: a blockchain-
envisioned signcryption scheme for healthcare iot ecosystems. In: 2020 International Con-
ference on Communications, Computing, Cybersecurity, and Informatics (CCCI), Sharjah,
United Arab Emirates. pp 1–6. https://doi.org/10.1109/CCCI49893.2020.9256705
12. Ghayvat H, Pandya SN, Bhattacharya P, Zuhair M, Rashid M, Hakak S, Dev K (2021) Cp-bdhca:
blockchain-based confidentiality-privacy preserving big data scheme for healthcare clouds and
applications. IEEE J Biomed Health Informat 1–1 (2021). https://doi.org/10.1109/JBHI.2021.
3097237
244 A. Verma et al.

13. Chauhan R, Jangade R (2016) A robust model for big healthcare data analytics. In: 2016 6th
international conference—cloud system and big data engineering (Confluence), pp 221–225.
https://doi.org/10.1109/CONFLUENCE.2016.7508117
14. Mahesh R, Meyyappan T (2013) Anonymization technique through record elimination to pre-
serve privacy of published data. In: 2013 international conference on pattern recognition, infor-
matics and mobile engineering, pp 328–332 https://doi.org/10.1109/ICPRIME.2013.6496495
15. Afifi MH, Zhou K, Ren J (2018) Privacy characterization and quantification in data publishing.
IEEE Trans Knowl Data Eng 30(9):1756–1769. https://doi.org/10.1109/TKDE.2018.2797092
16. Srinivasan NSC, GN (2015) A robust privacy preserving model for data publishing. In: 2015
International Conference on Computer Communication and Informatics (ICCCI), pp 1–6.
https://doi.org/10.1109/ICCCI.2015.7218095
17. Zeng P, Zhang Z, Lu R, Choo KKR (2021) Efficient policy-hiding and large universe attribute-
based encryption with public traceability for internet of medical things. IEEE Int Things J 1–1.
https://doi.org/10.1109/JIOT.2021.3051362
18. Rajendran K, Jayabalan M, Rana ME (2017) A study on k-anonymity, l-diversity, and t-
closeness techniques focusing medical data. Int J Comput Sci Netw Secur 17
19. Binjubeir M, Ahmed AA, Ismail MAB, Sadiq AS, Khurram Khan M (2020) Comprehensive
survey on big data privacy protection. IEEE Access 8:20067–20079. https://doi.org/10.1109/
ACCESS.2019.2962368
20. Abouelmehdi K, Beni-Hssane A, Khaloufi H, Saadi M (2017) Big data security and privacy
in healthcare: a review. Proc Comput Sci 113:73–80 (2017) https://doi.org/10.1016/j.procs.
2017.08.292, https://www.sciencedirect.com/science/article/pii/S1877050917317015, the 8th
international conference on emerging ubiquitous systems and pervasive networks (EUSPN
2017)/The 7th international conference on current and future trends of information and com-
munication technologies in healthcare (ICTH-2017)/affiliated workshops
21. Wang Q, Zhou D, Yang S, Li P, Wang C, Guan Q (2019) Privacy preserving computations
over healthcare data. In: 2019 international conference on internet of things (iThings) and
IEEE green computing and communications (GreenCom) and IEEE cyber, physical and social
computing (CPSCom) and IEEE smart data (SmartData). pp 635–640. https://doi.org/10.1109/
iThings/GreenCom/CPSCom/SmartData.2019.00123
22. Begum SH, Nausheen F (2018) A comparative analysis of differential privacy vs other privacy
mechanisms for big data. In: 2018 2nd International Conference on Inventive Systems and
Control (ICISC), pp 512–516 (2018). https://doi.org/10.1109/ICISC.2018.8399125
Network and Computing Technologies
Design of Facial Recognition Based
Touchless Biometric System

Priya Mishra, Shashwat Singh Mann, Mohit Sharma, and Madhulika Bhatia

Abstract In today’s era, usage of biometric systems for the entrance in a particular
premise is getting common. Fingerprint biometrics is one of the tools which is usually
practiced a lot. Face identification with ID card recognition is a method which is still
not introduced in various schools or universities. The paper presents an approach
for real time face detection and scanning ID cards to identify a person’s identity.
Algorithm like Principal Component Analysis (PCA) is used for reduction of face
space dimension and later used to obtain the image characteristics using Fisher Liner
Discriminant (FLD), generally known as Linear Discriminant Analysis (LDA). LBP
(Local Binary Pattern) is yet another technique used for face detection. The matched
temporary data with the already existing data set is used for system training. The
system testing is basically done by the process of feature vector and pattern matching.
Haar feature-based Cascade Classifier is a machine learning based approach where
a cascade function is trained from a lot of positive and negative images and detects
the objects in other images. Optical Character Recognition (OCR) is an effective
technique for image to text detection with Pytesseract being one of the main tools
used in OCR. Just like the techniques used for the face recognition, the results
found out through image to text techniques are compared with the existing dataset.
Combining both the techniques (face recognition as well as image to text recognition)
will result out in the design of a touchless biometric system.

Keywords Face recognition · Text recognition · Biometric system · Machine

learning

P. Mishra (B) · S. S. Mann · M. Sharma · M. Bhatia

Amity University, Amity Road, Sector-125, Noida, Uttar Pradesh 201301, India
S. S. Mann
e-mail: shashwat.mann@student.amity.edu
M. Sharma
e-mail: mohit.sharma31@student.amity.edu
M. Bhatia
e-mail: mbhadauria@amity.edu

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 247
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_19
248 P. Mishra et al.

1 Introduction

Due to the increased use and improved proficiency of face recognition systems, a lot
of study work has been proposed to make it more accurate and advanced in how the
detection algorithm works. The latest revolution being the Real Time Face Detection,
that is the capability to detect faces. Applications of face detection and recognition
aren’t just limited to features of entertainment and sharing photograph. The recent
advances in this work have shown implications of face recognition in many areas
such as security, biometrics, personal safety, law- enforcement, etc.
With the rise of a new pandemic on the whole world in 2020, people have become
anxious to touch the objects which do not belong to them personally. Corona virus
has affected human minds mentally as well as physically. People have been scared
of even touching small objects. A touched biometric system is a type of biometric
which uses human interaction physically to access a particular object. For example,
fingerprint system is used in various places for identification of a person. From
Aadhar Card centers to various school and colleges, people have been performing
this activity from a long time. What if we want to improvise and change the method
of biometric system to touchless i.e., without physical human interaction? What will
be the need and tools for it? To answers all these questions which arise in our minds,
we came up with a motivation of designing a touchless biometric which is a lot easier
to use and is user friendly where a person will need to show his/her face to a camera
in order to get entrance to a respective institute/school. Further, the method can be
used for marking attendance in a premise. Text recognition is yet another effective
method of using a touchless biometric system for various purposes where the text
written in the ID card of a person is compared with the dataset of that person already
stored.
The basic idea is to design a touchless biometric system that is capable of iden-
tifying or verifying a person using face and text scanning techniques. A database is
created which contains all the details of an individual, and the model is trained with
the existing dataset. If the biometric recognizes the face of a person with the matching
dataset that contains his/her information in it, the system will allow the person to
enter the premise or mark their attendance depending upon where the system will be
used for. The key strength of using a touchless biometric system is to reduce human
effort physically and use this system for security purposes like giving permission to
entry to a particular location only to the people an organization wants.
The challenges which a touchless biometric can come across is the accuracy of
detection of a face as well as taking a lot of time to detect the face. To overcome
these problems, the number of images to be trained must be more than 300 for
a single person in different angles and expressions. For the time taking problem,
several algorithms have been discussed and compared so that the least time taking
techniques are implemented for a better working of the touchless biometric system.
Basic steps involved in the working of the model are as follows:
Design of Facial Recognition Based Touchless Biometric System 249

1. Capturing: Image of the person/text is captured and further stored in the dataset
for comparison. A person needs to get near the camera and show his face so that
the camera captures the image and trains the image for further use.
2. Feature extraction: Feature extraction describes the relevant shape informa-
tion contained in a pattern so that the task of classifying the pattern is
made easy by a formal procedure. Facial feature extraction is the process
of extracting face component features like eyes, nose, mouth, etc. from a
human face image. Facial feature extraction is a very important method for
the initialization of face recognition.
3. Comparison: It is very important to check the image captured and the ID card
information is matching with the existing dataset. This step involves the compar-
ison of the person who is in front of the touchless biometric system with the
pre-existing dataset which is stored in the system containing his/her information.
4. Match/non-match: After the comparison, the permit/non-permit is implemented
according to the output of the detection.
Later, the trained dataset and the image are matched and compared with the
processed dataset available for the face recognition. Once the image is captured, the
person will have to scan his/her ID too. The person will only be given attendance or
be permitted if the image and ID card details matche with the dataset else the system
will reject the person’s entry.
The applications of touchless biometric system can be used in:
1. Various schools, colleges and institutes for the entry and exit purposes inside
the premise
2. A private company for keeping record of employee’s daily routine
3. For marking attendance
Figure 1 shows the architecture of how the touchless biometric system will work.
The first step includes the face image as well as ID card image input to our system. The

Fig. 1 Overview of the steps involved for implementing a touchless biometric system
250 P. Mishra et al.

extraction of face image is done either through Local Binary Pattern or Haar Cascade
technique. On the other hand, the ID card scanned is processed using OpenCV with
Optical Character Recognition technique. Both the inputs are then compared with
the existing data stored in the database and if the inputs get matched with the pre-
existing data of that particular person, the system will allow that person to enter a
premise or vice versa.

2 Related Work

The touchless biometric system works on the different concepts and techniques used
in field of Machine Learning and Deep Learning. Machine Learning is technique
where we train a machine/system to perform a particular work by itself without the
need of human interaction. Deep Learning uses the help of machine learning in the
working of algorithms. Various methodologies are being used to get better result
accuracy for face as well as for text recognition using Machine Learning. Some
research shows the use of MATLAB (Matrix Laboratory) being used for the working
of face recognition. A subfield of Machine Learning called NLP (Natural Language
Processing) is a technique of processing texts and words which is widely used for
many purposes but is yet another effective technique that can be performed for text
extraction. NLP uses Tokenization which is the splitting of a particular sentence and
then assigning each word into a particular category by applying the Parts Of Speech
Tagging (POS) to each split text.
There is a lot of scope for face recognition technique and can be used almost
anywhere including hospitals, a home, schools, etc. In the upcoming digital world,
a touchless system will be used almost everyday in a person’s life. There is a huge
demand of cost-effective machines capable of identifying person’s identity and that
can be done using face recognition and text recognition of an individual’s ID card.
The technique of CNN (Convolution Neural Network) which is an algorithm
used in Machine learning for detecting the faces. CNN is much fast and increases
the efficiency to eliminate errors too. After the face has been detected by the camera,
it is compared with the photo that is stored in the database of that student to update
the attendance. An excel sheet will update the record of attendance on a weekly or
monthly basis and then that file can be sent to the parents of each individual students
to know their performance in class [1].
A proposal showed an attendance management system based on face recognition
technology in the institute/organizations in order to obtain attendance of individuals
and to record their time of entry and exit using OpenCV library and FLTK (Light Tool
Kit). An image is taken as an input and the frontal face is extracted from the image
which is in RBG (Red Blue Green) format and is converted to gray scale of 50 × 50
pixels. The algorithm is used for finding patterns in the data for calculating similarities
between the image captured and the image that is already in the database [2]. Usually,
MATLAB deals with small images for processing the image faster [3]. The average
of all face images from converted trained image in matrix form and then calculating
Design of Facial Recognition Based Touchless Biometric System 251

the eigen faces value is done by Fisher Face method by applying PCA (Principal
Component Analysis) on the image and then applying LDA algorithm to calculate
the average value of each person [4]. Eye detection algorithm is used to reduce
false face detection rate. The facial recognition performances are also greatly proved
by using facial components alignment, contrast enhancement and image smoothing
[5]. A module consisted of text area extraction by using segmentation process and
OCR (Optical Character Recognition) process which will compare the character
recognition model with the trained data using tesseract tools [6]. Normalization is
performed where a printed document is converted into a gray scale image and then
finally converted into a binary image for image to text detection [7].

2.1 Literature Review

See Table 1.

3 Methodology

The methodology is divided into 3 parts:

(a) Methodology for Face Recognition:
Three different tools are used for face recognition namely Haar Cascade Classifier,
OpenCV library and LBP (Local Binary Pattern) algorithm (Fig. 2).
The first step is to acquire the image of an individual and then convert it into
grayscale by using OpenCV.
The second step is to detect the eyes, mouth and face of the person using Haar
cascade classifier/LBP and then normalize face images and train the data and store
information in a doc or XML file.
(b) Methodology for image to text recognition:
For text recognition, two tools are used namely OpenCV library, OCR (Optical
Character Recognition) (Fig. 3).
Firstly, the ID card image is scanned, and the method of noise removal and normal-
ization is implemented. Later, the process of segmentation feature extraction and
classification is adopted using optical character recognition. The extracted text data
is stored in a word or excel file.
After the image is scanned the most important step would be pre-processing
the module which includes Noise removal and Normalization of image, which will
increase the quality of image and also check for any rotational or uniform size in
the image text for better result in text recognition. After pre-processing, the filtered
image is later converted which can be understood by the computer. This is done by
segmentation, feature extraction and classification, where first individual character
252 P. Mishra et al.

Table 1 A brief review of related literature papers

S. No. Aim of the work Methods used in the Limitations Type of
work recognition
[1] Focus on making an CNN (Convolutional Lack of accuracy Face
automated attendance Neural Network) and only a single recognition
system proposal using machine learning
face recognition by module used
capturing the face of
each student and storing
it in the database for the
attendance
[2] An attendance PCA (Principal Limited usage of the Face
management system Component Analysis), methodology in recognition
based on face OpenCV, FLTK different areas of
recognition technology (Light Tool Kit) professional work
in the
institute/organizations
in order to obtain
attendance of
individuals and to
record their time of
entry and exit
[5] Focus on real time face ADA boost and Lack of result Face
recognition using many cascade classifier analysis and recognition
techniques at once as it accuracy prediction
deals with more
complex background
[7] Focus on how we can Artificial neural Hard to interpret the Text
do the recognition and networks, support model and does not recognition
extraction of words vector machine, etc. perform well on
written in images as classifiers and small data sets
linear discriminant
analysis, principal
component analysis,
etc. for feature
extraction
[6] ID card detection by Optical character Difficult to collect Text
using image recognition recognition, the data from recognition
techniques for mainly morphological citizens and less
Indonesian citizens by transformation accuracy for text
collecting their data technique size less than 12
from the citizen ID card Points
an individual has
[3] A proposed solution for Dilation, Sobel Lengthy process Face
face recognition of an operation, MATLAB which can be time recognition
individual person consuming for
detection of an
image
(continued)
Design of Facial Recognition Based Touchless Biometric System 253

Table 1 (continued)
S. No. Aim of the work Methods used in the Limitations Type of
work recognition
[4] Focus on fisher face Fisher face method, Complex process Face
method for recognition principal component and the paper only recognition
and detection of face analysis, linear focused on how to
discriminant analysis improvise the
method using fisher
face method

Fig. 2 Detecting and recognizing a face processing

Fig. 3 Text recognition process

is separated from image and then feature extraction is performed on that to extract
important data from the raw data.
(c) Comparing with the dataset:
The final step is comparing the extracted data of a person collected from the face
recognition and image to text recognition with the existing database. If the data is
matched, it will permit the person to access inside a premise or mark the attendance
according to where the system is used, else the access is denied.

3.1 Techniques and Tools

(a) OpenCV:
OpenCV is a python library used for object detection which includes face as well as
image to text detection. OpenCV is used for implementing the LBP algorithm as well
254 P. Mishra et al.

as the used for the Haar Cascade Classifier. Also, an OCR tool known as Pytesseract
is used for text recognition which uses OpenCV library for implementation. OpenCV
is basically used for real time face vision.
(b) LBP (Local Binary Pattern) Algorithm:
LBP algorithm is an effective algorithm used for detection of face by using the
result in form of binary number by calculating the pixels around the object we want
to detect. Due to less complexity, this algorithm turns out to be one of the least time
taking algorithm for detecting face.
LBP histogram can be defined as:

Hi = x, y I { f l(x, y) = i}, i = 0, . . . , n − 1 (1)

where x, y are the neighborhoods and H represents the histogram.

(c) ADA boost and Cascade Classifier:
ADA boost algorithm is used in face detection for gaining faster face recognition
output as well as for better accuracy result. The ADA boost algorithm is used to get
the best features to check the faces. Best features are first selected as weak classifiers
and then calculated together as a weighted combination of these features to make
a strong classifier. The strong classifier can be used to detect faces. In the equation
down below, hf is the threshold by the strong classifier to detect a face.
Given down below is the formula used for ADA Boost:
⎧ ⎫
⎨ T
1 
T
⎬
1 if αt h t (x) ≥ 2 αt
h f (x) = (2)
⎩ t=1 t=1
⎭
0 otherwise

In Eq. (2),
t initial time,
T final time,
hf final hypothesis,
x training data value.
Cascade classifiers are basically trained on ADA Boost here. The series of classi-
fier consists in a series of tests on the input features. The selected features are divided
into many stages and each stage is trained to become a strong classifier with the help
of the weak classifiers. At every stage, the Haar cascade classifier checks if there is
any image present or not and if not, the whole window will be discarded. Here, the
process of execution will be faster and the number of stages will give a predetermined
accuracy. Haar Cascade Classifier is used for face recognition just like in the local
binary pattern technique yet not as effective as LBP, it produces a great accuracy for
the time taken to recognize the face.
(d) Optical Character Recognition with Pytesseract:
Design of Facial Recognition Based Touchless Biometric System 255

OCR is a technique used for text recognition and extraction. It is used to identify
and recognize the text written inside an image by visualizing it and converting it into
an electronic form. Pytesseract is a tool used with OCR to scan and then recognize
the text written in the ID card and extract it in a text file format.
OCR cannot be used for face recognition as it is limited for extraction of words
written on an object. OCR is preferred over deep learning techniques just because of
the simplicity and better accuracy of extracted text image.
Different libraries are used in Python for the working of Face recognition and
text recognition code. The effective libraries used for implementing the touchless
biometric system are:
1. numpy: This python library is used for storing multi-dimensional arrays and
matrices. The text as well as face recognition method will require a large amount
of data to be stored which can be done easily using numpy library.
2. Tensor Flow: Tensor Flow is an effective python library which is used for object
detection. This library is used for text image detection.
3. Python Imaging Library (PIL): Python Imaging Library is used to enhance a
text written on an image. It increases the sharpness and contrast of the captured
text image so that the extraction of the words written on it can be done easily
without producing any errors which can happen if the text is blur or if the text
written on it is not sharp enough.

4 Results and Discussion

After running the setup of face detection using Haar Cascade and LBP (Local
Binary Pattern), it was found that the model runs successfully with somewhat no
instantaneous changes between the faces in the image. With several alterations and
researching through the web, even though the face detection for still image is very
accurate, for moving frame, the speed of changing the frames multiplies the difficulty
exponentially. The LBP algorithm was found to be the least time taking algorithm to
detect a face as compared to Haar Cascade classifier.
After successfully comparing the sample images from Fig. 4 and taking the mean
of the time taken, the results were as follows (Fig. 5).
The above graph represents the time graph of three different face samples. After
comparing the time taken with both the algorithms, it can be seen that LBP algorithm
is taking comparatively less time to detect the face as compared to Haar cascade
classifier resulting in better accuracy (Table 2).

5 Conclusion

The combination of face recognition with text recognition is a unique touchless

biometric technique which can be used in various places according to its need,
256 P. Mishra et al.

Fig. 4 Comparing Haar Cascade and LBP detection time

Fig. 5 Graphs comparing the time taken to detect the images

Table 2 Comparison of the techniques used with respect to time

Sr. No Technique used Mean accuracy of the time taken (From sample images taken
from the above)
1 Haar cascade classifier 0.129 s
2 LBP algorithm 0.037 s

whether it’s for personal use or for entry in a college/school. For any security
purposes, this system can be an advantage as it will not only check the facial detection
but will also scan the ID cards if necessary. It turns out to be user friendly because if
a person forgets to bring his ID card along, he/she can just use the facial recognition
tool to enter the premise. This system can also be used for marking attendance in
Design of Facial Recognition Based Touchless Biometric System 257

schools or in a company too. Combining face recognition with image to text recog-
nition can be beneficial as they give a positive output and combining both of them is
what makes the touchless biometric recognition system different than other systems
out there.

References

1. Nandhini R, Duraimurugan N, Chokkalingam SP (2019) Face recognition based attendance

system. Int J Eng Adv Technol (IJEAT) 8(3S). ISSN: 2249-8958
2. Saha A, Kar N, Karma M (2012) Study of implementing automated attendance system using
face recognition technique. Int J Comput Commun Eng 1(2)
3. Ronanki S, Gundu S, Baratam R, Mounika P, Kumar JR (2017) Face detection and identification
using svm
4. Anggo M, Arapu L (2018) Face recognition using fisher face method. J Phys Conf Ser
1028(1):012119. https://doi.org/10.1088/1742-6596/1028/1/012119
5. Zhang X, Gonnot T, Saniie J (2017) Real-time face detection and recognition in complex
background. https://doi.org/10.4236/jsip.2017.8200
6. Satyawan W (2019) Citizen ID card detection using image processing and optical character
recognition. In: The 3rd international conference on computing and applied informatics 2018
IOP Conf Ser: J Phys Conf Ser 1235:012049. IOP Publishing https://doi.org/10.1088/1742-
6596/1235/1/012049
7. Manwatkar PM, Singh KR (2015) Technical review on text recognition from images. In: IEEE
sponsored 9th international conference on intelligent systems and control (ISCO), Coimbutur,
https://doi.org/10.1109/ISCO.2015.7282362
Seeker Optimization for Linear Phase Fir
Filter Design

Harmandeep Kaur, Satish Saini, and Amit Sehgal

Abstract In this paper, filter design procedure is analyzed using seeker optimization
technique to design an optimal finite impulse response (FIR) filter with desired
parameter specifications. Using these algorithms, the optimum impulse response
coefficients of the different finite impulse response filter are determined to fulfill
their exact output. A fitness function based on the error is generated in which we
can specify the desired levels of δ p and δ s individually. Simulative results of these
techniques are presented and compared in conformity with their select band and reject
band ripples. The techniques developed are mainly useful for digital signal processing
applications as clearly indicated by the results that it forms a good platform for newer
applications.

Keywords Seeker · Bat · FIR filter

1 Introduction

With the various advancements in the transformation of digital signal, filters based
on the digital techniques are gaining much more importance instead of analog filters.
Digital filtrate is a system that uses digital signal as input and provides digital signal as
output after filtering. These are categorized into two: FIR and IIR filters reliant to size
of its impulse. Each filter possesses its own advantages and disadvantages. Designing
parameters for FIR filter, which are to be considered during the designing of filters,
are mentioned as: select band frequency, reject band frequency, ripples, fading, filter
coefficients, filter order, etc. [1–4]. FIR filter has numerous advantages over IIR filter
which leads to its popularity among the researchers. The implementation of FIR filter
is much simpler as it requires no feedback. By making the coefficients of the filter to be

H. Kaur (B) · S. Saini

School of Engineering (Electronics and Communication Engineering), RIMT University, Mandi
Gobindgarh, Punjab, India
A. Sehgal
School of Engineering and Technology, Sharda University, Greater Noida, UP, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 259
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_20
260 H. Kaur et al.

symmetrical, linear phase response can be achieved; thus, it allows consistent group
delay [5]. Filter designing involves determining the filter coefficients and filter order
which satisfy the required filter stipulations, for example, the frequencies of select
band and reject band, select band and reject band width, and select band and reject
band ripples as well as attenuation. The FIR filters feasibly be projected using various
techniques as an illustration window design technique [6–9], frequency sampling [6,
10] weighted least squares design [6], and Parks–McClellan method (equiripple or
optimal method) [2, 11]. Traditionally, windows method is the most common of
all these. But, in case of windows method, a fabricator will have to be negotiable
between one or two parameters.
So, optimization techniques gained a lot of interest of the candidate researchers
to design the filter with appropriate design conditions such as evolutionary optimiza-
tion techniques. Previously, the swarm-based optimization techniques and genetic
algorithms (GA) were used to implement FIR filter design such as particle-based opti-
mization (PSO). Various techniques in particular Parks–McClellan method, genetic
algorithm (GA) [12, 13], particle swarm optimization (PSO) [12, 14, 15], differ-
ential evolution (DE) optimization [16–18], teaching–learning-based optimization
(TLBO) [19–21] seemed to be exerted for the projection procedure of FIR filters.
Yang proposed an innovative evolutionary optimization technique identified to be
Bat algorithm [22]. Severino et al. designed a FIR filter by the usage of PSO and
Bat techniques [23]. It is being used in almost every field such as optimization [22,
24], filter designing [23], image processing [25, 26], data mining, feature selection,
fuzzy logic [27, 28], artificial neural networks [29, 30], and many more. SOA [31]
mimes the doings of individual hunt populace to resolve actual optimization issues.
It works on the basis of human searching to reach at the desired optimum solution. A
group of humans known as population are taken. Each individual in this are known
as seekers. Each seeker has its own center position, search radius, trust degree, and
search direction according to which it updates its position. Seeker optimization finds
its use in almost every field of study especially in engineering and technology and
are successfully implemented in various problems such as in [32] SOA method has
also been used for optimal reactive power dispatch problem. In [33], it has been used
for designing IIR filter, and as per the literature review, it is not been used much for
optimizing FIR filter parameters.
In the current paper, comparison is done between the seeker, Bat and PSO opti-
mization algorithms to design the FIR filter. The simulation results based on this
design are presented here. The article is organized as per the aforementioned manner.
Module 1 presents the introduction. Module 2 describes the FIR filter model. In
module 3, filter design using optimization technique is discussed. In module 4,
outcome of the simulative environment is presented. Finally, module 5 concludes
the paper.
Seeker Optimization for Linear Phase Fir Filter Design 261

2 Optimization Algorithms for FIR Filter

The major aim in the present study is to project an optimal digital FIR filter with better
design considerations with the use of seeker optimization technique. The symmetrical
nature of the coefficients requires only half of the coefficients to be optimized which
are then combined to form N + 1 count of coefficients. Various parameters that
are considered during the optimum filter designing are pass band and stop band
frequencies (wp and ws ), flat pass band, highest reject band attenuation, pass band
and stop band ripples (δ p and δs ), and short transition width.
Error function/fitness function/objective function is the major parameter to be
considered in all optimization algorithms. While designing the filter, coefficients
are changing the values in a manner with the purpose of the minimization of error
function. The Parks–McClellan algorithm uses the weighted approximate error for
FIR filter design as presented in (1).

E(w) = M(w)[K (wk ) − K i (wk )] (1)

K (wk ) illustrates the frequency output of desired filter, and (K i (wk )) symbolizes
the frequency output of actual filter. M (w) represents the weighting function. Fixed
proportion of δ p /δs is the key disadvantage of PM technique. An umpteen number
of error equations have been and are being used in variant researches as given in
[34–36]. The Lp standard approximation error is indicated as:
  1p

k
Error = [||K d (wk )| − |K i (wk )||] p (2)
i=0

Minimum value of the error equation will be interpreted as the optimum one. A
new set of coefficients will be generated using the optimal value of fitness function.
By individually specifying the required values for δ p and δs , flexibility in the error
equation may further be improved by specifying the error equation as presented in
[3]
  max
w≤w p |E(w)| − δ p + w≥ws (|E(w)| − δs )
J1 = max (3)

2.1 PSO Algorithm

PSO algorithm is a swarm-dependent technique based upon the swarm of birds or

school of fish that follow the path by learning from their fellow mates. Kennedy and
Eberhart [14] established it in 1995. It has the advantage of not getting trapped in
the local optimal solution but solves the problem at a global approach [15]. Every
262 H. Kaur et al.

individual varies their parameters in accordance with the group of individuals (known
as swarm), especially position which has two values—personal or local best (pbest)
and group (global) best (gbest). Each particle modifies the location and velocity of
its own in consonance with the following equations.
   
(g+1) g g g g
Ve j = w ∗ Ve j + C1 ∗ r1 ∗ pbest j − S j + C2 ∗ r2 ∗ gbestg − S j (4)

g
Ve j symbolizes the jth particle vectors velocity at gth iteration; w states weighting
factor; C 1 and C 2 are the cognitive functions known as positive weighting functions;
g
r 1 and r 2 represent the random values in within 0 and 1; S j is the occurant position
g
of jth particle vector h(n) at gth iteration; pbest j represents the local best of the jth
g
particle at the gth iteration; gbest represents the global best of the entire group at
the gth iteration.
The position is updated in correspondence to the undermentioned equations:
(g+1) g (g+1)
Xj = X j + Vj (5)

2.2 Bat Optimization Algorithm

Bat optimization is a search algorithm depending upon the behavior of bats and
their echolocation capability. It is proposed by Yang [22] and works according to
the search behavior of bats for their food. It senses the distance using echolocation
property and takes advantage of frequency equation. To utilize this algorithm for any
optimization problem, initialized values for velocity, position, minimum frequency,
wavelength, and loudness (speech intensity) values are set to find out the target.
Bat Motion and Variations of Loudness and Pulse Rates
Each bat has its initialized velocity vi , position pi , and pulse frequency qi in a solution
space.

qi = qmin + (qmax − qmin )β (6)

where qmax and qmin are maximal and minimal emission frequencies which are
assigned a value uniformly. Initially, value of a frequency is haphazardly assigned
for each bat and is drawn consistently from [qmax , qmin ]. The values for the vector β
∈ [0, 1] are considered to be the random value taken from a uniform distribution.
The new velocity position vit and position z it of the bats are updated at every time
step t in accordance with the velocity and position equations as follows:
 
vit = vit−1 + z it−1 −− z∗ f (7)
Seeker Optimization for Linear Phase Fir Filter Design 263

where z* states the present best position (result) globally that is taken after the
comparison of the entire solutions of n count of bats.

z it = z it−1 + z it (8)

A fresh solution is bring forth locally for each bat utilizing arbitrary walk after
the selection of a solution among the present best solutions.

z new = z old + εL t (9)

where ε symbolizes random count selected within −1 and 1, i.e., ε ∈ [−1, 1]

L it+1 = αL it (10)

pit+1 = pi0 1 − exp(−γ t) (11)

assuming L i and pi as the loudness value and pulse rate and are needs to be updated
in accordance with the proximity of the prey with an increase in the number of
iterations. On reaching near the prey, the decrease in loudness value is seen, but on
the contrary, the speed of emitted pulse rises. The values of loudness are specified
as per the convenience of the problem to be solved; usually, L o = 1 and L min = 0 are
chosen for any problem where L min = 0 indicates that the search of a bat for its prey
is successful and is currently not sending any pulses for further processing. α and γ
are constant values, and generally, for almost all simulative analysis, these are taken
to be equal, i.e., α = γ .

2.3 Seeker Optimization Algorithm

Seeker optimization (SOA) mimes the doings of individual hunt populace to resolve
actual optimization problems. It works on the basis of human searching to reach at the
desired optimum solution. A group of humans known as population are taken. Each
individual in this are known as seekers. Each seeker has its own center position, search
radius, trust degree, and search direction, and in accordance with these parameters,
each seeker updates its position [31–33]. The final decision-making is done while
considering these four parameters. Each seeker is initialized with a random value,
and afterward, their positions are changed as per the following equation:

Z id (t + 1) = Z id (t) + αid (t)βid (t) (12)

where αid (t) and βid (t) are the step length and search direction of the ith seeker and
dth dimension or variable. βid = 1 indicates that the ith seeker moves in the positive
direction on the dimension, whereas βid = −1 indicates its movement in negative
264 H. Kaur et al.

direction and βid = 0 shows that no movement of the ith seeker. Step length and
search direction are updated at each iteration depending upon the following factors.

2.3.1 Filter Design Using Seeker Algorithm

For designing FIR filters, undermentioned steps are implemented:

Step 1: Determine the controlling parameters for FIR filter such as select band and
reject band frequencies, order of filter, and select band and reject band ripples.
Step 2: Specify the parameters for seeker optimization such as maximum popu-
lation number (np ) of seeker strings, where each string has (N/2 + 1) count of
filter coefficients, maximal count of loops, and initialization for the values for s
number of seekers. Also, initialize the values of highest and lowest frequency and
minimum and maximum value of membership degree value (μmin and μmax ).
Step 3: Initialize the arrays for the position values.
Step 4: Calculate the fittingness value for any and all individuals.
Step 5: Upgrade the best suited location of all individuals and select adjacent
individual of all seekers.
Step 6: Compute the hunt trajectory as well as step size for all seekers, after that
amend the location of each seekers.
Step 7: Upgrade the location for all seekers and measure the fitness from these
updated values. Update the new solutions otherwise repeat the process from step
4.
Step 8: Generate the filter coefficients using these updated solutions by selecting
the best seeker among the entire group.

3 Results

In the current module, the outcome of the codes simulated on MATLAB environment
for designing an optimal FIR filter of the all four kinds of filters is presented. Filter
order is kept to be 40 after comparing the different orders for the same problem. The
sampling frequency is equal to f s = 1 Hz, and count of frequency samples is 512.
The required guidelines of the filter to be projected using optimization techniques are
taken as: δp = 0.1, δs = 0.01. The codes of all the optimization techniques are run for
100 iterations to procure the optimal output, i.e., number of iterations considered for
the optimization problem = 100. For low pass and high pass filter, cut off frequency
(normalized) is taken as 0.7 and 0.3, respectively. For band pass and band stop filters,
lowest and highest edge frequencies (normalized) are 0.3 and 0.7. Population size
for this design procedure is taken as 20. For PSO optimization, personal learning
and cognitive acceleration coefficient is considered to be equal to 2. Bat parameters
are taken as: Loudness is 0.5; pulse rate is 0.5; minimum and maximum frequency
Seeker Optimization for Linear Phase Fir Filter Design 265

is 0 and 2, respectively. In case of seeker, optimization parameters are as follows:

minimum and maximum membership degree is given as 0.1 and 0.4; minimum and
maximum frequency is 0.2 and 0.6, respectively.
Figures 1, 2, 3, and 4 compare the magnitude responses of actual filter and the
filters projected with PSO, Bat, and seeker optimization algorithms for FIR—low
pass, high pass, band pass and band select filters successively.
Figures 5, 6, 7, and 8 depict a comparison of convergence profile, i.e., the error
function for FIR-LP, FIR-HP, FIR-BP, FIR-BS filters consecutively.
Tables 1, 2, 3, and 4 successively present the various other comparative outcomes
of performance parameters of all techniques for 40th order FIR—low pass, high
pass, band select, and band reject filters. Almost lowest reject band obstacle and
select band ripple is achieved for seeker optimization algorithm while designing all
the kinds of filters. For low pass, high pass, select band, reject band filters seeker

Fig. 1 Comparison of actual filter and filter projected with PSO, Bat, and seeker algorithm for low
pass filter

Fig. 2 Comparison of actual filter and filter projected with PSO, Bat, and seeker algorithm for high
pass filter
266 H. Kaur et al.

Fig. 3 Comparison of actual filter and filter projected with PSO, Bat, and seeker algorithm for
band pass filter

Fig. 4 Comparison of actual filter and filter projected with PSO, Bat, and seeker algorithm for
band stop filter

Fig. 5 Evaluation of convergence profile for low pass filter using PSO, Bat, and seeker
Seeker Optimization for Linear Phase Fir Filter Design 267

Fig. 6 Evaluation of convergence profile for high pass filter with PSO, Bat, and seeker

Fig. 7 Evaluation of convergence profile for band select filter using PSO, Bat, and seeker

Fig. 8 Evaluation of convergence profile for band reject filter using PSO, Bat, and seeker
268 H. Kaur et al.

Table 1 Analysis of comparative variables for FIR-LP filter projected with various techniques
Technique Max reject band Max select band Max reject band Error fitness
attenuation(dB) ripple ripple
PSO −52.5 0.0027 0.002322 46.26
BAT −53.76 0.0029 0.002053 45.17
SEEKER −53.76 0.0029 0.002051 −2.006

Table 2 Analysis of comparative variables for FIR-HP filter projected with various techniques
Technique Max reject band Max select band Max reject band Error fitness
attenuation(dB) ripple ripple
PSO −53.12 0.0028 0.002208 56.96
BAT −53.73 0.0029 0.002053 34.96
SEEKER −53.73 0.0029 0.00201 7.18

Table 3 Analysis of comparative variables for FIR-BP filter projected with various techniques
Technique Max reject band Max select band Max reject band Error fitness
attenuation(dB) ripple ripple
PSO −49.71 0.0056 0.003268 108.4
BAT −9.87 0.0044 0.002461 92.02
SEEKER −52.18 0.0061 0.003426 87.9

Table 4 Analysis of comparative variables for FIR-BS filter projected with various techniques
Technique Max reject band Max select band Max reject band Error fitness
attenuation (dB) ripple ripple
PSO −52.97 0.0019 0.002241 86.96
BAT −53.73 0.005 0.002059 85.71
SEEKER −53.73 0.9997 0.003353 84.4

technique presents lowermost, or near to lowest reject band ripples of 0.002051,

0.00201, 0.003426, and 0.003353, respectively. Seeker design converges to the least
value of error fitness value among all techniques. For low pass, high pass, band pass
and band stop filter, it converges to min error count of −2.006 in 11, 7.18 in 57,
87.9 in 58 and 84.4 in 12 iterations. Thus, it performs best among all the design
approaches discussed in this works
Statistical parameters of FIR filters obtained by the projection procedure for all
the variant optimization algorithms are given in Tables 5, 6, 7, and 8 successively.
Table 5 Statistical data for FIR low pass filter with variant algorithms
Algorithm Select band ripple Reject band attenuation (decibels)
Max Average Variance SD Max Average Variance SD
PSO 0.0027 0.001 0.000000776 0.000881 −52.5 −54.086 6.40958 2.531715
Seeker Optimization for Linear Phase Fir Filter Design

BAT 0.0029 0.001614 0.000003481 0.00059 – 53.76 −54.954 2.41288 1.553345

SOA 0.0029 0.002414 0.0000000881 0.000297 – 53.76 −57.2325 7.588291667 2.754685
269
 270

Table 6 Statistical data for FIR high pass filter with variant algorithms
Algorithm Select band ripple Reject band attenuation (decibels)
Max Average Variance SD Max Average Variance SD
PSO 0.0028 0.002486 0.000000048095 0.000219 −53.12 −58.176 10.48643 3.238276
BAT 0.0029 0.001867 0.0000002667 0.000516 −53.73 −54.938 2.46017 1.568493
SOA 0.0029 0.001643 0.0000003329 0.000577 −53.73 −54.938 2.46017 1.568493
H. Kaur et al.
Table 7 Statistical data for FIR band pass filter with variant algorithms
Algorithm Select band ripple Reject band attenuation (decibels)
Max Average Variance SD Max Average Variance SD
PSO 0.0056 0.005 0.00000063 0.000794 −49.71 −58.1544 26.0031278 5.099326
Seeker Optimization for Linear Phase Fir Filter Design

BAT 0.0044 0.0039 0.0000003333 0.000577 −9.87 −53.1988 332.4233268 18.23248

SOA 0.0061 0.005433 0.0000009733 0.000987 −52.18 −59.4488 25.94118393 5.093249
271
 272

Table 8 Statistical data for FIR band reject filter with variant algorithms
Algorithm Select band ripple Reject band attenuation (decibels)
Max Average Variance SD Max Average Variance SD
PSO 0.0019 0.0014 0.0000001533 0.000392 −52.97 −55.7717 11.43669 3.381819
BAT 0.005 0.003422 0.000001701920 0.001305 −53.73 −57.6767 35.29782667 5.941197
SOA 0.9997 0.749125 0.248754 0.498753 −53.73 −57.6767 35.29783 5.941197
H. Kaur et al.
Seeker Optimization for Linear Phase Fir Filter Design 273

4 Conclusion

In the present article, the projection procedure of FIR digital filters is presented
with the aim to diminish the error function by optimizing the variant filter param-
eters with the use of PSO, Bat, and seeker optimization techniques. The procedure
runs for 100 iterations to minimize the error function by optimizing the filter coeffi-
cients. Different types of analysis are performed on the outcomes of the codes run on
MATLAB environment such as statistical analysis, analysis of convergence profile
and magnitude response analysis, and comparison of the variant parameters of filters,
namely ripples, attenuation, and as forth. Comparison is performed between the PSO,
Bat, and seeker optimization, and seeker optimization presents the best outcomes in
terms of all the criterions.

References

1. Ababneh JI, Bataineh MH (2008) Linear phase FIR filter design using particle swarm
optimization and genetic algorithms. Dig Sig Proc 18(4):657–668
2. Abdel-Rahman EM, Ahmad AR, Akhtar S (2012) A metaheurisic bat- inspired algorithm for
full body human pose estimation. In: Ninth Conference on Computer and Robot Vision, pp
369–375
3. Aggarwal A, Rawat TK, Upadhyay DK (2015) Design of optimal digital FIR filters using
evolutionary and swarm optimization techniques. Int J Electron Commun 70(4):373–385
4. Alam SM, Hasan MT (2010) Performance analysis of FIR filter design by using optimal,
blackman window and frequency sampling methods. Int J Electr Comput Sci 10(1):13–18
5. Antoniou A (n.d.) Digital filters: analysis and design. McGraw Hill, New York
6. Shenoi BA (2006) Introduction to digital signal processing and filter design. Wiley, Canada
7. Coulombe S, Dubois E (1996) Multidimensional windows over arbitrary lattices and their
application to FIR filter design. In: Proceeding IEEE international conference on acoustics,
speech and signal processing. Atlanta, GA, pp 2383–2386
8. Dai C, Chen W, Zhu Y (2006) Seeker optimization algorithm. In: International conference on
computational intelligence and security. Guangzhou, pp. 225–229
9. Dai C, Chen W, Zhu Y (2010) Seeker Optimization algorithm for digital IIR filter design. IEEE
Trans Industr Electron 57(5):1710–1718
10. Dai C, Chen W, Zhu Y, Zhang X (2009) Seeker optimization algorithm for optimal reactive
power dispatch. IEEE Trans Power Syst 24(3):1218–1231
11. Du ZY, Liu B (2012) Image matching using a bat algorithm with mutation. Appl Mech Mater
203(1):88–93
12. Gholami A, Ansari HR (2017) Estimation of porosity from seismic attributes using a committee
model with bat-inspired optimization algorithm. J Petrol Sci Eng 152(2):238–249
13. Harris FJ (1978) On the use of windows for harmonic analysis with the discrete fourier
transform. Proc IEEE 66(1)
14. Ifeachor E, Jervis B (2002) Digitalm signal processing: a practical approach. Pearson, UK
15. Jaddi NS, Abdullah S, Hamdan AR (2015) Optimization of neural network model using
modified bat-inspired algorithm. Appl Soft Comput 37(C):71–86
16. Karaboga N (2005) Digital IIR filter design using differential evolution algorithm. J Adv Sig
Proc 8(4):1269–1276
17. Karaboga N, Cetinkaya B (2006) Design of digital FIR filters using differential evolution
algorithm. Circ Syst Sig Proc 25(5):649–660
274 H. Kaur et al.

18. Kennedy J, Eberhart R (1995) A new optimizer using particle swarm theory. In: Proceeding
IEEE sixth international symposium on micro machine and human science. Nagoa, Japan, pp
39–43
19. Kennedy J, Eberhart R (1995) Particle swarm optimization. Proceeding IEEE international
conference on neural networks. Perth, pp 1942–1948
20. Lemma TA, Bin Mohd Hashim F (2011) Use of fuzzy systems and bat algorithm for energy
modelling in a gas turbine generator. In: IEEE colloquium on humanities, science and
engineering, pp 305–310
21. Litwin L (2000) FIR and IIR digital filters. IEEE Potentials, pp 28–31.
22. Liu G, Li YX, He G (2010) Design of digital FIR filters using differential evolution algorithm
based on reserved gene. In: Proceedings of the IEEE conference on evolutionary computation.
Barcelona, Spain, pp 1–7
23. Liu G, Li YX, He G (2010) Design of digital FIR filters using differential evolution algorithm
based on reserved gene. In: IEEE Conference on Evolutionary Computation. Barcelona, Spain,
pp 1–7
24. Najjarzadeh M, Ayatollahi A (2008) FIR digital filters design: particle swarm optimization
utilizing LMS and minimax strategies. In: IEEE international symposium on signal processing
and information technology. Sarajevo, Bosnia and Herzegovina, pp 129–132
25. Nayak MR, Nayak CK, Rout PK (2012) Application of multi-objective teaching learning based
optimization algorithm to optimal power flow problem. In: 2nd International conference on
communication, computing and security, pp 255–264
26. Parks TW, Burrus CS (1987) Digital filter design. Wiley, New York
27. Parks TW, McClellan JH (1972) Chebyshev approximation for non-recursive digital filters with
linear phase. IEEE Trans Circ The 19(2):189–194
28. Proakis JG, Manolakis DG (2000) Digital signal processing-principles, algorithms and
applications. Prentice-Hall, New Delhi, India
29. Proakis J, Ingle V (2011) Digital signal processing using MATLAB, 3rd edn. Cengage, USA
30. Rao RV, Savsani VJ, Vakharia DP (2011) Teaching–learning-based optimization: a novel
method for constrained mechanical design optimization problems. Comput Aided Des
43(3):303–315
31. Severino AG, Linhares LL, DeAraujo FM (2015) Optimal design of digital low pass
finite impulse response filter using particle swarm optimization and bat algorithm. In: 12th
international conference on informatics in control, automation and robotics, pp 207–214
32. Singh R, Verma HK (2014) Teaching-learning-based optimization algorithm for parameter
identification in the design of IIR filters. J Inst Eng 94(4):285–294
33. Slowik A, Bialko M (2007) Design of IIR digital filters with non-standard characteristics using
differential evolution algorithm. Bullet Pol Acad Sci Tech Sci 55(4):359–363
34. Tamiru AL, Hashim FM (2013) Application of bat algorithm and fuzzy systems to model
energy changes in a gas turbine. Stud Comput Intell 427:685–719
35. Yang XS (2010) A new metaheuristic bat-inspired algorithm. Nat Insp Cooperat Strateg Optim
284(2):65–74
36. Yang XS, Deb S (2010) Eagle strategy using L´evy walk and firefly algorithms for stochastic
optimization. Stud Comput Intell 284:101–111
Cloud Supported Secure e-Card College
Management System

Garima Jain and Rashmi Mishra

Abstract As the term, smart campus attracts professionals and academics from
multiple disciplines, and the technology keeps intervening in every aspect of life;
it becomes inevitable for the smart campus to take place and deploy the future
vision of smart cities. As a first step to achieving this vision, it is very important
to develop a clear understanding of a smart campus. The student e-card management
system (SEMS) is a barcode-based cutting-edge technology and innovative software
deployed in the AWS cloud environment. It uses the core infrastructure of the AWS
cloud, which includes AWS DynamoDB, AWS EC2 cluster, AWS S3, and the REST
API. There are many manual activities in the existing college management system
and lead to problems as they are time-consuming and expensive. The proposed idea
aims to design a student smart card that configures a bar code with each user’s ID
card, and that bar code can be read through the web, mobile app, and card reader.
With that said, the user’s ID card will be multi-functioned now and will serve many
purposes. The users of this application can be anyone who is associated with the
college. The ID card is used for financial transactions within the college’s boundary,
which ultimately serves the biggest commitment of having cashless transactions
across the country. The same ID card can be used to have books from the library
and many more activities. As this paper is a full-fledged cloud computing-based
system, we are also proposing an OTP-based transaction to provide the end-to-end
reliability of the financial transaction. This feature shows the perspective and influ-
ence of SEMS and its versatility, practicality, and usability. This research opens the
doors for future studies to gain a deeper insight into the type of decisions needed to
transform a traditional campus into a smart campus.

Keywords Bar code · REST API · Cloud computing · DynamoDB · Payments ·

SEMS

G. Jain (B)
Noida Institute of Technology and Engineering, Gr. Noida, UP 201306, India
e-mail: garimajain@niet.co.in
R. Mishra
GL Bajaj Institute of Technology and Management, Gr. Noida, UP 201306, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 275
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_21
276 G. Jain and R. Mishra

1 Introduction

In the current scenario, the smart card can play the role of a minicomputer that can
process the data and store it efficiently with the help of a chip called a microprocessor.
Student Management System is a framework for tracking all sorts of tasks done by
students on the premise of presence in class, bus tracking, hostellers can use for food
choice, accessibility of books online in library, college notifications, and events.
As the university has a clear mission of education, it is certainly a huge business
for enterprise, involving immense management activity. Technology changes the
utilization of information and the education system of fellows, which assists with
making their data and reports centralized.
It has to provide many different jobs for many people who are engaged in an
enormous variety of activities. The administrations need to work smoothly as per
the general learning of their clients; they should help, not disturb, the legal client
exercises, and they ought to be cost-effective.
Different university administrations incorporate payments, for example, for
parking, printing, and canteens. Cashless buy is definitely liked for money install-
ment, as it reduces the expenses of dealing with and moving money. Shrewd card
utilized as an electronic satchel gives a method for eliminating out money flow
from university exchanges. The possibility of scholars and staff flexibility between
colleges has increased a wide acknowledgment as communicated by the Bologna
Declaration.
SEMS is a barcode-based multi-platform application software fully deployed on
cloud clusters and supporting all the available user interfaces like android, IOS, web,
etc. A bar code will be configured with each user’s ID card. After the successful
read of the bar code, users are provided endpoint URLs performing specific tasks
through REST. In the era of cutting-edge technologies and increasing demand for
saving physical storage and maintaining the application’s complexity, SEMS uses
full-fledge clusters of AWS cloud to deploy the application. A light-weighted no SQL
database DynamoDB maintains the user’s record, AWS S3 cloud storage to keep the
documents, classroom notes, placement records, etc. Amazon MQ is operational to
send emails to the users if any notification email is needed to trigger users [1–3].
Users: Each entity of the college can be the user of this application. There is a
registration page for every new user. On successful completion of registration, the
user will be given a user ID which normally is the user’s card ID.
Barcode: This is the SEMS application’s main key. A bar code will be created
after successful registration, and that bar code will be configured on the user’s ID
card. For all the transactions, mainly financial transaction, this bar code will serve
as the backbone and makes this application unique from all the existing software of
the college management system.
REST API: It provides endpoint URLs to access the application and its database
and features.
• GET: Retrieves record from the table.
• POST: Inserts record into the table.
Cloud Supported Secure e-Card College Management System 277

• PUT: Modifies an existing record into a table.

• DELETE: Deletes a record from the table.
Encryption: To ensure the security of data and documents, the RSA encryp-
tion/decryption technique will be applied to document sets before uploading or
updating them to cloud storage, ensuring that attackers will not be able to steal
the papers.
AWS Cloud: AWS offers reliable, scalable, and inexpensive cloud computing
services.
DynamoDB: This application uses the Amazon DynamoDB database to store and
process the data. DynamoDB is a no SQL and light-weighted database that provides
fast and predictable performance with seamless scalability and saves you from the
burdens of operating and scaling a distributed database.
Elastic load balancing: To scale the application and balance the load, elastic load
balancing has been used to distribute incoming application traffic across multiple
targets automatically. It can handle the varying load of SEMS application traffic in a
single availability zone or across multiple availability zones.
AWS Lambda: AWS Lambda is a serverless compute service that runs SEMS
code in response to the events and automatically manages the underlying compute
resources for you.
Clusters: A cluster is a defined action for the combination of tasks or services or
related activities. SEMS uses an EC2 cluster that groups different container instances.
Amazon MQ: To manage the active queue from the incoming messages for one
of the modules “chat with experts,” this application uses Amazon MQ services.
Amazon s3: SEMS application uses Amazon s3 features to create a bucket for
each user to personalize the application based upon their requirements.
The working methodologies of Fig. 1 are described as below:
1. Every user of the college has an ID card associated with them. In this step, we
will configure a bar code with that ID card.
2. On the successful read of bar code by the system, the user requests go to the
cloud environment through a REST API call. These calls could be of any type
depends upon the environment
(a) If this is a GET call, then the requested data is fetched from the DynamoDB
database.
(b) For the POST rest call, the data sent from the user is first encrypted and
stored in the DynamoDB database in the corresponding table.
(c) PUT call to rest API updates the existing record of the tables in the
DynamoDB database.
(d) DELETE removes the data from a database.
3. The entire SEMS application is deployed in the AWS EC2 cluster, and elastic
load balancing is used to distribute the load and traffic across several targets.
4. As this application needs huge resources as the number of students increases
rapidly, AWS lambda is used to manage the reserve, which computes the
resources at run time.
278 G. Jain and R. Mishra

Fig. 1 High-level flow of SEMS

5. SEMS software provides the feature of the automatic triggering of emails and
messages with the help of Amazon MQ.
6. To keep the user documents in the cloud environment, each user is provided
a bucket titled “Users ID” with the help of Amazon S3. This bucket will be
password-protected and will keep sensitive documents in the encrypted format
only. The user will have to tell the system which document s sensitive and which
is not.
7. The proposed software supports OTP-based communication for all the financial
transactions and access to all the critical information.
Cloud Supported Secure e-Card College Management System 279

2 Literature Survey

The current ‘university management system’ [4] is either not automated or bulky.
They proposed a system that gives an idea to overcome multiple pitfalls and uses
the user’s ID as an important entity for financial transactions throughout the college
boundary. The problem may be face problems in the crashing system as they work on a
single point of contact. In [5] “Multi-Purpose Student Card System Using Smart Card
Technology Education,” the author shown in the implementation is just one sector
where can adopt smart cards. Others can also take on the adoption to improve their
functionality and usability. It may increase security issues. In [6], author explained
that comparison of planned and implemented schedules, systematic schedule adjust-
ments, and the survival models applied to ridership. In [7], author said that the card is
useful for the students in places like the library, canteen, stationery shops, and online
storage of important documents. From there, the potential and power of smart cards
their versatility and usability. In [8], the author works on the design concept, architec-
ture, and functionality of the “Students’ Electronic Card” computer system. In [9], the
author develops a linear Internet addiction (LIA) model and a neural network Internet
addiction (NIA) model to calculate students’ Internet addiction levels, respectively.
In paper [10], Zigbee technology is made use of, which requires one of the processors
to get connected to the Zigbee and therefore connect to the server for marking the
attendance. It can increase the hardware cost and maintenance cost. In paper [11],
the authors used Wsn’s for marking the attendance, which is ideally very costly for
colleges to implement in real-time just for attendance. Most colleges do not invest
a lot in the attendance marking process as they rely on professors to do the same.
In [12], the author performs a survey about the previously proposed systems and
thereby analyzes the drawbacks of those systems for proposing the advanced and
efficient solution to automation of attendance. In [13], the study is a research in
progress using Smart PLS-SEM about the adoption mechanism of SPS utilizing the
data mentioned above. None of the research work on data that is accessible from
virtually any Internet-connected device also requires higher security. In [14], the
author had given a significant study on smart campus initiative; they took data from
official university sources as a lack of resources. They give a comparison result on
a green campus, smart research, smart learning and smart management in which
the highest number is associated with smart management that is 58%. In [15], the
author explains smart campus in which he considers University Tun Hussein Onn
Malaysia data. According to its conceptual framework, some pillars like the student
pillar, research pillar, academic pillar etc., are considered as a parameter. The result
from the pilot study shows that Cronbach’s Alpha coefficient is 0.993. In [16], author
mainly focusing on factors choosing e-learning were: ease of use, cost, flexibility,
functionality, and range of features. In [17], the author designs into a student infor-
mation system that can use to input student scores online at SMPN 1 Curug. Authors
mention that the advantage of designing a system is to help teachers to enter grades
when system design is created; in [18], author proposed Integrated Cloud Education
System through a single-window interface called, namely multi-purpose electronic
280 G. Jain and R. Mishra

card which will help the student to avail online education as per their schedule. In
[19], author explains that how efficiently IoT and cloud infrastructure restructure
the traditional education and learning methods. IoT and cloud computing technolo-
gies can provide solutions for a smart and sustainable campus to improve student’s
learning methods and improve the efficiency of everyday activities in the Institution.
In [20], their article focuses on introduces technologies like cloud computing and
then analyzes the needs of the smart campus service platform. On this basis, they
proposed a smart campus service platform based on cloud computing technology
to promote the development of smart campuses in China. But their study fails to
work with security issues. In [21], the author studies reviews of research in different
smart card authentication methods and proposes an improved authentication method
to cover existing security weaknesses, including the security of verification. Finally,
they compared and analyzed their related work.

3 Modules

Figure 2 explains the detailed view of the modules. All the modules are using
distributed database so that the communication among those modules is not time-
consuming and will be decoupled from each other. “Home,” “faculty,” “student,”
“administrator,” “alumni cell,” “library,” “payment,” “others” are the main modules
of the proposed software where can access the first five modules through a web or
an app. In contrast, the barcode is made mandatory for the last three modules. We
can add more modules in the future on the demand of the organization.
Activities performed by the first five modules are mentioned below: -
• “Home”: This module is the home page of the proposed idea.
• “Faculty”: This module is particularly used by the faculties of the college and
facilitates the essential features for the faculty.
• “Students”: This module is particularly used by the students of the college and
facilitates the essential features for them.
• “Administrator”: All the administrator and management will use these modules
to perform there day to day activities.
• “Alumni Cell”: This feature is self-explanatory and can also be accessed by the
college’s ex-students and administrator and faculties members. They can help the
current students of the college to get placed in the companies. With this module, the
features like job referral, events provide communication between the ex-students
and the current students of the college. The module “events” is a chart server
between the ex and current college students where the current students ask help
from their seniors for a job, guide, aptitude, interview questions, etc.
The sub-activities performed by these five modules are described as follows:-
Cloud Supported Secure e-Card College Management System 281

Fig. 2 Modules of SEMS

• “My Account”: This tab gives the features like account nickname, Aadhar
number, email, mobile number, password, profile photo, university rank, and any
other personal details related to students, teachers, and staff members.
• “My Documents”: It has two features: Document upload and document down-
load.
The document upload: Process is shown in Fig. 3.
Document download: To download any document, through REST API
DynamoDB is hit. To download it successfully, the user will have to provide
OTP sent on a registered mobile number.
282 G. Jain and R. Mishra

Fig. 3 Document upload of SEMS

• My notes: These are the classroom notes uploaded by faculty members which
can be downloaded by the corresponding students.
• Course files: This is the teacher’s day-to-day teaching activity for a particular
subject teaching in the particular semester. Once the course is announced as
completed, the H.O.D can track the status.
• Attendance: The emails, text messages, or WhatsApp messages will be automati-
cally triggered to the student’s guardian if their attendance is less than a particular
percentage.
• Departmental activities: This feature enables a particular teacher to know the
departmental activities like FDP, teacher’s meeting, etc.
• Online help to students/chart with experts: If any student seeks online help from
a faculty member or subject matter experts, the student can go for the “chart with
experts,” and the proficient can help the student through “online help to students.”
This feature will be very helpful for the students doing a PhD or doing research
work in any field. If the students want remote help, they will highly appreciate
this feature.
Cloud Supported Secure e-Card College Management System 283

• Research papers/projects undertaken: This module is for the research

papers/patent work/projects done or projects get completed by the students under
that teacher’s guidance. All these documents can be kept in the amazon s3 bucket
with password protected after proper encryption.
• Library book reminder: It reminds the last date of the issued book which must
be returned before the last date.
• Director’s notice: Any notice issued by the director’s office will be available for
the concerned departments, students, etc.
• Admission cell/finance/registrar/placement cell: All these terms are self-
explanatory and will be used by the corresponding departments.
• Short attendance record: This feature is similar to the “Attendance” module,
and they will be sharing the same database so, no manual communication will be
done.
• Alumni cell: This feature is self-explanatory and can also be accessed by the
college’s ex-students and administrator and faculties members. They can help the
current students of the college to get placed in the companies. With this module,
the features like job referral and events provide communication between the ex-
students and the current students of the college. The module “events” is a chart
server between the ex and current college students where the current students ask
help from their seniors for the job, guide, aptitude, interview questions, etc.
• Library, payment, and others will be read through the barcode.
The activities performed by these three modules are defined as below:-
• Library: All the library-related activities like issue a book, return a book, late
fine, etc., will be handled by this module. Now, there is no need to do the manual
entry or no need to read the database tables. Just one bar read to ID card, and the
proposed software will show all the history of that card.
• Payment: This is the most magical part of the software and the heart of this
application.
Pay: Fig. 4 explains the working of the pay module. There are the following
steps to make a payment with the help of SEMS:-
Step 1: The user will go to the cash counter. The cash counter could be anything
in college like the canteen, financial department, library, etc.
Step 2: The user’s ID card will be scanned on the cash counter. Step 3:- On the
successful scan of the user ID card, its details and the available balance on the card
will be displayed on the screen. Else “Please add valid ID card” message will be
displayed.
Step 3: On the successful scan of the user ID card, its details and the available
balance on the card will be displayed on the screen. Else “Please add valid ID card”
message will be displayed.
Step 4: If the appropriate balance to make a payment is available on the card,
then a successful transaction is done. Else the “Insufficient balance, please recharge
your card” message will be displayed on the screen.
284 G. Jain and R. Mishra

Fig. 4 Pay to a party of SEMS

• Recharge a card: Figure 5 explains the working of the recharge a card module.
There are the following steps to recharge a card with the help of SEMS:-
Step 1: SEMS application will read the barcode from the user’s ID card.
Step 2: After the successful verification of ID card, registered email ID and
registered mobile are needed to proceed further. For an invalid ID card, a message
is displayed “Please first make a Registration.”
Step 3: Valid user will get the OTP on both:—Registered email ID and on a
registered phone number.
Cloud Supported Secure e-Card College Management System 285

Fig. 5 Recharge a card of SEMS

Step 4: On successful OTP verification, will choose the mode of payment, e.g.,
net banking, Paytm, debit /credit card, etc., and the message displayed that “Your
card has been recharged with rupees X.”
• Available balance: This module shows the balance available on a card.
• Mini statement: Figure 6 explains the working of the mini statement module.
There are the following steps to Mini statement with the help of SEMS:-
286 G. Jain and R. Mishra

Fig. 6 Flow of mini statement

Cloud Supported Secure e-Card College Management System 287

Step 1: User’s ID card is scanned from the barcode reader.

Step 2: On the successful read of the barcode, the date range to get a mini
statement of the associated card is entered.
Step 3: For the dates which are in range, the mini statement of the associated
card is sent on the registered email ID and a message will be prompted to the user
as “mini statement sent successfully on registered email ID.”
• Canteen: The ID card with bar code will be used to make any payment at the
canteen too and have the same flow as the “pay” sub-modules of the payment
module.
• Transport: The ID card with bar code will be used to make any payment at the
transport desk and have the same flow as of “pay” sub-modules of the payment
module.

4 Analysis the Gathered Data

Therefore, we had seen that IoT, cloud computing, and the latest technology like
big data analytics had proven their efficiency in participating smart applications
into a campus design to transform it into a smart campus. Although, despite these
technologies’ existence, little has been reported in the literature to define the complete
set of criteria that underpin the development of a smart campus. The smart card
enables secure, fast, flexible, and easy accessibility facilities in a smart campus.
As a case study for presenting the smart card in a campus facility that can store a
photo ID and e-wallet and enable library lending privileges, recreation, medical center
services, e-learning, alumni data, student data, and controlled access to university
facilities. Furthermore, some of the other applications addressed in the literature are
summarized in Fig. 7. Therefore, we can conclude that a smart card is an essential

Test Scenarion of Survey form

100 86 90 96 100
86 95
75

50

Fig. 7 Test scenarios from survey result using questionnaire

288 G. Jain and R. Mishra

Smart E-Card Survey Ranking

120
95 98 98
100 8789 928993
859189 9195 8987 919290
76 7984 75 7569
80 65 69 6568
74 697572
56
60
40
20
0

Very Important Moderately Important Slightly Important

Fig. 8 Smart e-card survey ranking using questionnaire

criterion for the smart campus framework, allowing for quick identification of all
transactions and a personal database possible to log in via the cloud.
These results, therefore, indicate that students are more interested in using elec-
tronic cards to facilitate and streamline services and processes than in physical
mobility and accessibility. So, Fig. 8 shows smart E-carding ranking.
It shows the average score attributed to the e-card criterion and its importance
within the different facilities of the university campus, such as its importance for
monitoring student attendance in a classroom, student residential activities, library
activities, loan, as a wallet to process payments and recode student data (e.g., student
information, admission, transcript, degree information, student, and activity records).
The results showed that the use of electronic cards to personal record data scored
the highest among respondents, followed by their use for library activities and
payments. For analyzing the consistency, the concluded result is analyzed by applying
Cronbach’s alpha (coefficient of reliability) using SPSS by using a formula.

M x̃
α= (1)
ỹ + (M − 1)x̃

The coefficient is analyzed to be 0.995, as we know the resulted reliability is

efficient when it is more than 0.70.

5 Conclusion

An ID card configured with the bar code-based software deployed in the AWS EC2
cluster using all the core features of cloud computing has been presented. The system
offers the advanced version of the recently proposed applications. It introduces the
least complex, time, and space-saving software that will be useful for all the users of
the college/university. AWS Dynamo DB no SQL database has been used to facilitate
Cloud Supported Secure e-Card College Management System 289

incremental scalability and high performance with the ease of cloud administration,
reliability, and table data model. REST API has been used to support the client–
server framework and facilitate communication between the user and the cloud.
From the point of view of management, this innovation will make their day-to-
day activity easier. At the same time, the students will enjoy the software’s core
features to boost up their career, and it will be helpful for students’ knowledge
enrichment. The user ID card serves many purposes: as can be treated as a firearm
for all transactions, including financial transactions inside the college premises. A
few modules like chat with experts and payment through ID card have been intro-
duced with any college management software. In the COVID scenario, this system
provides. Thus, the proposed education system model provides smart, economic, and
environmentally sustainable campus.

6 Future Scope

In the future, we willing to use blockchain-based digital technology and IoT in student
smart cards, which is an immense revolution as it helps educational institutions to
keep a clear digital transcript, a ledger of records, and payments for each student.
Furthermore, the technology RFID, which many researchers use and very costly, can
be replaced by Bluetooth with IoT to estimate the overall cost for optimal design. In
IoT, we will use sensors and actuators to flow data between sender and receiver.

References

1. Narula S, Jain A, Prachi (2015) Cloud computing security: amazon web service. In: 2015 Fifth
international conference on advanced computing and communication technologies, Haryana,
pp 501–505. https://doi.org/10.1109/ACCT.2015.20
2. Varia J, Mathew S (2014) Overview of amazon web services. Amazon Web Services, pp 1–22
3. Alqahtani A, Gull H (2018) Cloud computing and security issues—a review of amazon web
services. Int J Appl Eng Res 13(22):16077–16084
4. Jain S, Garg R, Bhosle V, Sah L (2017) Smart university-student information management
system. In: International conference on smart technologies for smart nation (SmartTechCon),
pp 1183–1188. https://doi.org/10.1109/SmartTechCon.2017.8358555
5. Omar S, Djuhari H (2014) Multi-purpose student card system using smart card technology.
In: Information technology based proceedings of the fifth international conference on higher
education and training, pp 527–532. https://doi.org/10.1109/ITHET.2004.1358229
6. Pelletier M-P, Trépanier M, Morency C (2011) Smart card data use in public transit: a literature
review. Transp. Res. Part C: Emerg. Technol. 19(4):557–568. https://doi.org/10.1016/j.trc.2010.
12.003
7. Singh A, Chopra A, Imran M, Singh R (2017) Designing and implementation of smart card
technology for the students of higher education. Int J Comput Sci Mob Comput 6(1):144–149
8. Materka A, Strzelecki M, Dȩbiec P (2009) Student’s electronic card: a secure Internet
database system for university management support. In: On internet–technical development
and applications. Springer, Berlin, Heidelberg. pp 59–72. https://doi.org/10.1007/978-3-642-
05019-0_8
290 G. Jain and R. Mishra

9. Peng W, Zhang X, Li X Using behavior data to predict the internet addiction of college students.
In: International conference on web information systems and applications. Springer, Cham, pp
151–162. https://doi.org/10.1007/978-3-030-30952-7_17
10. Badhe S, Chaudhari K, Kale S, Mane T (2016) Smart attendance management system. Int J
Comput Appl (0975–8887) Nat Conf Adv Comput Inf Technol 4(5):10–14
11. Alassery F (2019) A smart classroom of wireless sensor networks for students time attendance
system. In: IEEE Integrated STEM Education Conference (ISEC), Princeton, NJ, USA, pp
324–331. https://doi.org/10.1109/ISECon.2019.8881998
12. Upadhye PV, Nishigandha SP, Patmase BA, Vyawahare AR, Bharate A (2019) A survey paper
on various smart attendance systems. Int J Res Appl Sci Eng Technol 7(11):755–758
13. Morita M, Naruse K, Uesugi S, Okada H (2018) A study on adopting smart payment system. In:
5th international conference on behavioral, economic, and socio-cultural computing (BESC).
IEEE, pp 144–147. https://doi.org/10.1109/BESC.2018.00037
14. Musa M, Ismail MN, Md Fudzee MF (2021) A survey on smart campus implementation in
Malaysia. JOIV: Int J Inf Vis 5(1):51–56
15. Musa M, Ismail MN, Md Fudzee MF (2021) Smart campus implementation in Universiti
Tun Hussein Onn Malaysia: towards a conceptual framework. J Phys Conf Ser 1860(1). IOP
Publishing
16. Puška A et al (2021) Students’ satisfaction with E-learning platforms in Bosnia and
Herzegovina. Technol Knowl Learn 26(1): 173–191
17. Fauziah Z, Latifah H, Rahardja U, Lutfiani N, Mardiansyah A (2021) Designing student
attendance information systems web-based. Aptisi Trans Technopreneurship (ATT) 3(1):23–31
18. Paul, S, Bandyopadhyay K, Roy A (2021) A study on integrated cloud education system.
In: Advances in electrical and computer technologies: select proceedings of ICAECT 2020.
Springer Singapore
19. Revathi R, Suganya M, Gladiss Merlin NR (2020) IoT based cloud integrated smart classroom
for smart and a sustainable campus. Proced Comput Sci 172:77–81
20. Li Y (2020) Research on building smart campus based on cloud computing technology. In: 2020
5th International Conference on Mechanical, Control and Computer Engineering (ICMCCE).
IEEE
21. Dowlatshah K et al (2020) A secure and robust smart card-based remote user authentication
scheme. Int J Internet Technol Sec Trans 10(3):255–267
An Optimization of VM Allocation
and Migration in Cloud Computing
Systems for High Throughput and Low
Tolerance

Iram Warsi, Hiresh Gupta, and Monika

Abstract Virtualization is among the most popular procedures in the Cloud

Computing scenario since it involves creating an effective virtual server operating
system and storage devices. This will help the user by providing numerous machines
at a similar time and allowing different users to be assigned to different physical
requests of the programme. The efficient scheduling and optimization scenario for
resource allocations and virtualization will be discussed in this research paper. Cloud
Virtualizations also help with task burdens by replacing out-of-date computing proce-
dures with more accessible, cost-effective, and efficient alternatives. The efficiency
of the system is measured in terms of throughput, execution time, downtime, and
Service Level Agreements (SLA) violations. The proposed method is the ability to
accomplish high performance and is well suited to achieving the desired results.

Keywords Service level agreements (SLA) violations · VM migrations ·

Virtualizations · Cloud infrastructure

1 Introduction

Cloud computing is the transfer of on-demand services, such as applications, storage,

and processing power, through the web and on a pay-per-use basis. Cloud computing
services today include everything from Natural language processing (NLP) to
storage, connectivity, and computing power, artificial intelligence (AI), and ordinary
desktop apps. Cloud systems can now distribute any service that does not require

I. Warsi (B) · H. Gupta

Shri Ram Murti Smarak College of Engineering and Technology, Bareilly, Uttar Pradesh 243202,
India
H. Gupta
e-mail: hiresh.gupta@srmscet.edu
Monika
Deen Bandhu Chottu Ram University of Science and Technology, Murthal, Sonipat, Haryana
131001, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 291
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_22
292 I. Warsi et al.

Fig. 1 Cloud computing

connectivity

being physically adjacent to the processor hardware. Cloud computing provides a

wide range of services, from consumer features, for instance, aim of data analysis on
multimedia devices to enterprise benefits such as hosting big data and running all of
their workloads in the cloud. Cloud computing is becoming increasingly appealing as
a default option for various software programme manufacturers, who are gradually
giving their presentations as online services rather than discrete goods or models.
However, the drawback to cloud computing processes in the sense that additional
costs and risks for concerns, perhaps increasing some real-time process failures [1,
2] (Fig. 1).

2 Cloud Computing Diversities

There are mainly three types of diversities in the cloud system and are discussed
below:

2.1 Infrastructure-as-a-Service

There are several types of cloud computing models to choose from. This model
describes the fundamental framework of a cloud scenario, which includes virtual
servers, data storage backup, and networking. This is fantastic for businesses which
are dealing with modest applications from the beginning and wish to control almost
each but every component on their own, but it does necessitate that businesses have
the practical capabilities to control services at that level [3].
An Optimization of VM Allocation and Migration in Cloud … 293

2.2 Platform-as-a-Service

Platform-as-a-Service is the layer up version which deals with the storage, virtual
servers and comprises of the tools and software that cloud developers need to build
software-based applications that includes middleware, management of the databases
and tools used for the development [4].

2.3 Software-as-a-Service

It is concerned with the delivery of services based on applications that the majority
of people use on a regular basis. The final customer will interact with the service via
online applications or browsers; hence the basic hardware and operating system are
inadequate [5, 6].

3 Related Work

The VM Placement problem is a multi-objective optimization problem that tries to

optimize the conflicting objectives while also making competent use of multidimen-
sional resources; this was formulated by Xu and Fortes [7]. Users may avoid using
hotspots in virtualized data center environments using this technology, which also
helps to reduce energy usage. Users are able to deal effectively with the large solution
space by the proposed genetic algorithm modified. The innovative methodology is
based on the Fuzzy multi-objective evaluation, which assisted the user in combining
goals that were determined to be contradictory. In comparison with single objec-
tive methods and bin-packing algorithms, this method shows better performance.
Jamali and Malektaji [8] recommended a grouping of genetic algorithms that is
improved. A technique was introduced which is considered competent for encoding
and generating solutions. This strategy reduces the difficulty of managing a large
number of VMs, as well as resource waste and power consumption. The crossover
method has provided a positive result as there is a significant reduction in the cost and
operation efficiency. Here, the VM Placement is modelled is considered as a vector
packing problem which finds solutions to alleviate the utilization of power, turn off
the vacant servers and increase the competence of resources. A sub-optimal SLA
aware resources allocation optimization system was designed by Huang et al. [9].
To estimate the amount of resources that are utilized considering the SLA specifica-
tion more accurately with the help of prediction mechanism that uses support vector
regression (SVRs). Then, using a genetic algorithm-based technique, VM needs can
be changed dynamically. By using a system that uses SVR as a prediction mecha-
nism and then applies a genetic algorithm, we are able to lower the total waiting time
of the apps that are running. By providing the promised Quality of Service (QoS)
294 I. Warsi et al.

to the client, cloud providers can attain more profits by the proposed method. An
Optimal VM placement algorithm in which expense is reduced for hosting the VMs
in the environment which has several cloud providers taking into consideration as
per the demands of future and uncertainty of price; this was introduced by Chaisiri
et al. [10]. It arrived at its decision after solving stochastic integer programming with
two stages of resources. The effectiveness and performance of this method can be
evaluated with the help of Numerical analysis and simulations. As these tests show,
this method may be utilized for effective provisioning, it fits the requirements of both
customers and suppliers.
A novel approach was proposed by Tordsson et al. [11] which try to optimize
the placement of VMs across numerous cloud providers bearing in mind the criteria
specified by the clients an interoperability layer is provided by the cloud broker. It
aids in the placement of price–performance tradeoffs. Constraints with respect to
load balance, maximum budget and minimum performance. etc., are some client
specifications by which the allocation of VM is steered. When we look at the eval-
uation findings, we can see that when there is consumption in more than one cloud,
efficiency increases.

4 Proposed Work

This section discusses a potential approach for efficient virtualization for balancing
the load and reduced execution time for high-performance assessments. The proposed
solution is providing efficient solutions for the virtualization process in cloud
systems. The cloud system mainly facing the problem of overloading and under
loading hosts which are producing failures of the executions of the tasks and
processes. The load is also a major problem, as it is one of the most important
parameters for controlling the operation of cloud systems. Firstly the job specifica-
tions are initialized and then the machines are deployed with the VM. The machines
here are termed as the hosts on which the VM are deployed. Then the execution
of the services as requests starts through which the CPU utilizations are achieved.
Through CPU utilization the under load and overload of the hosts are obtained. After
evaluating the overload of the hosts then min–max scheduling is processed for the
completion of the tasks or the processes through which the minimum execution time
can be achieved but it can increase the overhead consumption in the migration process
through which the system’s effectiveness can be degraded. To overcome this situ-
ation the overhead consumption is reduced using moth flame optimization process
which increases the throughput of the virtualization process and also decreases the
downtime which reduces the energy consumption of the hosts and balances the load.
This improves the overall performance of the suggested solution and reduces the
likelihood of failure.
An Optimization of VM Allocation and Migration in Cloud … 295

5 Methodology Flow Diagram

See Fig. 2.

5.1 Step by Step Execution of the Research Methodology

Step 1: User requests for the service by initializing the job specifications.
Step 2: Initializing the VM and VM’s specifications for real-time scenarios, users
and their requests for different services can be in any number.

Fig. 2 Proposed methodology flow chart

296 I. Warsi et al.

Step 3: Analyzing the overload and under load hosts/consumptions based on the
usage by evaluating CPU utilizations. Perform min–max for minimum execution
of the jobs.
Step 4: If overload consumption increases.
Then call optimization.
Step 5: Else evaluate performance. Stop.

6 Result and Discussions

This section covers the simulations and results obtained for the proposed research.
The implementations are done in MATLAB tool and the cloud scenario is generated
which will be helpful for the virtualization and migration process for appropriate
SLA violations.
Figure 3 shows the proposed methodology is able to achieve low energy consump-
tion, which is the desired outcome, by measuring the energy consumption of the
machines for the executed jobs per second. For high executions of dynamic requests,
energy consumption must be low. If the energy consumption is high then the machine
gets overloaded and there can be a lot of failures which can degrade our performance
of the system.

Fig. 3 Energy consumption

An Optimization of VM Allocation and Migration in Cloud … 297

Figure 4 shows these machines Ids which are getting overloaded and under loaded.
The machines which are consuming more than 85% CPU utilizations then will raise
the flags and extract the ids of the machines which are getting overloaded.
Figure 5 shows the migration time of the machines which are getting overloaded
to the host which are ideal and is having high capacity and bandwidth to execute

Fig. 4 Machines overloaded

and under loaded

Fig. 5 Migration time

298 I. Warsi et al.

Fig. 6 Down time

the tasks. For the system to run efficiently, the migration time must be short. If the
migration period is prolonged, the virtual machine will take a long time to migrate
to the host, resulting in high energy usage.
Figure 6 shows the downtime of the system which shows that the proposed
approach is well suited for the execution of the tasks without any failure of the
hosts and virtual machines which means that downtime should be low If downtime
grows, real-time applications running on various cloud platforms may have a high
number of service failures, affecting the real-time environment dependent on servers.
Figure 7 shows the proposed service throughput which shows that the solution
is achieving throughput in migrations and execution process with high uptime and
throughput must be high for the efficient executions of the processes. If due to low
throughput, the execution time of the services increases then there will be heavy
loads in the queues which is not our desired output.
Figure 8 shows the data centres energy consumption on the daily basis. It is
one of the significant parameters through which we come to know the total energy
consumption used for the data centres. It can also be helpful in the analysis based on
power management for the data centres which is the key component of running the
real-time applications on the cloud.
Figure 9 shows the SLA violations using the proposed solution. It can be noticed
from Fig. 9 that the SLA violations are less which shows that there are less chances
of software failures as per the agreements. Service level violation occurs when the
An Optimization of VM Allocation and Migration in Cloud … 299

Fig. 7 Throughput

Fig. 8 Data centres energy consumption

300 I. Warsi et al.

Fig. 9 SLA Violations

services which are requested by the user or a customer have not happened on the
decided time interval. SLA is a very serious concern which can be managed for the
smooth workings of the contracts and services.
Table 1 demonstrates the suggested system’s performance, demonstrating that the
proposed technique is well suited for an efficient system with low latency and little
migration time, which is our intended output (Table 2).

Table 1 Proposed
Parameters Proposed
performance
Throughput 30
Downtime 1.5 ms
Migration time 0.01 s
SLA violation 0.008
Energy consumption 60 mJ

Table 2 Performance
Parameters Base Proposed
comparison
Throughput [11] 3.5 30
SLA violation [12] 0.015 0.008
An Optimization of VM Allocation and Migration in Cloud … 301

7 Conclusion

In a cloud computing context, virtualization is a strategy for connecting existing

resources in a network by dividing existing bandwidth into multiple channels, each
distinct and independent.
Virtual machine migration is one of the most important VM management tasks,
as its success has a direct impact on cloud data centre (DCs) energy efficiency [13].
Large-scale virtualized data centres were employed for VM’s power management,
with resource management groups divided into local and global managers [14]. Virtu-
alization refers to the creation of a virtual platform for the server operating system
and storage devices in Cloud Computing, and it provides a virtual stage of server
operating arrangements and storage solutions. The evaluation of a high-performance
cloud computing scenario is carried out in the proposed methodology. The new frame-
work can provide a stable solution for efficient virtualization and high throughput in
real-time cloud scenarios.

References

1. Jouini M, Rabai LBA (2019) A security framework for secure cloud computing environments.
In: Cloud security: concepts, methodologies, tools, and applications. IGI Global, pp 249–263
2. Jonas E, Smith JS, Sreekanti V, Tsai CC, Khandelwal A, Pu Q, Shankar V et al (2019) Cloud
programming simplified: a Berkeley view on serverless computing. arXiv preprint arXiv:1902.
03383
3. Manvi SS, Shyam GK (2014) Resource management for Infrastructure as a Service (IaaS) in
cloud computing: a survey. J Network Comput Appl 41:424–440
4. McGrath PM, Hicks M, Wiest T, McPherson DC (2014) Controlling utilization in a multi-tenant
platform-as-a-service (PaaS) environment in a cloud computing system. U.S. Patent 8,850,432.
Last accessed 30 Sept 2014
5. Ilager S, Wankar R, Kune R, Buyya R, GPU PaaS computation model in Aneka cloud computing
environments. In: Smart data: state-of-the-art perspectives in computing and applications
6. Valls MG, Cucinotta T, Lu C (2014) Challenges in real-time virtualization and predictable
cloud computing. J Syst Archit 60(9):726–740
7. Xu J, Fortes JAB (2010) Multi-objective virtual machine placement in virtualized data centre
environments: green computing and communications (GreenCom). In: IEEE/ACM Interna-
tional conference on & International conference on cyber, physical and social computing
(CPSCom). IEEE
8. Jamali S, Malektaji S (2014) Improving grouping genetic algorithm for virtual machine place-
ment in cloud data centers. In: 4th International e-conference on computer and knowledge
engineering (ICCKE). IEE
9. Huang JC et al (2013) An adaptive resource management scheme in cloud computing. Eng
Appl Artif Intell 26(1):382–389
10. Chaisiri S, Lee B-S, Niyato D (2009) Optimal virtual machine placement across multiple cloud
providers. In: Services computing conference, APSCC 2009, IEEE Asia-Pacific. IEEE
11. Tordsson J et al (2012) Cloud brokering mechanisms for optimized placement of virtual
machines across multiple providers. Future Gener Comput Syst 28(2):358–367
12. Patel N, Patel H (2017) Energy efficient strategy for placement of virtual machines selected
from under loaded servers in compute cloud. J King Saud Univ Comput Inf Sci
302 I. Warsi et al.

13. Mohammad M, Hemn K (2020) Efficient VM migrations using forecasting techniques in cloud
computing: a comprehensive review. Cluster Comput 23:2629–2658
14. Kim MH, Lee JY, Shah Raza SA et al (2021) Min-max exclusive virtual machine placement
in cloud computing for scientific data environment. J Cloud Comp 10:2
GraDex—Graph-Based Data Analytics
for Extractive Text Summarization

Maher Thakkar, Siddhant Patel, and Jai Prakash Verma

Abstract This paper aims to brief the reader about different Automatic Text
Summarization methods and their efficiency when it comes to providing mean-
ingful summaries. In this paper, we have conducted comparative research between
the BERT model for text embeddings along with K-Means clustering to identify
sentences closest to the centroid for summary selection, and a Word Frequency algo-
rithm that computes the frequency of appearing word, assigns appropriate weights
and selects sentences based on a threshold score. The purpose of this is to compare
the two different approaches, for Reviews and Feedback Analysis of different texts
and their summaries. Through our research, we were able to find that BERT outper-
forms the Word Frequency model according to all the evaluation metrics and this is
clearly demonstrated in the following sections of the paper.

Keywords Extractive summarization · BERT · Word frequency · Natural language

processing · Machine learning · Big data

1 Introduction

In recent years, there has been an exponential increase in the amount of textual data
over the internet. Different sources of information often result in redundancy, repe-
tition, and difficult comprehension for the average user. In such a scenario, manual
summarization can be supplanted by Automatic Text Summarization (ATS) to make
up for the time consumption and related costs. An ATS system provides a computer-
generated summary of the given input, covering important information and discarding
redundant and unnecessary data. The last few years have witnessed tremendous
research in this domain through a wide range of approaches [1]. Text Summarization
can be done in two ways: Extractive Summarization (ES) and Abstractive Summa-
rization (AS). In the former, one identifies important sections in the text, and the
summary generated is a subset of the original in the input. In the latter, the latter

M. Thakkar · S. Patel · J. P. Verma (B)

Institute of Technology, Nirma University, Ahmedabad, India
e-mail: jaiprakash.verma@nirmauni.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 303
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_23
304 M. Thakkar et al.

Fig. 1 General overview of the ATS system

uses advanced Natural Language Processing (NLP) methods to reproduce important

parts of the text after interpretation, examination, and analysis. It offers more cohe-
sion than the other approach and resembles the human approach, but the ease in the
process fetches the attention of a majority towards ES (Fig. 1).
We can observe many steps that constitute parts of the ATS system:
• Pre-Processing: Using various NLP techniques like stemming, lemmatization,
tokenization, etc. to filter the data for processing.
• Processing: Once unnecessary data has been filtered out, we perform the following
steps to identify and rank important sentences from the given text.
– Text Representation: Simplification of text analysis is preceded by an appro-
priate representation of sentences through graph, Bag of Words (BOW) model,
etc.
– Scoring of Sentences: Through chosen parameters, sentences are scored in
importance from the highest to the lowest, to reflect their importance in the
data.
– Selection of top-scored sentences: With things like user need in terms of
number of words and number of sentences, the top-ranked sentences are chosen
to match the desired length.
• Post-Processing: Once the important sentences are obtained, re-ordering and
concatenation can be applied as refinement techniques to obtain the final
computer-generated summary.
• Compare scores for generated summaries: The Recall-Oriented Understudy for
Gisting Evaluation (ROUGE) scoring algorithm is used to calculate the similarity
between generated summaries and the golden handwritten summaries provided
in the dataset. Use the ROUGE score to understand the quality of document
translation and summarization model [2].
In this paper, we present a BERT model-based Extractive Summarizer and
compare it with a Word Frequency Model that utilizes a different approach for gener-
ation of summary and compare the results to derive meaningful inference that could
help identify the best approach for the Extractive Summarization. The two models are
GraDex—Graph-Based Data Analytics for Extractive … 305

compared on the basis of a Review and Feedback Analysis on the Opinosis Dataset
which will be discussed in the subsequent sections of this paper. We discuss the
entire pipeline, wherein the input is taken, pre-processed, processed, and lastly post-
processed for the final resulting summary. The organization of this paper is as follows:
Related Work discusses the literature, related work, and recent advancements made
in relation to text summarization techniques; Methodology and Concepts describe the
methods used for review and feedback analysis; and the Execution section involves
the necessary graphs and tables supporting the analysis and inferences done by us.

2 Related Work

Several media platforms have contributed to the huge amount of textual data
and researchers have since been simultaneously researching the proposal and
improvements in several techniques of text summarization for concise and accurate
summaries [3].
There are various methods that are used for ATS; these include graph-based,
statistical-based, DL-based, semantic-based, Machine Learning (ML)-based, and
sentence centrality. The commonly used traditional approach for ATS is as follows:
1. Find the required features to be kept as weights, and calculate the value for each.
2. Assign scores to every sentence using the feature and corresponding weights,
as shown in Eq. 1.


n
Score(S) = Wi ∗ Fi (S) (1)
i=1

As a part of the earlier approaches, Krishnaven proposed a local scoring and

ranking approach using heading and subheadings in given data [4]. Sentences are
selected based on heading; upon compression ratio, we get an equal number of
sentences from each heading.
El-Kassas et al. [5] adopts a combination of various ATS methods (graph, statis-
tical, semantic, and centrality based) for the extraction of summary. The unsu-
pervised nature it follows does not call for the requirement of any training data.
Here, Recall-Oriented Understudy for Gisting Evaluation (ROUGE) is used as an
evaluation metric, and successful implementation obtained the best results among
state-of-the-art ATS models.
Fang uses an unsupervised graph-based model to compute the relationship
between words and corresponding sentences [6]. They have performed the exper-
iment of real-life datasets; the use of complex matrix operations and a redun-
dancy elimination technique provide theoretical convergence and an improvement
in quality.
306 M. Thakkar et al.

Liu used a Deep Learning (DL) approach using Generative Adversarial Networks
(GANs) with reinforcement learning, undergoes prediction of the abstractive
summary and discrimination from the ground truth summary for the evaluation results
[7]. In comparison, this model turned out to generate more abstractive and diverse
summaries from the given text.
A similar approach was adopted by Song, where they combined Long short-
term memory (LSTM) and Convolutional Neural Network (CNN) for an abstractive
summarization of the given text [8]. The focus here is shifted to semantic phrases for
a fine-grained approach, and the DL execution paves way for a more syntactic and
semantic approach in the end (Table 1).

3 Methodology and Concepts

The following subsections in this section will brief the reader regarding the frame-
work of the ES model used as well as the dataset and application considerations for the
same. It will also include necessary figures and flowcharts to make the understanding
of the framework much better and easier.

3.1 Dataset Considerations

“GraDex—Graph-based Data Analytics for Extractive Text Summarization”: The

application of this type of summarization has been done using a dataset that contains
multiple different, long reviews with multiple human written feedbacks each called
the Opinosis Dataset. In all, there are 51 topics with every topic having approxi-
mately 100 sentences (on average). The reviews were taken from various sources—
Tripadvisor (hotels), Edmunds.com (cars), and Amazon.com (various electronics).
The dataset file also comes with gold standard summaries used as a reference for the
summarization scores which will be discussed in the subsequent sections. The graph-
based BERT approach is seen to have a higher score when utilized over various itera-
tions of different reviews yielding successful analysis results which will be discussed
in the following sections (Fig. 2).

3.2 Text Pre-processing

Figure 3 gives an overview of the methodologies we have adopted as part of our

proposed approach. For better performance of our algorithm, it is essential for us to
convert the text into a more understandable form. The proposed tasks go as follows:
Table 1 Comparison of previous implementations of text summarization
Author Year Objective Cons 1 2 3 4 5 6
Fang et al. [9] 2016 Extractive text summarization for Currently limited to single document Yes Yes No No Yes Yes
word sentence co-ranking implementation
Krishnaveni et al. [10] 2017 Coherent ATS using current scoring Restricted to single document, heading No Yes No No Yes Yes
and ranking wise summary generation
Chen et al. [11] 2018 Text image summarization using Absence of pre- and post-processing No Yes No Yes No No
multi-model RNN steps of implementation
Liu et al. [12] 2018 Use of GANs for abstractive text Absence of pre- and post-processing No No Yes Yes No No
summarization steps of implementation
Mackey et al. [13] 2018 Using big data frameworks for ATS Lack of complete automation for the No Yes No No Yes Yes
GraDex—Graph-Based Data Analytics for Extractive …

summarization process
Gunawan et al. [15] 2019 Using maximum marginal relevance Low F-scores in certain categories of No No Yes No Yes Yes
and text rank for multidocument document
summarization
El-Kassas et al. [5] 2020 Using a graph-based framework for Complex approach due to incorporation Yes Yes No No Yes Yes
ATS for multiple approaches
1. Graph-based approach; 2. Extractive summarization; 3. Abstractive summarization; 4. DL approach; 5. Post-processing setup; 6. Pre-processing set up
307
308 M. Thakkar et al.

Fig. 2 Framework for graph-based extractive summarization

Fig. 3 Pre-processing
methodologies adopted

• Stemming and Lemmatization help reduce words to their bases in order to prevent
duplication and redundancy. There is a lot of unnecessary noise in the form of tags,
stop words, connectors, conjunctions, etc. which must be removed and eliminated
for better efficiency of our model.
• Tokenization is essential for the splitting of text into smaller constituents called
“tokens”.
• Following this, normalization (where necessary) helps in converting all text to the
same lower case, to have everyone on the same field.

3.3 Processing

This section contains all the necessary information regarding the methodologies
explained in the previous section along with the libraries and methods utilized for
the successful analysis of the Feedback and Reviews [14].
GraDex—Graph-Based Data Analytics for Extractive … 309

3.3.1 Word Frequency Model

It gives us the frequency of the word in each document in the corpus. It is the ratio
of the number of times the word is seen in a document analyzed to the total number
of words in that document. It gets increased as the number of instances of that word
within the document is increased.
For pre-processing, NLTK (stands for Natural Language Toolkit) is used for
building python programs that endeavor with human language data for putting
in statistical natural language processing (NLP). It accommodates text processing
libraries for tokenization, parsing, classification, stemming, tagging, and semantic
reasoning [12].
Libraries Used:
• We imported stopwords from the nltk.corpus package to remove the stopwords
defined in the English dictionary. The stopwords are English words which does
not add much meaning to a sentence. They can be ignored without abandoning
the meaning of the sentence.
• We also imported word_tokenize and sent_tokenize from the nltk.tokenize
package. Tokenization in NLP is the process through which a big quantity of text
is split into shorter parts called tokens. In word_tokenize() method, we extract the
tokens from a string of characters by utilizing tokenize. It gives back the syllables
from one word. One word can contain one or two syllables. In sent_tokenize we
will be splitting words from sentences.
• We imported PorterStemmer from the nltk.stem package. Stemming is the process
of producing lexical alternatives of a root/base word.
Functions used:
• create_frequency_table(text_string): In this, we will take the input which will
be a paragraph as a text_string. We will tokenize text into words, stem words
and then remove the stopwords. We will get a bag of word dictionaries (word:
frequency) as the output.
• score_sentances(sentences, freqTable): In this, we will take the input as a list
of sentences and word frequency table. We will be computing the score for each
sentence which would be equal to total word value upon the word count. As the
output, we will get a sentence score dictionary (sentence: score).
• find_average_score(sentenceValue): In this, we will be taking a sentence score
dictionary as the input. Then we compute the average sentence score equal to
the total sentence score upon the sentence number. We should get the average
sentence score threshold as the output.
• generate_summary(sentences, sentenceValue, threshold): In this, we will take
the input as the list of sentences and the sentence value dictionary. We should
get the output as the sentence whose score is greater than the threshold as the
summary.
• summarize_text_wf(text): In this, we will use a paragraph of text. The output
will be a summary of text according to the word frequency algorithm (Fig. 4).
310 M. Thakkar et al.

Fig. 4 Flow of word frequency model

3.3.2 BERT Model with K-Means Clustering

It is easier to cluster text documents with the help of BERT and K-means. We can
apply the K-means algorithm on the BERT embeddings to cluster documents. Similar
sentences are clustered based on their sentence embedding similarity [25].
In this approach preprocessing will not change the output predictions. There are
different cases:
1. Stemming or Lemmatization: BERT uses BPE (Byte-Pair Encoding) to shrink
its vocab size, so words like run and running will eventually be decoded to run
+ ##ing without the need of any separate function for stemming.
2. De-Capitalization: BERT provides two models (lowercase and uppercase)
where one changes your sentence into lowercase, and others will not trans-
pose related to the capitalization of your sentence. So, you do not require to do
any changes here, just choose the model for your use case.
3. Removing high-frequency words: BERT takes the Transformer model, which
does work on the attention principle. So, when you fine tune it on any problem,
it will look only at those words which will impact the output and not on words
which are common in all data.
Like the previous approach, we will be importing modules from the NLTK library
as well as the Scikit-learn library for the K-means clustering algorithm.
Libraries Used:
• Here we also imported Torch which is the open-source machine learning library,
a scientific computing framework, and a script language.
GraDex—Graph-Based Data Analytics for Extractive … 311

• Just like before, we have made use of the word_tokenize and sent_tokenize
modules from the nltk.tokenize package.
• We will be importing BertTokenizer and BertModel from pytorch_pretrained_bert
to instantiate a BERT pretrained model with the help of the BertModel class.
• We will be importing K-Means and pairwise_distances_argmin_min from
sklearn.metrics. The k-means clustering method is an unsupervised machine
learning technique used to get to know clusters of data objects in a dataset. pair-
wise_distances_argmin_min is used to calculate minimum distances between one
point and a set of the points. This function calculates for each row in X, the index
of the row of Y which is closest (according to the specified distance). The minimal
distances are also returned.
• We will be importing numpy. NumPy is used for matrix manipulations and data
handling which is one of the main purposes of this library.
Functions used:
• bertSent_embedding(sentences): We will take a list of sentence tokens as the
input. A list of latent vectors, each vector is a sentence representation should be
the output. Bert model produces 12 layers of latent vector, the ‘last layer’ method
is used here, other choices include average last 4 layers, average all layers, etc.
• kmeans_sumIndex(sentence_embedding_list): We will take a list of embedded
sentence vectors as the input. The output will be a list of indices of sentences in
the paragraph, representing the clustering of the key sentences. K Means is used
here for clustering.
• bertSummarize(text): We will take a paragraph as string as the input. The
output will be the summary including a few key sentences using BERT sentence
embedding and clustering.

4 Results and Evaluation

This section contains the results of the Feedback and Review Analysis done on the
Opinosis Dataset using different models and their evaluation metrics. The neces-
sary plots and data tables are included in the following subsections along with the
inferences (Fig. 5).

4.1 Principle and Evaluation Metric

For researching the two approaches for generation of an extractive summary, we

first consider a single review, out of the 51 different types of reviews each having
approximately 100 lines of text, from the dataset as the reference input to be given
to the two models. In the dataset, we also have the gold standard human-written
312 M. Thakkar et al.

Fig. 5 Flow of BERT model with k-means clustering

summaries for each review to compare our model-generated summaries with. Now,
for the evaluation of the generated summaries, we choose the evaluation metric known
as ROUGE, i.e., Recall Oriented Understudy for Gisting Evaluation. ROUGE is a
popularly used set of metrics used for evaluating automatic text summarization and
machine translation software in the field of natural language processing. The metrics
compare a system-generated summary against a reference.
ROUGE has five different evaluation metrics viz.:
• ROUGE-N: Overlap of N-grams between the generated and reference summaries,
e.g.,
– ROUGE-1 mentions the overlap of unigram(each word) in between the system
and reference summaries.
– ROUGE-2 mentions the overlap of bigrams between the system and reference
summaries.
• ROUGE-L: Longest Common Subsequence (LCS) based statistics. Longest
common subsequence problem takes into account sentence level structure
similarity naturally and identifies longest co-occurring in sequence n-grams
automatically.
• ROUGE-W: Weighted LCS-based statistics that favor side-by-side LCS(s).
• ROUGE-S: Skip-bigram-based co-occurrence statistics. Skip-bigram is any set
of words in their sentence order.
GraDex—Graph-Based Data Analytics for Extractive … 313

• ROUGE-SU: Skip-bigram plus unigram-based co-occurrence statistics.

In our implementation, the ROUGE-N and ROUGE-L metrics were evaluated,
and the results were compared. ROUGE score is composed of:
• Precision: # of overlapping words/total words in the reference summary
• Recall: # of overlapping words/total words in the model generated summary
• F1 -score: # measure of model’s accuracy on dataset.

4.2 Evaluation Results

For the comparison of evaluation results, the subsequent parts of this section will
include tabular data and waveform analysis for the same.
Tables 2, 3 and 4 depict the precision, recall and F 1 -score of ROUGE-1 gram,
ROUGE-2 gram, and ROUGE-L evaluation metrics for both the models. We can
observe that the Word Frequency Model performs slightly better in terms of ROUGE
scoring than the BERT and k-means model. However, on further analysis done by
calculating the mean ROUGE score for the entire Opinosis dataset which contains
51 different reviews with each of their handwritten summaries for reference, we find
that the BERT model performs considerably better than the Word Frequency Model.
The tabular data for the mean ROUGE-1 score, for the entire dataset, of the two
models are as shown in Table 5.
In Fig. 6, we can see a bar graph which depicts the ROUGE-1 mean F 1 -score’s
comparison for the BERT model and the Word Frequency Model as well as a distribu-
tion plot that gives us an idea about the ROUGE-1 F 1 score density for the same. We
can clearly see that the BERT model yields a higher mean F 1 score as well as a better
density curve with wider distribution. Since the F 1 score is a weighted average of the
precision and recall values, where an F 1 score reaches its best value at 1 and worst

Table 2 Comparison of
Parameter BERT and K-means Word frequency model
performance parameters of
cluster model
ROUGE-1 gram evaluation
metric F 1 -score 0.1111 0.1176
Precision 0.0639 0.0666
Recall 0.4230 0.5

Table 3 Comparison of
Parameter BERT and K-means Word frequency model
performance parameters of
cluster model
ROUGE-2 gram evaluation
metric F 1 -score 0.0102 0.0182
Precision 0.0058 0.0103
Recall 0.04 0.08
314 M. Thakkar et al.

Table 4 Comparison of
Parameter BERT and K-means Word frequency model
performance parameters of
cluster model
ROUGE-L evaluation metric
F 1 -score 0.1406 0.1560
Precision 0.0849 0.0924
Recall 0.4090 0.5

Table 5 Comparison of
Parameter BERT and K-means Word frequency model
mean performance
cluster model
parameters of ROUGE-1
evaluation metric for the F 1 -score 0.0867 0.05336
entire dataset of reviews Precision 0.0483 0.0282
Recall 0.5839 0.7329

Fig. 6 ROUGE-1 mean F 1 score and score density for the two models

value at 0, we can say that the BERT model even without hyperparameter tuning
performs better than a conventional frequency model for summary generation.
Similar to the analysis done in Figs. 6 and 7 shows similar results that indicate that
the BERT model has a higher mean precision score as well as a better distribution
meaning that the BERT model was able to generate summaries that were more precise
as per logic and understanding rather than the literal meaning of correct prediction.
Figures 6 and 7 both indicate that when a larger amount of data is to be analyzed
and summary needs to be generated, the F 1 score and precision of the BERT are
much better when the entire data is considered.
GraDex—Graph-Based Data Analytics for Extractive … 315

Fig. 7 ROUGE-1 mean precision and precision density for the two models

5 Conclusion

The purpose of this research paper was to study the performance of the Word
Frequency Model and the BERT and K-means model in the generation of an extrac-
tive summary when a single text as well as multiple texts of large sizes are applied
to each of them. The data gathered was tabulated and plotted on waveforms to derive
useful inferences that may help in further research or the usage/application of the two
models in their respective areas of requirement. The study that was conducted utilized
a BERT model that was not fine-tuned to provide ideal results, however, the generated
summaries and their respective performance parameters still proved to be better than
the Word Frequency Model in the case of larger data size. With that, we can conclude
that in areas of higher precision and quality is required, the BERT model is the ideal
choice of ATS system at the cost of more resource utilization. However, in scenarios
of limited hardware/software capabilities, the Word Frequency Model proves to be
more than sufficient in the classification and generation of summaries.[15]

References

1. Virapat K, Bowen T, Yiming N (2020) Automatic text summarization of COVID-19 medical

research articles using BERT and GPT-2. https://arxiv.org/abs/2006.01997
2. Verma JP, Atul P (2017) Evaluation of unsupervised learning based extractive text summa-
rization technique for large scale review and feedback data. Indian J Sci Technol 10(17):1–6.
ISSN: 0974-6846
316 M. Thakkar et al.

3. Verma JP, Atul P (2017) An extractive text summarization approach for analyzing educational
institution’s review and feedback data. Int J Comput Appl, June 2016
4. Wolf T, Sanh V, Rault T (2019) PyTorch pretrained BERT: the big & extending repository of
pre-trained transformers. https://github.com/huggingface/pytorch-pretrainedBERT
5. El-Kassas WS, Salama CR, Rafea AA, Mohamed HK (2020) Edge-summ: graph-based frame-
work for automatic text summarization. Inf Process Manag 57(6):102264. https://doi.org/
10.1016/j.ipm.2020.102264; http://www.sciencedirect.com/science/article/pii/S03064573203
01047
6. Kheruwala HA, Shah JV, Verma JP (2020) Comparative study of sentiment analysis and text
summarization for commercial social networks. In: Gupta S, Sarvaiya JN (eds) Emerging
technology trends in electronics, communication and networking. Springer, Singapore, pp
213–224
7. Zhang JJ, Chan HY, Fung P (2007, December) Improving lecture speech summarization using
rhetorical information. In: IEEE Workshop on automatic speech recognition & understanding,
2007 ASRU, pp 195–200
8. Ming Z, Pengfei L, Yiran C, Danqing W, Xipeng Q, Xuanjing H (2020) Extractive
summarization as text matching. In: ACL 2020. https://arxiv.org/abs/2004.08795
9. Fang C, Mu D, Deng Z, Wu Z (2016) Word-sentence co-ranking for automatic extractive text
summarization
10. Krishnaveni P, Balasundaram SR (2017) International conference on computing methodologies
and communication (ICCMC) - automatic text summarization by local scoring and ranking for
improving coherence. [IEEE 2017 international conference on computing methodologies and
communication (ICCMC) - Erode, India (2017.7.18–2017.7.19)]
11. Chen J, Zhuge H (2018) Abstractive Text-Image Summarization Using Multi-Modal Atten-
tional Hierarchical RNN
12. Liu L, Lu Y, Yang M, Qu Q, Zhu J, Li H (2017) Generative adversarial network for abstractive
text summarization
13. Mackey A, Cuevas I (2018) Automatic text summarization within big data frameworks. J
Comput Sci Coll 33(5):26–32
14. Wei L, Pei-Ran S, Rui-Li J (2020) A BERT based single document extractive summarization
model. J Comput 31(2):241–249
15. Gunawan D, Harahap SH, Fadillah Rahmat R (2019) Multi-document summarization by using
textrank and maximal marginal relevance for text in Bahasa Indonesia. In: 2019 International
conference on ICT for smart society (ICISS), vol 7, pp 1–5. https://doi.org/10.1109/ICISS4
8059.2019.8969785
Smart Traffic Monitoring with Fog
and Cloud Computing

Varsha Choudhary, Yashwant Singh, and Pooja Anand

Abstract The connectivity among billions of real and virtual things to the Internet
with unique identities, brought the term Internet of Things (IoT) which is constantly
changing the world. IoT penetration in major services is making our lives easier by
managing things from anywhere at any time. Consequently, this ubiquitousness has
vigorously led to exponential increase of data, questioning its efficient storage and
analysis. However, Cloud Computing has been playing a substantial role to manage
the colossal army of data but becomes inefficient for time-sensitive applications like
Smart Traffic Monitoring. Perhaps, this would result in cloud overloading and brings
bandwidth of network at saturation point. This necessitates the use of fog computing
with Smart systems to aid efficient utilization of cloud computing paradigm. Fog
computing is a decentralized computing paradigm that extends cloud services to
the edge of the network. In this article, Fog and Cloud computing based integrated
solutions for smart traffic monitoring are studied, to overcome the downsides of
real-time analytics.

Keywords IoT · Fog computing · Cloud computing · Smart traffic · Monitoring

1 Introduction

Internet of things (IoT) has merged the physical and the digital world together by
changing the physical world unit into the smart world via the global connectivity to
control anything from anywhere around the world. Taking the former, the Internet is

V. Choudhary (B) · Y. Singh · P. Anand

Department of Computer Science and Information Technology, Central University of Jammu,
Samba, J&K 181143, India
Y. Singh
e-mail: yashwant.csit@cujammu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 317
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_24
318 V. Choudhary et al.

a global network of networks made up of millions of personal, institutional, educa-

tional, commercial, and government networks ranging in size from local to world-
wide, all connected via a wide range of electrical, wireless, and optical communica-
tion technology. With the latter, the things which may be considered any material or
person in the physical universe. Thus, we can define IoT as an interconnected world-
wide network that brings interactions between people, things, and things-to-things,
which may be anything else in the world, by assigning them a distinct identity or
name [1].
IoT is more than just linking devices to the Internet. Data analytics methods are
used to process data generated by IoT devices in order to extract useful information
about the system and its users, the operations, and surroundings. Cloud servers store
and analyze the vast majority of IoT data. Cloud computing systems are highly
scalable and may be adjusted on-demand in a “pay-per-use” approach, lowering the
expenditure required to construct the necessary analytics application. Existing data
analytics techniques can handle massive amounts of data stored in centralized cloud
storage. However, for a variety of reasons, cloud data processing fails to match the
current requirements of smart systems [2]. Generally, the cloud servers being located
at a larger physical distance causes a delay, proving to be unreliable for real-time
applications. Moreover, these billions of smart nodes must be generating data, over-
burdening the data center, and saturating the network bandwidth. To exclusively meet
the IoT dynamics, fog computing must assist the cloud to provide better response
time and higher transfer rate. Mainly the fog nodes are just like in between placed
intermediators sharing the load of cloud nodes for efficient systems.
In this article, several cloud and fog computing based integrated approaches are
studied in context to Smart Traffic Monitoring system. The sensor nodes gather
multiple sources of traffic information for data analysis, manage traffic processes,
and save the proper choice for future information presenting by tracking traffic
flow, managing traffic processes, and storing the proper choice for future infor-
mation presentation. These computing-based systems aid in optimizing the road
networking systems, which allowed quick and better communication systems with
better response time in terms of data analysis [3]. The rest of this paper is organized as
follows. Section 2 summarizes the computing technologies for Smart Traffic Moni-
toring, followed by Sect. 3 covering the computing technologies-based solutions for
smart traffic monitoring and their relative tabular comparison. Section 4 presents the
emerging challenges in smart traffic monitoring techniques, followed by conclusion.

2 Computing Technologies-Based Solutions for Smart

Traffic Monitoring

The authors have studied several research papers published from 2016 to date on
IoT, fog, and cloud computing in the traffic monitoring system for critical analysis.
They have been taken from prominent publications and databases such as Springer,
Smart Traffic Monitoring with Fog and Cloud Computing 319

Elsevier, IEEE, ResearchGate, Google Scholar, etc. They are detailed as under this
section. Ruimin Ke et al. suggested a technique for identifying and estimating traffic
flow metrics (such as velocity, intensity, and volume) using UAV footage that involves
four phases. They are Ego-motion irregularity, poor estimate precision in busy traffic,
and powerful computational complexity being addressed by the suggested structure.
The work showed positive results [4]. Diao et al. proposed a unique hybrid model
for accurately forecasting passenger flow with many steps ahead of the view of time,
origin-destination spatial, recurrence and self-similarity, and historical probability
distributions. The findings revealed that the hybrid model can enhance accuracy on
average of 20–50% [5].
In another noticeable work, Alsrehin et al. concentrated on transportation manage-
ment techniques that relied entirely on machine learning and data mining technology
for identifying and forecasting traffic. The steps for establishing intelligent trans-
portation and control systems were given by the authors. The model also specifies
the relationships between road intersections, such as how traffic lights near to one
another contribute to one another [6]. Tunio et al. presented their work by employing
various image processing methods to govern real-time traffic in the MATLAB tool.
The purposed model uses LEDs to display the red and green light signals, as well as
seven segments to display the signal green’s decrementing timer. Furthermore, the
device outperforms current conventional systems in terms of precision, performance,
and ease of use [7].
In [8], Bui et al. proposed the RSDC, inspired by recent breakthroughs in city
sound classification with CNN-based approaches, a novel methodology for conges-
tion classification utilizing road sound datasets has been developed. The implementa-
tions, in particular, show promising results, with accuracies ranging from 92 to 95%
for classifying traffic congestion over various time slots. Huang et al. [9] described a
high-level architecture for vehicular Fog Computing. Also, the security and forensic
concerns, as well as mitigating measures were focused. The application examples
of vehicular Fog Computing and their potential advantages were also discussed [9].
Sookhak et al. had given a definition for Fog Vehicle Computing (FVC) as to supple-
ment the fog computing’s compute and storage capabilities, as well as a complete
framework for the FVC and a variety of important technologies. Effective results
have been found. Some open problems and potential directions in the sense of FVC
were also highlighted for future study [10].
Additionally, Nahri et al. proposed IoV networked infrastructure for real-time
traffic data analytics based on three main layers. The proposed framework consisted
of real-time data collection and analytics modules. Advantages, disadvantages,
results, security issues were also discussed [11]. In [12], Naha et al. addressed Fog
and Cloud Computing patterns along with their technological differences. In this
paper, Fog Computing Taxonomy was also suggested, with debating on the current
research papers and their limitations. The authors have explored open problems and
weaknesses considering Fault Tolerance, Resource Scheduling and Allocation, Tool
Simulation, and Fog-based Microservices. Mohammad et al. introduced the impor-
tance of fog computing in relation to cloud computing and IoT with a model for
mitigation traffic congestion to lower the latency. The authors presented a highly
320 V. Choudhary et al.

dispersed infrastructure and applications that were backed up by the use of loca-
tion identification and mobility. Specifically, focusing on the drawbacks of cloud
computing in context to smart environments [13].
Darwish et al. presented a three-dimensional structure (intelligent computing,
real-time big data analytics, and Internet of Things) to provide real-time ITS big
data analytics in an IoV context. Furthermore, the advantages and disadvantages
of IoV and intelligent computing systems have also been mentioned. In addi-
tion, the different edge computing technologies were compared [14]. Qiong and
others demonstrated an intelligent traffic light management system based on fog
computing. It analyzed and reported the existing traffic conditions at the crossroads
and nearby intersections using a Fog Computing platform with promising results
[15]. In another work, Sangmek et al. used a microcontroller board to measure and
manipulate data in a wireless environment using the architecture design of the Fog-
Cloud Computing traffic model. It entailed surveying relevant studies for a Fog-Cloud
Computing traffic model research, establishing architecture for USN, describing the
Fog-Cloud Computing traffic model design documentation, and evaluating progress
[16]. Erskine et al. proposed the Cuckoo search algorithm (CSA). Several hybrid
optimization algorithms were used comprising firefly algorithm (FA), firefly neural
network, and main distribution establishment (KDE) to authenticate both network
and node tier against all attacks for truthfulness in VANET. Furthermore, the results
were also evaluated [17].
Ning et al. developed a three-layer VFC model in which vehicles serve as Fog
nodes, allowing for real-time traffic management, congestion detection, and other
functions. Followed by some research issues and problems related to VFC-enabled
traffic management were described and highlighted [18]. In [19], Hussain et al.
proposed an Integer Linear Programming (ILP) design for determining the best loca-
tion, power, and quantity of Fog Computing Nodes (FCN) in order to reduce total
network latency and energy usage, with illustrating the operation of the VFNP model
and its related constraints. In recent work, Bhardwaj et al. presented a unique smart
Fog Computing based taxonomy to control vehicular traffic. Mobile edge computing
and mobile Cloud Computing, cloudlets, microdata centres were used in their work.
The role of smart Fog Computing and important features of its architecture were
also discussed with significant results and comparative analysis [20]. Following this
work, Vergis et al. used fog computing technologies for low-cost vehicular traffic
monitoring and proposed a three-tier architecture-based system compromising of
mobile tracking system (MTS), Information gathering system (GPS), and Fog devices
(FD) [21]. Thakur et al. discovered vehicle congestion detection strategies relied on
connected vehicles technology [22]. Tang et al. developed a VFC framework that
was studied from a variety of perspectives, including latency optimization, utility
maximization, and so on. The research shows positive results [23].
In Table 1 few basic traffic monitoring techniques are compared. Some of them
address the use of Machine Learning (ML) and other algorithms like Data Mining
(DM), and Artificial Intelligent (AI). It is found that unmanned aerial vehicles (UAVs)
show high accuracy in terms of performance. Additionally, using these algorithms
shows positive results in traffic monitoring. Table 2 presents the relative contrast of
Smart Traffic Monitoring with Fog and Cloud Computing 321

Table 1 A relative comparison of the basic traffic monitoring techniques

Authors Algorithms used Software/Tools/Model Year Result
used
Ke et al. [4] Optical flow, Unmanned aerial 2019 96 and 87% accuracy
K-means vehicles (UAVs) in estimating
clustering and average speed and
connected graph vehicle count
algorithms
Diao et al. [5] Predictive Novel hybrid DTMGP 2019 20–50% accuracy
analytics model, Gaussian process improvement
algorithm
Alsrehin et al. ML, data mining, Traffic intelligence tool, 2019 –
[6] and AI algorithms OpenTraffic, MOTUS,
Waze, CARTO-Waze
connector, SUMO
Tunio et al. [7] Image processing MATLAB, OTSU’s 2020 Accurate,
algorithm method time-saving, and
ease of use
Bui et al. [8] ML algorithms CNN architecture using 2020 92–95% accurate
(KNN, SVM, RF, SoftMax
XGboots)

Table 2 A comparison of the Fog and Cloud computing based smart traffic monitoring techniques
Authors Algorithms used Software/Tools/Model Year Result
used
Huang et al. [9] Traffic data fusion Fog-assisted traffic 2017 –
algorithm, Intelligent control system
traffic light control,
and Dynamic routing
algorithm
Darwish and Image processing OpenStreetMap, 2018 Energy efficient
Bakar [14] algorithms, Vehicle MongoDB, NodeMCU,
detection, and length Thinger. Io, Magnetic
estimation algorithm Sensors, LCD Unit
Qiong et al. Intelligent control JAVA compiler tool My 2018 Reduced traffic
[15] algorithm, Eclipse congestion
Optimization
algorithm
Erskine and Cuckoo search MATLAB 2019 Convenient for
Elleithy [17] algorithm (CSA), reducing delay
firefly algorithm (FA)
Hussain et al. Optimization GAMS CPLEX solver 2019 Limit energy
[19] algorithm usage
Vergis et al. Data reduction GPS technologies 2020 Provide accurate
[21] algorithms estimation
Tang et al. [23] Greedy offloading Heuristics based 2020 Beneficial results
decision algorithm, scheduling strategies
322 V. Choudhary et al.

the different works in the literature. The readers might quickly recognize already
accessible contributions in IoT-based Fog and Cloud Computing smart traffic moni-
toring systems in this table. They primarily focused their efforts on the Internet of
Things (IoT) enabled fog and cloud computing architectures for smart traffic system,
algorithms used in the process, and the latest technologies and tools used. From the
studied literature it is noticed that there are few works, which precisely emphasize
Image processing algorithms with higher accuracy. Furthermore, these works in the
literature provide insight into IoT, Fog Computing, and Cloud Computing and the
latest techniques augmented for more satisfactory results.

3 Computing Technologies for Smart Traffic Monitoring

Smart Traffic Monitoring is all about controlling the traffic in a smart way with
the use of IoT and AI. A smart traffic management system includes smart bridges,
smart highways, smart street lights, sophisticated 3D vision cameras, and automated
traffic signaling systems. In most countries, in cities and urban regions, transportation
congestion is now becoming a huge problem. Traffic congestion puts a strain on
municipal infrastructure and has a detrimental influence on people’s socioeconomic
life as a result of time spent stuck in traffic. The most serious issue with traffic
congestion is delaying of HPVs, in which the driver will have difficulty reaching his
or her destination in the shortest possible time, particularly in urban areas. Priority
Vehicles (HPV) have been significantly impacted, resulting in vehicles in massive
number to reach at their destinations, in the circumstances in which human lives
are at stake. Since the HPV has no special priority, a set of traffic lights with a
predetermined schedule creates more difficulties in showing up on time [2].
A centralized traffic monitoring and management system is needed to solve this
issue which involves infrastructure that allows for the recording and updation of
congestion data [2]. Smart traffic monitoring prioritizes moderate traffic condi-
tions by analyzing real-time traffic situations, providing congestion-free traffic, and
enhancing conventional ticketing with an electronic E-bill payment system with
speed sensors to notify passengers about speeding. Additionally, provides safe and
punctual public transportation, reduces emissions, and aids in moving in the right
traffic directions via GPS and GIS at intersections and narrow road ends. Figure 1
depicts the smart traffic monitoring system.

3.1 Cloud Computing for IoT

Cloud Computing implies, in the simplest words, storing and accessing data and
programmes over the internet from a remote location or device instead of the hard
drive of our computer. This so-called remote location has many features that are
Smart Traffic Monitoring with Fog and Cloud Computing 323

Fig. 1 Smart traffic monitoring system

substantially different from a traditional remote computer, such as scalability, elas-

ticity, etc. The Cloud is just an Internet metaphor. That is called Local Storage and
Computing when we store data on or run a programme from the local computer’s
hard drive. We need to access our data or programmes over the Internet for it to be
called Cloud Computing. The end result is the same, but Cloud Computing can be
done anywhere, and via any computer with an online connection [24]. Facebook,
Instagram, Twitter, Dropbox, Google Images, Google Drive, iCloud, Gmail, etc.,
are the simple examples of Cloud Computing used by common people in everyday
life. It provides scalability, versatility, agility, and simplicity, which is why its use in
companies is growing rapidly [25].
Vehicular Cloud Computing is designed to aid vehicles with traffic protection as
well as performance. There is tremendous potential for vehicular Cloud Computing
to shift the structure of contemporary vehicular communication. Explicitly, vehicles
underused services may be shared with other vehicles in order to control traffic
during congestion. Vehicle Clouds are divided into three large groups, that is, the
V2V Cloud, the V2I Cloud which is vehicle to infrastructure, and the combination
of vehicle Clouds with other commercial clouds, i.e., integrated Clouds. V2V clouds
are the one in which on the road or in parking lots, vehicles form a cloud to share data
for common services. They are further divided into dynamic vehicular clouds and
324 V. Choudhary et al.

static clouds. Dynamic clouds include moving cars and vehicular sensor clouds. A
cloud of automobiles parked in a parking lot is an example of a static vehicular cloud.
These clouds are often utilized for processing and storage. V2I Clouds can be used
for participatory sensing and cooperative sensing. Integrated clouds are formed when
certain clouds, including such mobile computing and internet clouds, are linked to
vehicular clouds [26]. Via Cloud Computing, network service providers can transfer
tens of millions or even billions of messages instantly. Thus, Cloud Computing
technology will be the IoT promoter with so many features like cost-saving, flexible,
and secure [25].

3.2 Fog Computing for IoT

Cloud computing has numerous intrinsic abilities and is now extensively employed,
having a huge impact on both businesses and people. However, it also has some
drawbacks, including its restriction of communication between the Cloud and the end
devices. A new digital technology, Fog Computing, has been implemented to solve the
problems facing Cloud Computing. In Fog Computing applications, the data collected
is not sent to the Cloud server but to nearby processing devices thereby improving the
service quality and also reducing the latency [13]. Fog Computing is characterized as
a paradigm of distributed computing that expands the services offered to the edge of
the network by the Cloud [27, 28]. Cisco suggests the idea of Fog Computing where
fog nodes are composed to form a Fog network. Any computer having processing,
storage, and network access can be these nodes. The Fog nodes, with main road
or maybe an oil plant, can be deployed anywhere, anywhere on the lane. Within
milliseconds, they function on IoT data. These nodes transport processing infor-
mation to the cloud allowing historical study and relatively long-time preservation.
This method is mostly conducted for productivity purposes (latency reduction and
network bandwidth conservation), as well as for safety and enforcement causes [29].
Figure 2 depicts IoT-based integrated Fog and Cloud Computing structure in smart
traffic monitoring.

4 Emerging Challenges

• Downtime: Downtime is often cited as one of the Cloud server’s main drawbacks.
The traffic monitoring system is unable to cope with the effects of an outage or
slowdown. In Cloud Computing, every component is available online, exposing
potential vulnerabilities. Even the most powerful teams are subjected to tremen-
dous breaches and privacy violations from period to period. This whole system is
expensive too as smart vehicles are costly [9].
• Technological faults: Geofencing, like any technology, has its faults. Most
geofencing systems have found that it has worked most, if not all, of the time.
Smart Traffic Monitoring with Fog and Cloud Computing 325

Fig. 2 Fog and cloud

computing structure in smart
traffic monitoring

Sometimes geofence technology has consistent problems. Geofence can also

cause privacy concerns. It can drain cell phone data and battery life too [13].
• Cost: In order to become more sensitive to the traffic entering the intersection,
there could be additional costs associated with improving traffic light control.
However, the Gaussian process extensively used is computationally expensive
[20].
• Network: In addition to mobility and pollution metrics, one of the challenges of
smart traffic monitoring is to ensure that network metrics perform well. Standard
vehicular communications produce negative effects on the output of all metrics
in the congestion situation. The highest rate is anchored by these conventional
vehicle communications, while emissions are impacted by vehicle rises [30].
• Interoperability: It refers to the capacity of two devices to communicate with one
another, and it is a major issue in traffic management systems. It takes time to
read information during Vehicle to Vehicle (V2V) contact because a link can be
lost, resulting in data loss [31].
• Security: In VANETs and IoV, there are many kinds of threats that can pose a
major issue to vehicular networks. The use of sensor nodes have some drawbacks
like sensor node devices’ fairly simple operation makes them easy to deploy,
and it also exposes them to fraudulent security assaults, as there are often no
comprehensive security systems for these sensors. UAVs are also vulnerable to
hacking [11, 32].
326 V. Choudhary et al.

5 Conclusion

The Internet of Things (IoT) has now become a popular topic of discussion and
marketing strategy in today’s world with its exponential rise and growing sustain-
ability issues. By witnessing its applications everywhere, we have explored its
usability in traffic systems. One of the biggest challenges of the twenty-first century is
traffic jam and therefore there is an urgent need to combat the traffic congestion issue
with automated smart solutions. The intelligent traffic control system isn’t limited
to decreasing traffic congestion, air pollution, and regulating road traffic; it also
needs continuous monitoring and preserving the public’s safety and welfare. This
paper covered the introduction to Smart Traffic Monitoring, brief idea of Fog and
cloud computing in context to their usage in this environment. Then the basic smart
traffic monitoring techniques based on fog and cloud computing were studied. Some
important techniques for smart traffic monitoring like VANET, GPRS, UAVs, Geo
fencing, smart fog computing, etc., were also analyzed. Additionally, the emerging
challenges like privacy, downtime, attacks, etc., in computing technologies for smart
traffic monitoring were highlighted. Via a few improvements and technical imple-
mentations, the smart traffic management system will provide intelligent information
and analysis to current problems. As a prominent tool in every field, IoT is growing
and an intelligent traffic management system will be the future traffic management
scheme for the smart city.

References

1. Madakam S, Ramaswamy R, Tripathi S (2015) Internet of things (IoT): a literature review. J

Comput Commun 3(5):164–173
2. Lah AAA, Latiff LA, Dziyauddin RA, Kaidi HM, Ahmad N (2017) Smart traffic monitoring
and control architecture and design. In: 2017 IEEE 15th Student conference on research and
development (SCOReD)
3. Rabby MdKM, Islam MM, Imon SM (2019) A review of IoT application in a smart traffic
management system. In: Proceeding of the 2019 5th international conference on advance in
electrical engineering (ICAEE), Sept 2019
4. Ke R, Li Z, Tang J, Pan Z, Wang Y (2019) Real-time traffic flow parameter estimation from UAV
video based on ensemble classifier and optical flow. IEEE Trans Intell Transp Syst 20(1):54–64
5. Diao Z, Zhang D, Wang X (2019) A hybrid model for short-term traffic volume prediction in
massive transportation systems. IEEE Trans Intell Transp Syst 20(3):935–946
6. Alsrehin NO, Klaib AF, Magableh AWS (2019) Intelligent transportation and control systems
using data mining and machine learning techniques. IEEE Access 07:49830–49857
7. Tunio MH, Memon I, Mallah GA, Shaikh NA, Shaikh RA, Magsi Y (2020) Automation of traffic
control system using image morphological operations. In: IEEE Xplore, 2020 International
conference on information science and communication technology
8. Bui KNAM, Oh H, Yi H (2020) Traffic density classification using sound datasets: an empirical
study on traffic flow at asymmetric roads. IEEE Access 08:125671–125679
9. Huang C, Lu R, Choo K-KR (2017) Vehicular fog computing: architecture, use case, and
security and forensic challenges. IEEE Commun Mag 105–111
Smart Traffic Monitoring with Fog and Cloud Computing 327

10. Sookhak M, Yu FR, He Y, Talebian H, Sohrabi Safa N, Zhao N, Khan MK, Kumar N, Fog
vehicular computing: augmentation of fog computing using vehicular cloud computing. IEEE
Veh Technol Mag 12(3):55–64, 7994678
11. Nahri M, Boulmakoula A, Karima L, Bath AL (2018) IoV distributed architecture for real-time
traffic data analytics. Procedia Comput Sci 130:480–487
12. Naha R, Garg S, Georgakopoulos D, Jayaraman P, Gao L (2018) Fog computing: survey of
trends, architectures, requirements, and research directions. IEEE Early Access 06:47980–
48009
13. Mohammed TS, Khan OF, Ibrahim AS, Mamlook R (2019) Fog computing-based model for
mitigation of traffic congestion. Int J Simul Syst Sci Technol. ISSN: 1473-804x online, 1473-
8031 print
14. Darwish TSJ, Bakar KA (2018) Fog based intelligent transportation big data analytics in the
IoV environment. IEEE Trans Cybern 06:2570–2580
15. Qiong W, Fanfan H, Xiumei F (2018) The intelligent control system of traffic light based on
fog computing. Chin J Electron 27(6)
16. Sangmek C, Boonnam N (2019) Fog–cloud computing traffic model and performance evalua-
tion for ubiquitous sensor network infrastructure, Oct 2019. IEEE. https://doi.org/10.1109/ICS
EC47112.2019.8974817
17. Erskine SK, Elleithy KM (2019) Secure intelligent vehicular network using fog computing.
MDPI Electronics 8:455. https://doi.org/10.3390/electronics8040455
18. Ning Z, Huang J, Wang X (2019) Vehicular fog computing: enabling real-time traffic
management for smart cities. IEEE Wirel Commun 26(1):87–93
19. Hussain M, Alam M, Bega M (2020) Vehicular fog computing-planning and design. Procedia
Comput Sci 167:2570–2580
20. Bhardwaj A, Goundar S (2019) IoT enabled smart fog computing for vehicular traffic control.
EAI Endorsed Trans Internet Things 5(17):162221
21. Vergis S, Komianos V, Tsoumanis G, Tsipis A, Oikonomou K (2020) A low-cost vehicular
traffic monitoring system using fog computing. Smart Cities 3(1):138–156. https://doi.org/10.
3390/smartcities30100008
22. Thakur A, Malekian R (2019) Fog computing for detecting vehicular congestion, an internet
of vehicles-based approach: a review. IEEE Intell Transp Syst Mag 11(2):8–16
23. Tang C, Wei X, Zhu C, Wang Y, Jia W (2020) Mobile vehicles as Fog nodes for latency
optimization in smart cities. IEEE Trans Veh Technol 69(9):9364–9375
24. Chandrasekaran K, Gujarati D, Porter D (2010) Essentials of cloud computing, p 577. ISBN:
978-0-07-337584-7
25. Srivastava P, Khan R (2018) A review paper on cloud computing. Int J Adv Res Comput Sci
Softw Eng 8(6)
26. Ahmad I, Noor RMd, Ali I, Imran Md, Vasilakos A (2017) Characterizing the role of vehicular
cloud computing in road traffic management. IJDSN 13(5)
27. Dastjerdi AV, Gupta H, Calheiros RN, Ghosh SK, Buyya R (2016) Fog computing: principle’s
architectures and applications. arXiv:1601.02752
28. Yi S, Li C, Li Q (2015) A survey of fog computing: concepts, applications and issues. In:
Proceedings of the 2015 workshop on mobile big data. ACM, pp 37–42
29. Dhingra S, Madda RB, Patan R, Jiao P, Barri K, Alavi AH (2020) Internet of things-based fog
and cloud computing technology for smart traffic monitoring. ScienceDirect JID: IOT 09(11)
30. Butun I, Österberg P, Song H (2020) Security of the internet of things: vulnerabilities, attacks
and counter measures. IEEE Commun Surv Tutorials 22(1):616–644
31. Mehmood Y, Ahmad F, Yaqoob I, Adnane A, Imran M, Guizani S (2017) Internet-of-things-
based smart cities: recent advances and challenges. IEEE Commun Mag 55(9):16–24
32. Hussein ARH (2019) Internet of things (IOT): research challenges and future applications. Int
J Adv Comput Sci Appl (IJACSA) 10(6):77–82
A Review of Precision Agriculture
Methodologies, Challenges,
and Applications

Mohammedshakil Malek, Bachwani Dhiraj, Deep Upadhyaya,

and Dixit Patel

Abstract Internet of things (IoT) furnished recent cultivated procedures can alter
the prevailing agriculture mechanisms by computerizing the complete technique of
farming since crop-management, aquatic irrigation to creating superior judgments
in accordance with real-time monitoring of conservational circumstances, situations
of soil and surroundings of landscape during the prevailing periods, technology
spurred preciseness agriculture explanations have spurred a framework towards stag-
nant and instruction physical agriculture procedures to automatic preciseness posi-
tioning agricultural procedures by the current innovations like agricultural IoT and
AI, WSN assisted harvest checking along with its control and cloud computing.
During the projected assessment, a demanding and elaborated calculation has been
performed to recognize the scope of the study and study the modern knowledge
sparked applications and policies. Furthermore, in this paper, five strands IoAT-
based PA model is presented. The first strands represent the physical level maneuvers,
second strands define safety rules, third strands explain effective data-management
performs, fourth strands deliver operative irrigation prototypes, and the last strands
deliberate technology-aided facilities for aquatic management. Lastly, alongside
upcoming guidelines, definite investigation of the performed literature review in the
visual form accompanied by acquired reviews and academic training were delineated.

Keywords Internet of agricultural things (IoAT) · Precision agriculture (PA) ·

Crops irrigation

M. Malek (B)
F. D. (Mubin) Institute of Engineering and Technology, Gandhinagar, Gujarat, India
B. Dhiraj
Civil Engineering Department, Parul University, Vadodara, Gujarat, India
D. Upadhyaya
Civil Engineering Department, GEC, Godhara, Gujarat, India
e-mail: deep.upadhyaya@gtu.edu.in
D. Patel
Civil Engineering Department, GEC, Surat, Gujarat, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 329
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_25
330 M. Malek et al.

1 Introduction

By 2050, the number of inhabitants on Earth is assessed to ascend to a huge amount of

9100 million occupants. The existing population of 7300 million has plundered high
right around two-thirds of Earth’s assets [1]. Over the most recent 60 years, because
of popularity for wares, like fiber, petroleum, food, wood, and new water, extra-
terrestrial has been allocated to horticulture as compared to eighties and nineties
time period. The realities disclose that there is presently an unmistakable short-
fall in regular assets and if the continuous consumption of assets isn’t managed
in the correct manner, the people in the future won’t have anything to encourage
[2]. Rainwater system is an essential worry for the improvement of agribusiness in
agricultural nations. To determine the water system problem, individual specialists
have put forth a few attempts by proposing imaginative strategies, for example, drip
water system, mechanized sprinkler innovation, and smart. “Smart Farming tech-
nique (otherwise called accuracy agribusiness (PA) utilize most recent innovations”,
for example, IoAT, cyber physical system, big data analytics, intellectual sensing and
frameworks, and AI procedures to resolve the concern towards asset management
using demonstrating effective administration of accessible assets.
Besides, most recent advanced methodologies have additionally contributed
colossally in taking care of the issue of water deficiency and in the improvement of
irrigation system approaches, for example, improvement of a computerized frame-
work for agriculture, which advises farmer of specific vulnerabilities ahead of time
and supports, they control such bothersome circumstances [3, 4]. A central segment
of the PA framework is field planning. The idea of accuracy farming isn’t old; it was
induced before 25 years when inventive innovations, for example, remote sensing
networks and GPS were introduced [5].
PA is a creative strategy for gathering, scrutinizing, and highlighting information,
which assumes a vital part in settling on crop the executives’ choices [6]. Present-day
agriculture is as of now confronting another period with expanding creation needs
and lessening accessibility. The remarkable expansion in population carries with it
the need for a more prominent measure of creation; the capacity to control assets
precisely permits more noteworthy control of costs/benefits and assists with checking
the exhaustion of characteristic assets, for example, streams and soil supplements
[7]. PA permits observing of environment entities and to have an exact reaction on
visualization of plantation behavior to tackle explicit issues for each situation [8].
As demonstrated in Fig. 1, four primary parts of accurate agriculture cycle which
involve every one of the various procedures utilized are Sense-Analyze-Predict-Act.
Figure 1 represents the transformation of a cloud stage administration, is incorporated
by sensors, which thus forecasts the outcome using the assistance of various models
and ultimately advises the farmer which assists them with making vital moves and
further their innovation advanced farming. As portrayed in Table 1, four parts of the
accurate agribusiness cycle, characterize how the cycles are completed and beneath
we see what each parameter accomplishes.
A Review of Precision Agriculture Methodologies, Challenges … 331

Fig. 1 Precision agriculture life cycle

Table 1 Statistical display of articles according to their research fields

Major category Sub-category Sub-sub category No. of articles
Technology Hardware Type of sensors 31
NFC 7
Sensor networks 28
Software Middle-ware 24
Database 23
Architecture Hardware/Network 12
Software 9
Process 10
System 12
General 12
Application Smart infrastructure 46
Challenges Security challenges 8
Privacy challenges 6
Legal challenges 6
General challenges 18
Business models 7
Future directions 30
Overview/Survey 14

1.1 Precision Agriculture Policies and Performs

1.1.1 Irrigation System

Taking everything into account, innovative water system frameworks are
vital in the whole harvest managing process, which incorporates planning
332 M. Malek et al.

of seedbeds, usage of supplements, pesticides and insect sprays, growth,

harvest and development, and re-development of ranch. A profoundly exact
and flexible irrigation system framework is highlighting a significant part of
the PA procedure.
1.1.2 Quality Land Management System
It is the utilization of tasks, performs, and medicines to ensure soil and
improve its exhibition. It is significant, straightforwardly and by implication
both, to efficiency and natural manageability of crops [9] and the board assists
us with controlling significant agricultural parameters, i.e., vegetal well-
being, its shelter, and soil dampness that assists with giving a greater depic-
tion to mix of advancements and farming related sensors to reduce expenses
and increment harvests, for example, sensors for pH [10]. This sensor is able
to gauge the focus on intensities of oxygen, particle and hydrogen that will
help ranchers during anticipating land squabble circumstances [11].
1.1.3 Weather Anticipating System
The exact ground-level climate data support for surveying the useful-
ness of the rancher’s grounds and becoming further proficient in pest and
infectious prevention, by staying away from obliteration of harvests and
expanding the functionality and productivity in their activities [12, 13].
Expanding necessity towards IoT to interrelate every one of the parts of
both the virtual and real sphere has brought about organizations and buyers
to vigorously depend on IoT-related advances to create headways in their
individual arenas.
1.1.4 Data Insight
Expansion in proficiency and usefulness is the fundamental target of IoT-
initiated advances. Gathering adequate measures of information to utilize it
as an instrument to increase buyer involvement is the Holy Grail that most
organizations depend on. The assortment of sensor investigation to build
the information extension guarantees numerous advantages in cultivating,
chiefly decrease in delays because of outside correspondence, which at last
aides ranches arranged in country regions with helpless availability [14, 15].
1.1.5 Sustainable Computing
With the developing requirement aimed at sustainability of assets, the
objective of bigger rated associations is to make the entire world a superior
spot. In the wake of investigating numerous utilizations of IoT advances,
the Forum of World Economic tracked down that 84% were tending to,
or might actually discourse, the Sustainable Development Goals of UN.
Sustainable processing assists with vanquishing a portion of the world’s
developing obstacles, furthermore, handles those issues by giving arrange-
ments, like Smart Grid. A savvy network comprises of all the most recent
creating/created IoT manageability innovations. It identifies and responds to
nearby alterations in use using an electric inventory system which utilizes
computerized correspondences innovation.
1.1.6 Data Security
A Review of Precision Agriculture Methodologies, Challenges … 333

Data Security has a prime job in such turns of events. The extent of
information and abilities to incites, overall aides cultivate the fundamental
correspondence between each one of those worried about food creation and
the climate based on study by the ranchers and ground laborers in the overall
logic, wherever [16]. Playing out a portion of the information investigation
straightforwardly inside the ranch additionally has number of advantages
from the data security perspective. Just a synopsis of the homestead checking
information will be shared by outsider. This will complicate getting explicit
touchy data from the information shared.
1.1.7 Data Analytics
Data scrutinization helps with creating significant, preemptive, and bene-
ficial choices which can expand openings and effectiveness in the whole cycle
of smart cultivating. In spite information examination require relevant statis-
tics to be effective, and information which is inadequate or inaccurate will
give experiences that are not completely dissected. Data from sensors fixed
at filed, assortment of information at every stage, and financial elements of
choices will keep on being basic for achievement of information examination.
1.1.8 Cyber Physical Systems
Digital actual frameworks will shape an incorporated climate of digital
and actual items which helps ranchers progressively observing of home-
steads, soil management and crop-management by demonstrating assortment
of data.

2 Legacy Systems

2.1 Crop Monitoring System Using IoAT

The 3D organizes hub associated with the yield distinguishing proof component
addresses different IoAT frameworks. This framework manages the claim normal
for IoT and highlights the prospective it needs to increase the accuracy agribusiness
strategies. The mishap the anticipated model has the consideration of water, envi-
ronment and soil sensor in a similar class as yield recognizable proof, supplement
regulator and energy regulator.

2.2 Camera-Chessboard Array Preparation

The principle obstacle in such sort of plan is assisting to manage the overlaying
of IoT data using each single item on the perspectives on the camera. This design
course of action goes through a specific strategy to see if some forecast blunders
occur in the arrangement of cameras. The product, utilized to adjust such cameras, is
a prior MATLAB tool compartment program that supports in discovering the camera
334 M. Malek et al.

forecast blunders. Subsequent to handling its programs and making virtual square
outlines of their showcase and associations, a realistic of the vegetal using camera
perception is made, that assists the ranchers with for all intents and purposes looking
to their crops daily.

2.3 Rural Architectural Network

It is a WSN empowered organization intended to expand crop-management doings

to consider justified harvesting choices [17, 18]. The proposed architecture gives
versatility and covers huge topographical areas. It utilizes 6LoWPAN centered WSN
grid for detecting and inciting associated tasks.
Progressing upon past surveys for a multi-camera course of action on a smaller
ranch, in this research the measure increment and notice various sensors hubs dissi-
pated through an enormous farmland region. These hubs play out the detecting and
impelling tasks all the while, expanding the adaptability and the inclusion range.
The utilization of IPv6 for this arrangement has been fundamental. In addition to
the fact that it solves the location space and auto-setup entanglement, but on the
other hand is joined by the amendment level to present the 6LoWPAN convention
heap. Little parcel dimension, little data transfer capacity (amongst 250/40/20 kbps),
and other lattice geography downsides are just the beginning of the detriments of
the 6LoWPAN organization. Despite the fact that it has its own merits, for example,
energy redeemable overutilization of little fueled attending hubs, the key drawback
of 6LoWPAN is the absence of use that uses 6LoWPAN due to the broad preparation
which it needs to acquire acclimated with the innovation and the information on IPv6
convention. Information on heap and the usefulness of IPv6 is an absolute neces-
sity aimed at the finishing clients dealing with such design. Directing arrangements,
execution investigation, the structural intricacy of the proposed CLAD convention,
impacts of climate over the 6LoWPAN related organization and the confirmation
of nature of administration are for the most part completely gave in the led study
to really carry out haze and distributed computing innovations in their anticipated
system. In the new occasions, the utilization of applications related to IoAT has
expanded dramatically because of its low-energy prerequisites and inclusion huge
topographical locales. Somewhat recently, LoRaPAN and LoRaWAN advances have
supplanted customary correspondences conventions, for example, Wi-Fi and Classic
Bluetooth [19–22].
For the most part talking, in the haze processing models, IoAT based gadgets
are associated with heterogeneous gadgets in closeness to IoAT clients. The signif-
icant difficulties in current applications of IoAT in mist registering climate are the
executives of accessible assets and booking of routine assignments [18, 23].
A Review of Precision Agriculture Methodologies, Challenges … 335

2.4 Irrigation System Through Aerial Sight and Central Pivot

This system, despite the fact that is the nearest thing we need to amplify IoT execution,
ordinarily the last uses of these advances are not easy to understand. The comprehen-
sion of the huge potential outcomes of these innovations is a vital factor in finding and
applying them. Moderation of the information hole is vital to utilize the tested inno-
vation proficiently and suitably. With this, security issues, information misfortune,
and control above lengthier diffusions are additionally extraordinary concerns that
must be managed. This sort of system is exceptional in the constant dissemination of
the amount of water in the whole farmland region [24, 25]. Each proposed framework
enjoys its benefits and burdens, a couple of significant benefits and drawbacks that
accompany the execution of this framework are for watering bigger geological fields.
The proposed design requests just 60% of water when contrasted with the regular
water system frameworks with minimal expense and higher productivity.
It is basic that customary overhauling and upkeep ought to be given to such
frameworks to stay away from the chance of glitches, which requires colossal starting
capital venture [24, 26]. Once more, if the extraordinary issues of this proposed
model are relieved and vital advances, for example, consideration of sponsorships
to ranchers are taken to spur them to roll out progressive improvements to their
homestead and, that is the point at which the agriculture area will begin to develop
the affirmative manner. Further, we take a gander at keen smart water management,
which consolidates different advancements, with the objective to add the effort of
IoT innovations and reason any continuous issues with the current designs. Brilliant
water the board framework can give constant notice of the water system and yield
conditions to ranchers, which may help ranchers in taking in the nick of time choices.

2.5 Anticipated Methodology

This part examined an IoAT framework that has been anticipated in this study to give
an outline of an IoT initiated PA framework, which stages using an order of levels,
adjustable administrations, and administrations that can be reproduced for normal
IoT frameworks alike. As of small sensor establishment over the farm to definite
overall benefit of end clients, such design gives an outline or a plan that can be used
for comparable IoT frameworks.
I. Stage I: Physical Stage
An IoT actual stage gadget, for example, actuators, sensors, and regulators
communicate and robotizes the interaction of treatment, aquatic system and the
usage of insecticides and insect poisons for superior harvest development the
executives [27]. The IoAT framework consolidates a few customary advances,
for example, NFC, distributed computing, RFID, WSNs to computerize the
harvest the executives measures [28]. WMSN and WSN models transfer data
using remote networks comprising of numerous interconnected hubs that are
336 M. Malek et al.

utilized to detect, measure, and speak through IoT incorporated innovations,

which assumes a vital part in any IoT incited tests [29].
These designs are extremely exact as they follow a particular diagram of
the relative multitude of required segments, which must be incorporated for
effective signal transmission [30]. WMSN or WSN enjoy their benefits and
impediments and though these segments are the innovators of node processing,
expansion in proficiency, usability and consistency is consistently a worry
[31]. In addition, to comprehend the administration of anticipated IoT Data
prototypes, it is fundamental to dive further into rules independent entity-
management, which happens at a reasonable stage. WQM is a significant
obstacle in WSN related systems. Such frameworks request dependability,
manageability, and ideal nature of administration to convey WQM information
over longer distances [32–34].
When all is said in done, IoAT empowered WSN frameworks comprises
of various correspondence frameworks that are needed for directing, power
source reinforcement, actuators, and sensors for communicating data associ-
ated to crop, and easy to understand GUI-related plans [35]. The preparing
chunk comprises of a miniature regulator, which benefits, in legitimate and
number-crunching activities, whose outcomes are put away in the memory.
In addition, the result of the preparing chunk is utilized with the correspon-
dence chunk through an outsider application so the rancher can settle on better
choices relying on their space of attention. The sensor chunk encloses various
kinds of sensors, for example, pH sensors, moistness sensors, hotness sensors;
which guarantees that the information collected is solid and can be exam-
ined deliberately utilizing numerical calculations [36]. As a rule, the regular
IoAT framework incorporates three significant advancements: (a) actuators and
sensors (b) mist registering and edge figuring (c) distributed calculating stock-
piling. These advancements are able to give applicable data to GUI related
gadgets and it additionally intersects the diverse IoAT engineering [37].
II. Stage 2: Network Stage
This stage manages information assortment, the board, information protec-
tion and security, and IoAT device management [38, 39]. Information
obtaining, security, and the executives: Protocol and programming segments
for information procurement are the vital attributes of this stage further-
more to safety and device-management purposes. This stage uses commu-
nication agents like Mosquitoes. Mosquitoes are proficient to execute IoT
based conventions like MQTT. MQTT is a lightweight community IoT conven-
tion, which helps IoAT frameworks in intersecting the regulators, center prod-
ucts, switches, doors, workers, and detecting gadgets. MQTT is a generally
utilized convention for asset and energy requirement IoAT applications [40,
41]. In the existing times, technologies, for example, LoRaWAN has empow-
ered long-range remote correspondences around the scope of 25–30 km. It
needs low force and lesser transmission capacity for composite remote corre-
spondences. Irrespective, it is a troublesome errand for LoRaWAN to pledge
present multimedia coursing above inferior transmission capacity for uses, like
A Review of Precision Agriculture Methodologies, Challenges … 337

climate observing [42, 43]. LoRaWAN utilizes IPV6 convention to speak with
IoAT based frameworks. Moreover, somewhat recently, with the appearance
of various correspondence conventions IFTTT, Z-Wave, NFC, Thread, RFID,
SigFox, ZigBee, and interlinking of digital and actual articles have gotten
simpler. Thread and Z-Wave are brilliant correspondence convention which
is being utilized in shrewd houses that needs to interlink house machines,
for example, ACs, fans, freeze, lights, dishwasher, etc. [44, 45]. NFC is a
label-based innovation which is utilized for distinguishing different gadgets
and short-range interchanges. NFC is able to interrelate and regulate detecting
gadgets straightforwardly through advanced mobile phones [46]. ZigBee is
a two-way correspondence convention that is generally utilized for uses, for
example, tree directing, vector steering, and low-rate information interchanges
[47].
III. Stage 3: Data Management Stage
This stage manages information conveyance, handling, and capacity. It uses
software management firmware stage-like FIWARE. FIWARE is a stage for the
turn of events and worldwide sending of Internet utilization of the upcoming
time. It gives an absolutely free, public, and open architecture just as a bunch of
details which permit designers, specialist co-ops, organizations, and different
associations to foster items [48, 49]. FIWARE COSMOS is the orientation
execution of the FIWARE’s big data study general enabler. It is expected to
convey implies for investigating stream and batch information [50, 51].
FIWARE ORION and FIWARE CYGNUS is an apache flume related center
product that is liable for overseeing constant information stockpiling, which is
generally utilized for overseeing ancient data. It primarily manages information
assortment, information handling, and setting mindful data trade [48]. FIWARE
ORION is a C++ execution of FIWARE stage that is competent to deal with the
whole lifetime pattern of information the board. It incorporates capacities, for
example, overseeing questions, routine updates and memberships. FIWARE
QUANTUM LEAP is a component that can accumulate information as time-
arrangement, for example, ngsi-tsdb [48].
IV. Stage 4: Irrigation and Water Distribution Methodologies
Ordinary farming plans are required to utilize robots and soil sensors to
get logical data identified with soil conditions. The traditional models likewise
use calculation insight methods for quality of water administration and cloud
computing investigation for wellbeing or execution of output based uses [19].
These methodologies use various logical techniques and instruments to give
constant relevant data to farming specialists [52, 53].
(a) Edge Computing: Edge figuring is a productive information securing
method, which is generally utilized in mechanization related IoAT appli-
cations to give ongoing scientific calculations and notices identified with
soil conditions, natural conditions, and landscape-related data [43, 45,
54]. It is a portable detecting system which gives constant checking
338 M. Malek et al.

of distant areas which will help rural specialists’ in settling on farm-

land related choices [55]. Somewhat recently, individual analysts have
proposed different approaches to carry out edge and fog registering
created IoAT frameworks [56].
(b) Irrigation Models: A water system mechanism (drip irrigation model)
comprises various water systems (drip irrigation) techniques which are
utilized by agricultural experts in the existing eras. The five strategies
to the water system methodology are (i) exterior water system (ii) sprin-
kling water system (iii) drop water system (iv) sub-surface water system
and (v) water system using drone. On account of surface water system,
the water is consistently appropriated above the dirt apparent because
of the impact of gravitational stream. Sprinkling water system looks
like a rain style activity where water is splashed around the ranch field.
Trickle irrigation is a methodology wherein water is conveyed as drops
or little streams. Smart trickle water system frameworks are boon to
execute for ranchers and cause exceptionally less initial expense. In
such frameworks, because of mechanization, no human mediation is
required which likewise brings about moderation of water wastage [57–
59]. Ranchers can oversee and control such frameworks utilizing smart-
phones. On account of subsurface water system method, the water is
provided beneath the layer of soil. The water is generally provided
inside the zone of plant root. Drone technology helped water utilizing
IoAT helped innovations, for example, WSNs [60–62]. Agribusiness
explicit drone in keeping an elevated perspective on crop creation and
assists screen with trimming development. Improving the general ranch
proficiency, the robots likewise give an arrangement where ranchers can
screen their yields occasionally, as indicated by their preferring. The
pesticides are the principle obstacle drones can intercept [63, 64]. Bugs
are identified, and various spaces of farm are gotten in practically no time.
The lone concern drone helped water system makes is that of protection.
There is no authority required for drones entrance, passing over some-
body’s possessions using a camera and mic appended to it, is said as
simple as possible, and could be a possible protection infringement.
V. Stage 5: Water Use Facilities
Easy-to-use GUI-related plans are given to ranchers and merchants to
improved agreement and quicker admittance to water use administrations.
Next, the pecking order of IoT in farming is trailed, the last year carries
the appropriation of such developed administrations, just as how nicely the
anticipated framework encounters client prerequisite and satisfies designer
and client assumptions [65]. The engineers utilize the updated information
assembled through this trial to additionally progress their IoT environment
segments, and the clients (ranchers) who are actually occupied by such inno-
vation on an everyday premise, advantage from the use of these administrations
and help give significant client experience information towards the executed
A Review of Precision Agriculture Methodologies, Challenges … 339

test. Because of energy imperatives of water quality administration detecting

gadgets a solid energy reaping philosophy is needed to deal with the force rein-
forcement prerequisites. Be that as it may, the productive execution of energy
gathering technique is as yet an open issue for individual analysts [32]. A
productive steering convention is needed to oversee smooth directing of data
between detecting gadgets, regulators, switches, entryways, and distributed
computing based information stockpiling workers [66]. On the new occasions,
it has been seen that the land is contracting slowly which presses the regular
assets. In such situations, most recent advances can help ranchers in getting
context-oriented data of assortment of yields for in the nick of time crop the
board choices [66].
Besides, the factual portrayal of all explored information has appeared in
the additional areas, which provides understanding of the definite directed
overview of most recent IoAT related reviews [67, 68]. In the wake of assessing
altogether the handling stages and determining how an IoT framework should
work for explicit assignments, investigation of the inborn association among
IoAT-related frameworks and information is addressed [69]. In the decision-
making process, investigation is the utmost significant advance. All the infor-
mation created through IoT gadgets is just of worth on the off chance that it
is exposed to examination. Unlawful perceptions and significant ends are the
things which IoAT frameworks plan to remove while carrying out an informa-
tion examination (DA) measure. As demonstrated in Table 1, this paper has
done the conveyance and characterization of the IoAT articles as for various
classifications, which are explains subsequently. It additionally helps with
recognizing the difficulties which undermine IoT scattering and open explo-
ration issues which steers innovation the correct way for coming eras [70,
71].
(a) Technology: IoT is a network, which is implanted with innovation and
can gather and trade information [71]. Like an environment, it connects
different frameworks with the shared factor of IoT installed innovations
and assists us with broadening IoT advancements farther than our PCs
and cell phones.
(b) Applications: Uses of this layer in such engineering have a tremendous
window for development. IoT gadgets are probably going to be valuable
in association amongst individuals and the gadgets nearby them [72].
Based on following everyday exercises of a person, to observing their
rest for more proficient and sound living, the utilizations of IoT are
boundless and consistently have an extent of additional development
[73, 74].
(c) Challenges: The difficulties in the field of IoT are largely scattered.
To effectively carry out IoT implementations, these difficulties must be
managed and insignificant or totally removed [75, 76]. From security,
protection, and legitimate difficulties to challenges, which overall influ-
ence each of these viewpoints, are to be searched from this engineering.
340 M. Malek et al.

Problem related to encryption, communication compatibility, identity

management, universal liability problems, and data possession are the
significant challenges which are to be engaged in.
(d) Business Models: Through modifications in IoT advances to expand
usefulness and effectiveness in everyday life, plans of action likewise
must be adjusted to match IoT assumptions [77]. For ecological devel-
opment and upcoming improvements in the arena of accuracy farming,
rural-centered enterprises are needed to distinguish feasible innovation-
related agricultural plans that can give ongoing checking of harvests and
give most recent soil conditions. The directed review distinguishes the
deficiencies between the traditional horticulture based frameworks and
developments received by farming based ventures [78].
(e) Future Directions: Despite the fact that anticipating the eventual fate
of farming-related architecture is disputable, the idea of this headway is
the main thrust for the development of the Web of Things [79, 80]. For
coordination of clever items with this present reality, the simplicity of
the cycle of advancement, expansion in combination among gadgets and
defeating difficulties to build the attainability of thoughts are the great
components, which must be encountered through.
(f) Survey/Overview: Here, comprehensive definite study of IoAT tech-
nologies are discussed and presented in Table 1.
Table 1 depicts statistical data distribution about different categories and sub-
categories with respect to the surveyed articles and discussions. Figure 2 signifies a
graphical presentation of the quantity of papers considered for journal and conference
papers. Similarly, Fig. 3 illustrates the dispersal of all the various sources based on
which the study and analysis have been incorporated.
This statistics demonstration contributes various insights addicted to the prevailing
investigation which headways and its records were used for this study and all

35 32
30

25

20
14
15

10

0
Journal paper Conference paper

Fig. 2 Type of paper reviewed for the study

A Review of Precision Agriculture Methodologies, Challenges … 341

Fig. 3 Pie chart ICCCI Big Data &

Springer MICC MIPRO
representation of different 2% 2% 2% Society
publications 2%
2%
Indepedentent
4%
ICGT SPICC
4%
John Wiley & IEEE Journal
Sons 39%
7%
MDPI
(Sustainability
Journal) MDPI (Sensor
5% Journal)
13% IEEE
Conference
18%

the synchronized reference replicas based on which this study has been extracted.
Figure 4 elaborates a global depiction of the previous studies of the similar research
works.

Fig. 4 Global scenario for similar research work

342 M. Malek et al.

3 Conclusions and Future Enhancement

Imperative agribusiness initiatives affect the advanced existing practices, methods,

and applications. The most recent exploration discoveries additionally recommend
that suitable utilization of accuracy based practices can make an incredible effect on
the efficiency grade and sustainability factors [81–83]. This research paper expects to
examine the various innovation-based water system frameworks (irrigation systems),
ongoing modern practices and developments, different accuracy agribusiness-based
stages alongside the conversation of a recently presented methodology. This paper
discussed various contextual investigations that were likewise talked about to evaluate
the adequacy of the inheritance frameworks[84–92].
In this domain, maximum research were focused on regulator of several agricul-
tural crops and the remote monitoring, necessities of its resource, like required level
of water, pesticides and insecticides quantity, and fertilizers usage to ascertain the
indices of vegetation. Notwithstanding, this paper prepared a certified endeavor to
dissect the current techniques as well as have proposed a viable model to accomplish
better outcomes.
Though, the paper has finished a sincere effort for analyzing the prevailing
approaches and also suggested an efficient mechanism to attain superior outputs.
In future, a secure precision agriculture framework can be implemented to address
IoT-related security challenges.

Conflicts of Interest The authors declare no conflicts of interest.

References

1. Shah B, Malek S (2018) Identification of significant factors affecting the design of smart village.
In: Proceedings of the 2nd international conference on women in science and technology:
creating sustainable career, Birla Vishvakarma Mahavidyalaya, Vallabh Vidyanagar, India
2. Nandurkar S, Thool V, Thool R (2014) Design and development of precision agriculture system
using wireless sensor network. In: International conference on automation, control, energy and
systems (ACES), Hooghly
3. Ayanelam YF, Malek S (2018) A critical review of effective project management using 5D
BIM. In: Smart construction. 1st International conference on engineering and technology:
smart engineering, Parul University, Baroda, India. McGraw Hill Education, pp 155–158
4. Kochai Z, Malek S (2018) An assessment of time and cost overrun in Afghanistan’s construc-
tion. In: Smart construction. 1st International conference on engineering and technology: smart
engineering, Parul University, Baroda, India. McGraw Hill Education, pp 159–166
5. Chaudhari K, Malek S (2018) Critical review on delay and cost overrun in construction projects.
In: Smart construction. 1st International conference on engineering and technology: smart
engineering, Parul University, Baroda, India. McGraw Hill Education, pp 167–172
6. Fountas S, Aggelopoulou K, Gemtos T (2016) Precision agriculture: crop management for
improved productivity and reduced environmental impact or improved sustainability. In: Supply
chain management for sustainable food networks
7. Zhang L, Dabipi I, Brown W (2018) Internet of things applications for agriculture. Wiley
A Review of Precision Agriculture Methodologies, Challenges … 343

8. Ghayvat H, Mukhopadhyay S, Gui X, Suryadevara N (2015) WSN-and IoT-based smart homes

and their extension to smart buildings. Sensors 15:10350–10379
9. Ananthi N, Divya J, Divya M, Janani V (2017) IoT based smart soil monitoring system for
agricultural production. In: IEEE
10. Malek M, Mohibali S, Bachwani D (2021) Identification, evaluation, and allotment of critical
risk factors (CRFs) in real estate projects: India as a case study. J Proj Manag 6(2):83–92.
https://doi.org/10.5267/j.jpm.2021.1.002
11. Malek MS, Gundaliya PJ (2020) Negative factors in implementing public–private partnership
in Indian road projects. Int J Constr Manag. https://doi.org/10.1080/15623599.2020.1857672
12. Malek M, Gundaliya P (2021) Value for money factors in Indian public-private partnership
road projects: an exploratory approach. J Proj Manag 6(1):23–32. https://doi.org/10.5267/j.
jpm.2020.10.002
13. Malek M, Bhatt V, Patel A (2020) Global, local and national growth of road projects through
PPP. Test Eng Manag 83(March–April 2020):25837–25860
14. Memon A, Malek M, Patel A (2020) A review on risk and risk management technique in
construction projects. Int J Disaster Recovery Bus Continuity 11(1):2695–2701
15. Patel R, Bachwani D, Malek M (2020) Human resource management, a vital tool for labour
management in construction industry. Stud Indian Place Names 40(9):201–206
16. Parikh R, Bachwani D, Malek M (2020) An analysis of earn value management and
sustainability in project management in construction industry. Stud Indian Place Names
40(9):195–200
17. Khan S, Malek M, Phugat S (2019) Factors influencing time and cost overrun in construc-
tion of government buildings: Indore as a case study. CIKITUSI J Multi Res 6(4):288–295.
16.10089.CJMR.2019.V6I4.19.2857
18. Nautiyal V, Shah Y, Malek S (2019) Application of pavement management
system in HDM-4: a critical review. Int J Manag Technol Eng 9(4):1850–1859.
16.10089.IJMTE.2019.V9I4.19.27237
19. Macwan J, Bhavsar A, Malek M (2019) Smart project scheduling for construction projects
using AI. Int J Tech Innov Mod Eng Sci 5(5):14–19
20. Agrawal A, Malek M (2018) Principles of sustainability in construction project management.
Int J Adv Innov Res 5(4):45–52
21. Gajera D, Malek M (2018) A study on consumer behaviour in real estate for Vadodara city.
Univers Rev 7(12):956–969
22. Kumbhani H, Malek M (2018) Cost management process in construction industry. Suraj Punj
J Multi Res 8(12):151–158. 16.10089.SPJMR.2018.V8I11.16.3145
23. Saiyed F, Malek S (2017) The realty growth trajectory of India and China—inexorable
comparison. In: Proceedings of the 2nd international conference on construction, real estate,
infrastructure and project management, NICMAR, Pune, India
24. Upadhyay U, Malek M (2018) Identification of factors responsible for organizational
transformation of GSRDC. Pramana Res J 8(12):189–201. 16.10089.PRJ.2018.V8I12.18.2063
25. Malek S, Akalkotkar P (2014) Need of PPP in Indian road sector. Int J Emerg Technol Appl
Eng Technol Sci. In: Proceedings of the international conference on information, knowl-
edge & research in engineering management and sciences (IC-IKR-EMS), Kalol, Gujarat,
India. KITRC, pp 300–305
26. Saiyed F, Pitroda J, Malek S (2016) Assessment of very critical risk factors in real estate
sector by criticality index. In: Proceedings of the international conference on engineering:
issues, opportunities and challenges for development, Umrakh, Gujarat, India. SNPITRC, pp
193–205
27. Jariwala K, Malek M (2018) Total quality management in construction of textile market in
Surat city. J Appl Sci Comput 5(12):699–706. 16.10089.JASC.2018.V5I12.453459.1500186
28. Bachwani D, Malek M (2018) Parameters indicating the significance of investment options in
real estate sector. Int J Sci Res Rev 7(12):301–311. 16.10089.IJSRR.2018.V7I12.287391.2950
29. Rathod S, Malek S (2018) Six sigma in construction industry: literature review. Int J Manag
Technol Eng 8(12):913–920. 16.10089.IJMTE.2018.V8I12.17.2102
344 M. Malek et al.

30. Pandya SP, Prajapati MR, Thakar KP, Assessment of training needs of farm women. Gujarat J
Ext Educ 25(2):169–171
31. Pandya S, Ghayvat H, Sur A, Awais M, Kotecha K, Saxena S, Jassal N, Pingale G (2020)
Pollution weather prediction system: smart outdoor pollution monitoring and prediction for
healthy breathing and living. Sensors 20:5448
32. Kochai Z, Malek M (2018) Causes of schedule delay and cost overrun in Afghanistan
infrastructure projects. Int J Adv Innov Res 5(3):18–25
33. Malek S, Akalkotkar P (2016) Factors contributing to successful public private partnership
projects for highway projects in Indian context. Int J Sci Res Dev 3(12):25–29
34. Pandya S, Ghayvat H, Kotecha K, Awais M, Akbarzadeh S, Gope P, Smart home anti-theft
system: a novel approach for near real-time monitoring and smart home security for wellness
protocol. Appl Syst Innov 1(4):42
35. Araya SY, Malek S (2018) Application of building information modelling in Indian construction
projects—a critical review. Int J Creative Res Thoughts 6(1):979–987
36. Tadewos SG, Malek S, Patel D (2018) Challenges of building construction in Mekelle city,
Ethiopia: from contractor’s perspective. J Emerg Technol Innov Res 5(3):420–424. https://doi.
org/10.6084/m9.jetir.JETIR1803080
37. Aynalem YF, Malek S (2018) Developing an effective project concept in 5D environment. J
Emerg Technol Innov Res 5(5):372–380. https://doi.org/10.6084/m9.jetir.JETIR1805660
38. Barot V, Kapadia V, Pandya S (2020) QoS enabled IoT based low cost air quality monitoring
system with power consumption optimization. Cybern Inf Technol
39. Ghayvat H, Pandya S, Patel A (2019) Proposal and preliminary fall-related activities recognition
in indoor environment. In: 2019 IEEE 19th International conference on
40. Pandya A, Malek S (2018) A review for the causes of delay in Indian construction projects. Int
J Adv Eng Res Dev 5(3):1–6
41. Akbarzadeh S, Ren, Pandya S, Chouhan A, Awais M (2019) Smart aging system
42. Shah B, Malek S (2018) A recapitulation on exigency of smart villages in Indian ambience.
Int J Adv Eng Res Dev 5(3):1–8
43. Saiyed F, Pitroda J, Malek S (2015) Reaping the returns: review on risk management in real
estate. Int J Futuristic Trends Eng Technol 3(1):1–8
44. Patel V, Malek S (2018) A review on identification of safety factors in construction projects.
Int J Adv Eng Res Dev 5(3):1–9
45. Malek S, Pathan N, Mal H (2013) Risk management in construction industry. Indian J Appl
Res 3(6):377–379
46. Parekh P, Malek S (2018) A critical literature review on implementing lean construction
concepts in infrastructure projects. Int J Adv Eng Res Dev 5(3):1–7
47. Chaudhari K, Malek S (2018) Effectiveness of public private partnership in infrastructure
development in India: a review. Int J Adv Eng Res Dev 5(3):1–12
48. Hada N, Malek S (2018) Identification of risk factors in government housing project: a review.
Int J Adv Eng Res Dev 5(3):1–6
49. Ghayvat H, Pandya S (2018) Wellness sensor network for modeling activity of daily livings—
proposal and off-line preliminary analysis. In: 2018 4th International conference on computing
50. Jethwa B, Bhavsar A, Malek S (2017) An overview of factors affecting risk management in
PPP based infrastructure projects. Int J Eng Dev Res 5(2):1737–1742
51. Awais M, Kotecha K, Akbarzadeh S, Pandya S (2018) Smart home anti-theft system
52. Jethwa B, Bhavsar A, Malek S (2017) Critical review on risk management in PPP based
infrastructure projects. Int J Eng Dev Res 5(1):357–361
53. Patel M, Pandya S, Patel S (2017) Hand gesture based home control device using IoT. Int J
Adv Res
54. Parmar D, Pitroda J, Malek S (2017) Use of BIM (building information modelling) as an
integrated tool to plan, design and manage critical construction projects. Int J Adv Res Innov
Ideas Educ 3(1):1145–1153
55. Malek S, Saiyed F (2017) A study on incipient trends: the real estate sector of Ahmedabad city
(Gujarat). Int J Res Granthaalayah 5(1):304–315
A Review of Precision Agriculture Methodologies, Challenges … 345

56. Patel D, Malek S (2016) Application of performance management system in infrastructure

sector of construction industry. Int J Eng Dev Res 4(2):1998–2000
57. Kalkani K, Malek S (2016) Analyzing causes of delay in construction projects. Int J Innov Res
Sci Technol 2(12):257–262
58. Pandya S, Yadav AK, Dalsaniya N, Mandir V, Conceptual study of agile software development
59. Malek S, Akalkotkar P, Zala L (2016) Gujarat PPP: attractive factors for adopting public private
partnership for highway projects. Int J Futuristic Trends Eng Technol 3(2):1–6
60. Garg D, Goel P, Pandya S, Ganatra A, Kotecha K (2002) A deep learning approach for face
detection using YOLO. In: 2018 IEEE PuneCon
61. Sur A, Sah RP, Pandya S (2020) Milk storage system for remote areas using solar thermal
energy and adsorption cooling. Mater Today Proc
62. Cohen JM, Pandya S, Krasenbaum LJ, Thompson SF (2020) A real-world perspective of
patients with episodic migraine or chronic migraine prescribed AJOVY in the United States.
HEADACHE
63. Sur A, Pandya S, Sah RP, Kotecha K, Narkhede S (2020) Influence of bed temperature on
performance of silica gel/methanol adsorption refrigeration system at adsorption equilibrium.
Part Sci Technol
64. Cohen JM, Pandya S, Tangirala K, Krasenbaum LJ (2020) Treatment patterns and character-
istics of patients prescribed AJOVY, Emgality, or Aimovig. HEADACHE
65. Wandra K, Pandya S (2014) Centralized timestamp based approach for wireless sensor
networks. Int J Comput Appl
66. Upadhyay S, Malek S (2016) Adoption and implementation of total quality management in
construction industry. Int J Eng Dev Res 4(2):1256–1260
67. Patel WD, Pandya S, Koyuncu B, Ramani B, Bhaskar S (2019) NXTGeUH: LoRaWAN based
NEXT generation ubiquitous healthcare system for vital signs monitoring & falls detection.
In: 2018 IEEE PuneCon
68. Dandvate HS, Pandya S (2016) New approach for frequent item set generation based on Mirabit
hashing algorithm. In: 2016 International conference on inventive
69. Swarndeep SJ, Pandya S (2016) Implementation of extended K-Medoids algorithm to increase
efficiency and scalability using large datasets. Int J Comput Appl
70. Pandya S, Shah S, Joshi N, Ghayvat H, Mukhopadhyay SC, Yap MH (2016) A novel hybrid
based recommendation system based on clustering and association mining. In: 2016 10th
International conference on sensing technology (ICST)
71. Patel S, Singh N, Pandya S (2017) IoT based smart hospital for secure healthcare system. Int
J Recent Innov Trends Comput Commun 5
72. Ghayvat H, Pandya S, Shah S, Mukhopadhyay SC, Yap MH, Wandra KH (2016) Advanced
AODV approach for efficient detection and mitigation of wormhole attack in MANET. In: 2016
10th International conference on sensing technology (ICST)
73. Swarndeep Saket J, Pandya S, Implementation of extended K-Medoids algorithms to increase
efficiency and scalability using large dataset. Int J Comput Appl
74. Bhola YO, Socha BN, Pandya SB, Dubey RP, Patel MK (2019) Molecular structure, DFT
studies, Hirshfeld surface analysis, energy frameworks, and molecular docking studies of
novel (E)-1-(4-chlorophenyl)-5-methyl-N -((3-methyl-5-phenoxy-1-phenyl-1H-pyrazol-4-yl)
methylene)-1H-1, 2, 3-triazole-4-carbohydrazide. Mol Cryst Liq Cryst
75. Patel W, Pandya S, Mistry V (2016) i-MsRTRM: developing an IoT based intelligent Medi-
care system for real-time remote health monitoring. In: 2016 8th International conference on
computational
76. Wandra KH, Pandya S (2012) A survey on various issues in wireless sensor networks. Int J Sci
Eng
77. Shah JM, Kotecha K, Pandya S, Choksi DB, Joshi N (2017) Load balancing in cloud computing:
Methodological survey on different types of algorithm. In: 2017 International conference on
trends in electronics and informatics (ICEI). https://doi.org/10.1109/ICOEI.2017.8300865
78. Malek S, Akalkotkar P (2016) Driving forces leading to the adoption of PPP: perspective from
Gujarat (India), Hong Kong and Australian practitioners. Int J Innov Res Sci Technol 2(9):6–11
346 M. Malek et al.

79. Patel RR, Pandya SP, Patel PK, Characterization of farming system in North West agro climatic
zone of Gujarat state. Gujarat J Ext Educ 27(2):206–208
80. Pandya S, Ghayvat H, Kotecha K, Yep MH, Gope P (2018) Smart home anti-theft system:
a novel approach for near real-time monitoring. smart home security and large video data
handling for wellness protocol
81. Malek S, Akalkotkar P (2016) Negative factors adopting public private partnership for highway
projects: a case study of Gujarat. Int J Sci Technol Eng 2(8):30–35
82. Ghayvat H, Awais M, Pandya S, Ren H, Akbarzadeh S, Chandra Mukhopadhyay S, Chen
C, Gope P, Chouhan A, Chen W (2019) Smart aging system: uncovering the hidden well-
ness parameter for well-being monitoring and anomaly detection, smart aging system: uncov-
ering the hidden wellness parameter for well-being monitoring and anomaly detection. Sensors
19:766
83. Saket S, Pandya S (2016) An overview of partitioning algorithms in clustering techniques
84. Andrew R, Malekian R, Bogatinoska D (2018) IoT solutions for precision agriculture. In:
MIPRO, Opatija
85. Benyezza H, Bouhedda M (2018) Smart irrigation system based ThingSpeak and Arduino. In:
International conference on applied smart systems. ICASS, Médéa
86. Patil K, Kale N (2016) A model for smart agriculture using IoT. In: International conference
on global trends in signal processing, information computing and communication. IEEE
87. Ashwini BV (2018) A study on smart irrigation system using IoT for surveillance of crop-field.
Int J Eng Technol 7:370–373
88. Shah K, Malek M, Phugat S (2020) Alternate building materials used in construction industry.
Stud Indian Place Names 40(9):184–189
89. Khan BR, Malek M, Tidke A (2020) Risk management in construction industry. Stud Indian
Place Names 40(9):165–171
90. Patel K, Malek S (2016) Planning, scheduling and resource management for residential real
estate projects in Ahmedabad by using Microsoft Project. Int J Eng Dev Res 4(2):189–194
91. Pandya S, Sur A, Kotecha K (2020) Smart epidemic tunnel—IoT based sensor-fusion assistive
technology for COVID19 disinfection. Emerald
92. Joshi N, Kotecha K, Choksi DB, Pandya S (2018) Implementation of novel load balancing
technique in cloud computing environment. In: Computer communication and informatics
(ICCCI)
Power Quality Enhancement and Low
Voltage Ride Through Capability
in Hybrid Grid Interconnected System
by Using D-Fact Devices

T. Joshiram and S. V. R. Laskhmi Kumari

Abstract One of the major difficulties detected in the distribution system in present
days has been power quality. These days, most individuals are utilizing the urbane
electrical devices which are dependent on the semiconductor devices, these devices
humiliate the power quality. Hence, there is a need to recover the voltage profile.
In this paper, the photovoltaic (PV) plant and the wind turbine generator (WTG)
are connected to the same point of common coupling (PCC) with a nonlinear load.
The unified power quality conditioner (UPQC) [1] is familiar as the best solution
for moderation of voltage sag associated problems in the highly taped distribution
system. This effort grants the simulation modeling and analysis of innovative UPQC
system for solving these problems. UPQC to increase the power quality and recover
the low voltage ride through (LVRT) capability of a three-phase medium voltage
network connected to a hybrid distribution generation (DG) system. UPQC associated
to same PCC. Unlike fault condition setups are tested for improving the efficiency and
the quality of the power supply and compliance with the requirements of the LVRT
grid code. It inserts voltage in the distribution line to reserve the voltage profile and
guarantees constant load voltage. The simulations were led in MATLAB/Simulink to
show the UPQC-based future approach’s usefulness to smooth the distorted voltage
due to harmonics [2].

Keywords Active power · DC link voltage DFIG · Unified power quality

conditioner · LVRT · Power factor · Photovoltaic voltage stability · Reactive
power · Reactive components · Total harmonic distortion · Sag · Swell · D-FACTS

1 Introduction

Power Quality—The Problem

If possible, the goal of power industry is to supply a purely sinusoidal voltage at fixed
amplitude and fixed frequency [1, 2]. Whereas it is the responsibility of the dealer to
deliver a nearly sinusoidal voltage with smaller amount variation in amplitude and

T. Joshiram · S. V. R. Laskhmi Kumari (B)

VR Siddhartha Engineering College, Kanuru, Vijayawada, Andhra Pradesh 520007, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 347
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_26
348 T. Joshiram and S. V. R. Laskhmi Kumari

frequency, the customer additionally has a section to play in making such a situation
[3]. At the PCC, both the service organization and the client have a few components to
conform to. To overcome these constraints and assurance the stability of the electric
power system related with a great deal of variable energy resources, the executives
of energy system have relied upon examining distinctive particular courses of action
[4]. Quite possibly the main grid code necessities are the low voltage ride through
(LVRT) capacity, which means that the renewable energy power change system
should remain associated during grid faults and supply reactive power to help the
grid [5].
The prospects of power generation from hybrid energy systems are winding up
being incredibly promising and dependable [2]. A DFIG and flywheel energy storage
system was premeditated in [3] and the future control system was planned to ensure
that the grid power is remote from wind power output fluctuations. To boost the LVRT
capability of a grid-integrated DFIG-based wind farm [3]. Additional devices usually
used in distribution networks to defend grave loads counter to voltage disturbances are
known as D-FACTS and comprise: DSTATCOM (Static Dispensing Compensator),
dynamic voltage restorer (DVR) and unified power quality conditioner (UPQC) [1].
Unified power quality conditioner (UPQC) is one of the most prevalent solutions cast-
offs currently [5]. The UPQC as isolated device in as quick control device to control
the active power and reactive power and it keeps the sensitive load from preeminent
disturbances [5]. The UPQC incorporates of joined activity of (DSTATCOM) and
dynamic voltage restorer (DVR) [1]. The DSTATCOM is compensating the reactive
power and harmonics in the load side and DVR mitigates the voltage sags/swell in
the source side [5]. From this future method the foremost disturbances are condensed
and too the control voltage sag instigating from the supply side [5]. To reimburse
for the harmonics in the load current by injecting the required harmonic currents.
To normalize the power factor by injecting the required reactive current [6]. This
paper dowries a simulation study to regulate the worth of the UPQC to restrained
voltage disturbances, reduce their effect on the total stability of the transmission and
distribution network, and recover the LVRT capability when the network is connected
to a hybrid PV-wind system [7].

2 Future Simulated Methods and Explanation

The ideal of photovoltaic energy or wind turbine or together relies upon the avail-
ability of the sustainable asset after some time and furthermore is bounty at the spot
of establishment [4]. The future topography is displayed in Fig. 1. It contains of a
500 kW PV farm interconnected to a distribution system through a 3 phase PWM
inverter with a 3 phase AC choke filter and a move forward transformer. The DFIG
has a negligible produce power of 500 kW and is associated with the matrix at the
PCC through a move step-up transformer and providing the heap. Therefore, the
appraised total power delivered by the hybrid system is 1 MW [2] (Fig. 2).
Power Quality Enhancement and Low Voltage Ride Through … 349

Fig. 1 MATLAB simulation with UPQC and a load connected to grid

Fig. 2 PV farm of 500 kW connected to grid via an inverter associated to 500 kVA − 400 V/30 kV
transformer [2]

The PV farm is exposed to the solar irradiance shown in Fig. 3 and the WTG is
consecutively with a wind speed of 12 m/s throughout the simulation time of 5 s
additional information required by the volume editor [1].

Fig. 3 Solar irradiance at 25 °C [5]

350 T. Joshiram and S. V. R. Laskhmi Kumari

2.1 Photovoltaic Installations

The PV plant array contains of 16 series modules and 102 parallel strings (model:
SunPower SPR-305-WHT) [2].
The PV model castoff in the paper is created on the 2-diode equivalent circuit
revealed in Fig. 4. The PV cell total current in the equivalent circuit revealed in
Fig. 4 is stated by [2].
 z+I R   y+I R 
v+I R
I = IPh − IPh e q nkt
− 1 − Is1 e q nkt
−1 (1)
RP

where n is the idealist factor. Presumptuous that all the cells are equal and working
under the similar operating circumstances [8].

Gs
Rp.field = Rp.cell (2)
Gp

The PV power transformation control is made on maximum power point tracker

(MPPT) which endures high efficiency power transfer and be dependent upon
together the solar irradiance and the electrical qualities of the load [9]. Figure 5
displays the I–V and P–V characteristics of the PV module for different stages of
solar irradiance [5].
As depicted in Fig. 5, the most extreme force is positive by the significant square
shape region, PMP = VMPIMP, reachable from the I–V trademark. The synchronizes
of the VMP are start [8].
   
dP  d(I E)  dI 
= = 1+ =0 (3)
dE  E=Emp dE  E=Emp dE  E=Emp

Then IMP is resolute by evaluating Eq. 1 at E = E MP [10].

The PV array model is mounted to contain 12 series modules and 102 equal strings
sought after to ship 500 kW at an irradiance of 1000 W/m2 and a DC voltage of VDC

Fig. 4 PV cell circuit model [2]

Power Quality Enhancement and Low Voltage Ride Through … 351

Fig. 5 V –I and P–V characteristic of PV module under dissimilar solar irradiance and at 25 °C
with MPPT points in strings [5]

= 896 V applied to the boost converter as uncovered in Fig. 7. There are various
MPPT control techniques future underway, some of them are a lot of the same in
relations of their functioning rule [9] (Fig. 6).
The INC strategy can be seen as a superior kind of the Perturb and Observe (P
and O) [11]. The point of the power bend is gotten from:

dPPV dIPV IPV

=0⇒ =− (4)
dE PV dE PV E PV
dPPV d(IPV E PV ) dIPV
= = IPV + E PV (5)
dE PV dE PV dE PV

Multiplying 2 sides of Eq. (5) by 1/E PV gives:

Fig. 6 Flowchart of the incremental conductance algorithm [2]

352 T. Joshiram and S. V. R. Laskhmi Kumari

Fig. 7 Displays the circuit model of the DC–DC boost converter used in this effort [7]

1 dPPV IPV dIPV

= + = G PV + dG PV (6)
E PV dE PV E PV dE PV

where G and dG mean the conductance and gradual conductance correspondingly. A

flowchart identifying with the INC calculation is uncovered in Fig. 6. The calculation
can follow the MPP and remains there till an adjustment of [dI]_(PV) or [dV ]_PV
occurs because of a modification in climatic conditions [11].
The grave upsides of the inductor and capacitances of the expected lift converter
are: L 1 = 30 mH, C = 100 µF, C 1 = C 2 = 120 mF. To affirm greatest force reflection
from the PV source, the converter interfacing the PV framework to the network
should be cultivated of self-changing its own limitations progressively. The 3-level
voltage source inverter topology revealed in Fig. 8 is simulated in this work [9].
The Vdc boost converter orientation output voltage is customary at 714 V and the
IGBT 3-level inverter uses PWM technique, (3.3 kHz carrier frequency) converting
DC power from 714 Vdc source to 400 Vac, 50 Hz. The grid is connected to the
inverter through an inductive grid filter and a low frequency transformer to step-up
the voltage from 0.4 to 3 0 kV in order to reduce losses when PV energy is transmitted
to the grid [9] and to filter out harmonic frequencies. The 12 pulses required by the
inverter are generated by the discrete three phase PWM generator [14] (Fig. 9).

Fig. 8 3-level inverter topology [25]

Power Quality Enhancement and Low Voltage Ride Through … 353

Fig. 9 Output three-level inverter unfiltered and filtered voltage waveforms [5]

2.2 Modeling of Wind Plant

1. Aerodynamic modeling of the wind turbine

The unique energy from the wind is caught by the wind turbine and changed over
to mechanical force Pm [9, 12]. The wind power plant involves of a solitary DFIG-
based wind turbine delivering 500 kW with 400 Vac produce voltage. The energy or
power of a wind turbine might be indefatigable by various means [27]. The force Pm
captured by the wind turbine is a control of the sharp edge sweep, the pitch point,
and the rotor speed [7] (Fig. 10).
The power or torque of a wind turbine may be resolute by numerous income.
The power Pm taken by the wind turbine is a purpose of the blade radius, the pitch
angle, and the rotor speed [13]. Figure 13 displays the simulated power curves for
dissimilar wind speeds [5] (Figs. 11 and 12).

1
Pm = πρCP (λ, β)R 2 ϑ 3 (7)
2
 
116 21
Cp (λ, β) = 0.22 − 0.4β − 5 e λi (8)
λi

Fig. 10 Doubly fed induction generator [3]

354 T. Joshiram and S. V. R. Laskhmi Kumari

Fig. 11 Basic schematic representation of the UPQC

Fig. 12 Proposed UPQC model [8]

1 1 0.035
= − 2 (9)
λi λ + 0.08β β +1

2.3 Double Fed Induction Generator (DFIG) Modeling

The Doubly Fed Induction Generator (DFIG)-based wind turbine with variable-speed
variable-pitch control game plan is the most extreme common wind power generator
in the wind power industry. This machine can be worked also in grid associated or
independent mode. In this task an exhaustive electromechanical model of a DFIG-
based wind turbine associated with power grid or just as independently worked
wind turbine system with incorporated battery energy stockpiling is set up in the
MATLAB/Simulink area and its adjusting generator and turbine control erection is
executed. Natural [8].

Model of DFIG
The DFIG contains of stator winding and the rotor twisting outfitted with slip rings.
The stator is giving 3-stage protected windings produce up a picked post plan and
is associated with the matrix through a 3-stage transformer. The same to the stator,
the rotor is additionally worked of 3-stage protected windings. The rotor windings
Power Quality Enhancement and Low Voltage Ride Through … 355

Fig. 13 FLC-based series

converter controller

FPO

are associated with an outside fixed circuit through a bunch of slip rings and brushes
[11].

3 UPQC Topology

The arrangement of the both DSTATCOM and DVR can handle the power quality
of the source current and the load bus voltage. Furthermore, if the DVR and
DSTATCOM are associated on the DC side, the DC bus voltage can be managed by
shunt associated on the DSTATCOM while the DVR supplies the necessary energy
to the load if there should arise an occurrence of the transient’s disturbance in source.
The setup of such a device is displayed in Fig. 11. The DG is related in among
the dc connection of the UPQC. The recreation of the arranged method has been
endorsed out by MATLAB/SIMULATION [11].

4 Formation of Hybrid UPQC

The arranged technique embraces of the consolidated rating star connected

transformer-based UPQC appended with DG. The 3 phase source is taken from
the 1 MW system. The series converter is associated through the series reactor from
the line correspondingly through the shunt reactor is shunt converter. The series and
356 T. Joshiram and S. V. R. Laskhmi Kumari

shunt converter is connected with common DC link connection and capacitor. The
arranged UPQC model is displayed in Fig. 12 [13].

4.1 Series Converter Control

The various disturbances like switching operation and different faults occur in the
distribution system causes voltage sags and swell. It influences the customer equip-
ment cruelly. The series converter compensates the voltage sags and swells in the
distribution system. The fuzzy logic controller-based series converter controller is
displayed in Fig. 13. The DC link limited voltage is contrasted and the reference
voltage by comparator [7]. The blunder signal acquired from the comparator is
handled with FLC 1. The real worth of voltage in phase a, b, c is prepared with
the magnitude of the injected voltage in series converters. The real value of voltage
in phase a, b, c is handled with the magnitude of the injected voltage in series
converters. This output value is contrasted and output of FLC 1 by comparator. The
amplitude of voltage is utilized for reference current estimation [7].

4.2 Shunt Converter Control

Because of expanding in nonlinear load and power electronic equipment in distri-

bution system causes harmonics. This harmonic and is compensate by the shunt
converter. The dc link voltage is detected and contrasted with reference voltage [11].
The error signal is handled and it considered as the magnitude of the 3 phase supply
current references. The reference current is determined by utilizing the unit vector
in phase, with the real supply voltage the 3 phase unit vector in phase is inferred as
in [12].
vsa vsb vsc
u sa = , u sb = , u sc = (10)
vsm vsm vsm

where vsm is the amplitude of supply voltage. vsa , vsb , vsc are the three phase supply
voltage. u sa , u sb , u sc are the multiplication of three phase unit current vectors. The 3
phase shunt current for compensation of harmonics as shown in 10 [12]. The design of
UPQC depends on the parameter specification of the distribution system. The 1 MW
grid is considered in the system. The fifth, seventh and eleventh order harmonics are
made in this plan. The decreased rating star connected transformer is connected with
the UPQC, whereas the industrial and domestics loads are associated in close to the
shunt converter side [7] (Fig. 14).
Power Quality Enhancement and Low Voltage Ride Through … 357

Fig. 14 MATLAB Simulink model of the DVR [2]

5 Results and Discussion

The UPQC has simulated using the proposed hybrid UPQC with DG. The source
voltage waveform before and after connecting the UPQC are analyzed. It noticed that
the source voltage is distorted before connecting the UPQC and it becomes sinusoidal
after connecting the UPQC. The voltage waveform on source side without UPQC
is shown in Fig. 15 and with UPQC is shown in Fig. 16. It has clearly shown that
the voltage sag and swell present in the waveform is compensated after connecting
the UPQC. The voltage sags and swell present in the load side are also reduced,
due to source side compensation [9]. Hence, the power quality of the system can be
improved (Figs. 17, 18 and 19; Table 1).
DVR is proved to compensate voltage levels under faulty conditions. Voltage
harmonics has been reduced considerably. Harmonics generated at source side has
THD of 30.5% which has been compensated to 3.6% at load end. Even the voltage
sag during fault duration has also been compensated to a desired level [1]. UPQC
is proved to compensate current and voltage levels under faulty conditions. Voltage
and current harmonics have been reduced considerably. Current harmonics generated

Fig. 15 Load voltage waveform without UPQC

358 T. Joshiram and S. V. R. Laskhmi Kumari

Fig. 16 Load voltage waveform with UPQC and DVR [2]

Fig. 17 THD without UPQC in load side [6]

Fig. 18 THD with UPQC and DVR on load side

Power Quality Enhancement and Low Voltage Ride Through … 359

Fig. 19 THD with UPQC in source side

Table 1 THD % comparison with and without D-FACTS

Location Without UPQC With DVR (%) With DSTATCOM With UPQC (%)
(%) (%)
THD in source 30.5 30.5 30.5 1.1
side
THD in load 38.83 3.69 4.12 1.21
side

at load side has THD of 30.24% which has been compensated to 1.21% at PCC.
Voltage harmonics generated at source side has THD of 1.45% which has been
compensated to 1.06% at load end [2]. The power quality is improved and the power
oscillation overshoot reduction control of rotor speed and preventing the system from
having a DC link overvoltage and thus increasing the stability of the power system
in accordance with LVRT requirements [2].
Table 2 shows the system parameters.

Table 2 System parameter

Parameters Values
[2]
MV grid Grid L–L voltage 30 kV
Frequency [2] 50 Hz
DFIG Power 500 kW
Frequency 50 Hz
Nominal L–L voltage 400 V
Photovoltaic Plant capacity 500 kW
Vdc 896 V
Parallel strings 102
Series modules 12
UPQC Capacitances 1100 uF
Filter inductance 30 mH
Filter capacitances 28 uF
Star connected transformer 20 kV
360 T. Joshiram and S. V. R. Laskhmi Kumari

6 Conclusion

This paper presents a hybrid UPQC and DVR in distribution systems for simul-
taneous compensation of load current harmonics, voltage sag/swell and source
neutral current. The performance of proposed UPQC and DVR has been investi-
gated through extensive simulation studies. From these studies it is observed that
the proposed scheme completely compensated the source current harmonics, load
current harmonics, voltage sag/swell and neutral current [2]. Even the current and
voltage level during fault duration has also been compensated to a desired level
[3]. Future scope. The more advanced controllers such as fuzzy controller, artificial
neutral network, AUPF, ISCT, AGCT, IGCT theories can also be used with UPQC
to make the system more effective [9].

References

1. Abas N, Dilshad S, Khalid A, Power quality improvement using dynamic voltage restorer.
IEEE Access. https://doi.org/10.1109/ACCESS.2020.3022477
2. Benali A, Khiat M, Allaoui T, Denaï M, Power quality improvement and low voltage ride
through capability in hybrid wind-PV farms grid-connected using dynamic voltage restorer.
https://doi.org/10.1109/ACCESS.2019
3. Karthikeya P, Gonsalves R, Senthil M (2019) Comparison of UPQC and DVR in wind turbine
fed FSIG using asymmetric faults. Int J ELELIJ 3(3)
4. Pota HR, Hossain J (2019) Robust control for grid voltage stability high penetration of
renewable energy, 1st edn. Springer, Berlin, pp 1–11
5. Swain SD, Ray PK (2019) Improvement of power quality using a robust hybrid series active
power filter. IEEE Trans Power Electron 32(5)
6. Improvement of power quality using a hybrid UPQC with distributed generator. In: 2016
International conference on circuit, power and computing technologies (ICCPCT). IEEE. 978-
1-5090-1277-0/16/$31.00©2021
7. Dosela MK, Arson AB, Gülen U (2019) Application of STATCOM-supercapacitor for low-
voltage ride-through capability in DFIG-based wind farm. Neural Comput Appl 28(9):2665–
2674
8. Kosala M, Arson AB (2021) Transient modelling and analysis of a DFIG based wind farm with
supercapacitor energy storage 78:414–421
9. Noureddine O, Ibrahim AMA (2021) Modelling, implementation and performance analysis
of a grid-connected photovoltaic/wind hybrid power system. IEEE Trans Energy Convers
32(1):284–295
10. Rashid G, Ali MH (2021) Nonlinear control-based modified BFCL for LVRT capacity
enhancement of DFIG-based wind farm. IEEE Trans Energy Convers 32(1):284–295
11. Dosela MK (2021) Enhancement of SDRU and RCC for low voltage ride through capability
in DFIG based wind farm 99(2):673–683
12. Dosela MK (2021)Nonlinear dynamic modelling for fault ride-through capability of DFIG-
based wind farm 89(4):2683–2694
13. Ghosh S, Malasada S (2021) An energy function-based optimal control strategy for output
stabilization of integrated DFIG flywheel energy storage system. IEEE Trans Smart Grid 8(4)
Optimal LFC Regulator for Frequency
Regulation in Multi Area Power System

V. Santoshi Raju and P. Venkatesh

Abstract In a power system, frequency control is more sensitive to the variations of

load. So, it needs proper balancing of generation and load demand. Linear controllers
such as proportional integral (PI) do not account the non-linearity of the system.
Their performance with non-linear power system may be ineffective. This project
will present Sliding Mode Control (SMC) which has distinguished properties of
accuracy, robustness and easy tuning performance. Due to these advantages SMC
technique is effectual for LFC and most vigorous to the base parameter and uncer-
tainties. The anticipated controller is set side by side with Proportional-Integral
Derivative (PID), Tilt-Integral Controller (TID) and Fractional Order PID (FOPID)
controllers for performance analysis and Grey Wolf Optimization (GWO) will be
used to tune the parameters of controllers in Load Frequency Control (LFC). The
proposed controller is applied to two area single unit LFC power system consist of
thermal unit. To validate the usefulness of controllers, Integral of the Time Weighted
Absolute Error (ITEA) performance index will be considered and 1% step load
change will be applied. The resulting power system will be modelled and simulated
in MATLAB/SIMULINK environment. The proposed work is effective in main-
taining the frequency deviation, undershoots and overshoots of the power system to
zero in a considerably less time with SMC compared to other controllers.

Keywords Proportional-integral-derivative (PID) · Tilt-integral-derivative (TID) ·

Fractional order-proportional-integral-derivative (FOPID) · Sliding mode
controller (SMC) · Grey wolf optimization (GWO)

1 Introduction

Frequency deviation control in the power system is the most general strategy for
the efficient operation of power systems [1–5]; hence the power demand should be
coordinated to the power generation whenever source and load varies [1]. The Load

V. Santoshi Raju (B) · P. Venkatesh

VR Siddhartha Engineering College, Kanuru, Vijayawada, Andhra Pradesh 520007, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 361
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_27
362 V. Santoshi Raju and P. Venkatesh

Frequency Control (LFC) will make sure that the steady state errors in the power
system will be maintained zero throughout the operation for a two area power system,
where the two areas are coupled all the way through a tie-line [6]. The operation of
LFC also includes to curtail the unexpected tie-line power flows between interrelated
adjacent areas and also to curb the transient variations of multi area frequencies. The
most intricate problem in LFC is when power transaction takes place between inter-
connected multi areas, when uncertainties rises in distributed generator and model
parameters [7]. In a conventional LFC, Area Control Error (ACE), also termed as
control input signal compromises of multi area tie-line power exchanges and local
area frequency [8]. In order to meet up with the grid performance levels, the widely
used controller to control the ACE is Proportional-Integral (PI) controllers. In order
to tune the parameters of PI controller many intelligent optimization techniques are
used which will enhance dynamics of the controller and also used to improve the
robustness of the controller in conditions power system function state variations [9]
and [10].
In the contemporary days, bunch of recent direct strategy came into existence
like optimal control techniques [2], distributed control techniques [3], robust control
technique [11] and hybrid control algorithms, are significantly used for LFC design
[12]. On comparing all the controllers, Sliding Mode Control (SMC) dynamics has a
unique feature which can be modelled regardless of disturbances and system parame-
ters, which will enhance the robustness and response speed of LFC [4]. For a two area
single unit LFC, second-order SMC and extended disturbance observer is proposed
in [13].
This paper proposes a second-order SMC algorithm with an additional extended
disturbance observer for a two area LFC scheme. In order to trim down the intri-
cacy of the power system, the load change and the tie-line power are considered
as one single parameter, i.e. lumped disturbance parameter so that the order of the
power system will be reduced then the extended disturbance observer is used to
estimate the lumped disturbance parameter. SMC requires a sliding surface calcu-
lated through state variable transformation which compels the frequency deviation to
zero without an integral unit. Here, sliding surface matrix is used to tune the system
dynamics and desirable sliding surface matrix can be calculated through optimal
sliding manifold design or Eigen value assignment method. During load variations,
if the scheduled power requires any changes, the modelled scheme will work more
efficiently. In order to eliminate the chattering effect, the second-order sliding mode
control technique along with super-twisting algorithm is engaged which will compel
the sliding surface to reach the sliding surface, respectively. Therefore, the modelled
robust LFC can effectively utilize the benefits of SMC and thus it permits very low
real-time computational burden.
Optimal LFC Regulator for Frequency Regulation in Multi Area … 363

2 Grey Wolf Optimization (GWO)

Grey wolf optimization (GWO) technique is the latest meta-heuristic methods

proposed by Mirjalili et al. in 2014 [5]. Grey wolfs lives in packs and is consid-
ered to be at the peak of the food chain. In Grey wolfs, they are further divided into 4
sub-groups, i.e. alpha (α), beta (β), delta (δ), and omega (ω) and alpha (α) category is
considered to be the top of all the wolf’s in ranking and hence alpha (α) is considered
to be the healthy solution and the second, third and fourth best are beta (β), delta (δ)
and omega (ω) [14].
Generally, Grey wolf’s follow a particular and rare strategy to attack the prey.
Initially, they hunt, chase and attack the prey, and then they blockade the prey and
finally attack the prey.
The encircling behaviour of Grey wolf can be represented in Eqs. (1) and (2):
→ − → 
 = −
D
→ −
C . X p (t) − X p (t) (1)

−
→
X (t + 1) = X p (t) − A.
D  (2)

where T is present iteration, C, D and A is vector coefficient, X p is prey’s tracking

vector and X is Grey wolf’s tracking vector.
The vector A and C can be estimated as shown in Eqs. (3) and (4):

a .−
A = 2 →
r 1 − a (3)

C = 2.−
→
r 2 (4)

r1 and r2 are the arbitrary vectors and ‘a’ decreases from 2 to 0 during the repetition.
The chase is leaded by alpha (α), beta (β), delta (δ) and omega (ω). One of the
advantages of GWO algorithm is it is simple to apply due to its simplex composition,
low required memory and computational necessity [14].

3 Fractional Order Controllers

PI controller is the most conventional controller and is widely used to tune the
parameters of LFC in power system. Day by day, as the order of LFC increases and
multiple areas are inter-connected through the tie-line which enhances intricacy of
the power system and degrading the performances of the orthodox controllers. In
order to increase the effectiveness of the power system, a non-integer order control
or a fractional order control (FOC) came into existence which is purely based on
fractional calculus. There are different kinds of FOC’s like Tilt Integral Controller
(TID), fractional order PID (FOPID).
364 V. Santoshi Raju and P. Venkatesh

3.1 Tilt-Integral-Derivative (TID) Controller

TID controller is the FOC which is used to finely tune the LFC parameters. Since
it is a non-linear controller, TID generally works on three controllers (T, I, D) and
an additional parameter (n) is used here for tuning purpose. TID controller is almost
same like a PID controller, but the proportional characteristics are substituted
 1 by
−n
tilt a proportional characteristic (represented by the transfer function s ) which
provides a frequency function called feedback gain and it is tilted with reference to
the gain of the traditional controller. On comparison to PID, TID controller offers
high level of flexibility to control variables.

T T
g(s, β) = 1 + + Ds (5)
S n S

where
⎡⎤
T
⎢I ⎥
β=⎢ ⎥
⎣D⎦ ∈ R
4
(6)
n

and n ∈ R and n = 0 and 2 < n > 3, u(s) is control signal, r (s) is reference signal e(s)
is error signal, y(s) is output signal and g(s, β) is transfer function of TID controller
[s ∈ Z , β ∈ R].

3.2 Fractional Order PID (FOPID) Controller

FOPID is the extension of PID controller which is purely based on fractional order
differential calculus. FOPID gives better response than conventional controllers due
to presence of five parameters that gives good design flexibility to design the deriva-
tive and integral components. λ and μ are additional parameters to the conventional
PID in FOPID and can be expressed as P I λ D μ . The two additional parameters λ of
integration and μ of derivative also made the tuning of the new FOPID controller
more flexible.
The transfer function of FOPID controller can be represented as Eq. (7):

U (s) 1
G(s) = = Kp + KI λ + KDsμ (7)
E(s) s

where K p is proportional gain, K D is differential gain, K I is integral gain, λ is degree

of integration and μ is degree of differentiation.
Optimal LFC Regulator for Frequency Regulation in Multi Area … 365

The PI λ Dμ controller is more accurate and gives a prospect to further regulate the
variations in control system.

4 Sliding Mode Controller (SMC)

The ith area system dynamics of a multi area inter-connected power system can be
represented as shown below (Eqs. 8–11) [4]. The frequency and the power exchange
between the LFC’s of inter-connected areas in a multi area power system should be
maintained constant throughout the operation [4].

1 1 Di 1
 f˙i = Pmi − PLi −  fi − Ptie,i (8)
2Hi 2Hi 2Hi 2Hi
1 1
 Ṗmi = Pgi − Pmi (9)
Tti Tti
1 1 1
 Ṗgi = Pci −  fi − Pgi (10)
Tgi Ri Tgi Tti
N
 Ṗtie,i = 2π Ti j  f i −  f j (11)
j=1, j=1

This LFC model is as it is taken from [4] because the research studies reveal that
this LFC model is more practical and reliable and will give results without disturbing
the accuracy of the system.
Where i is the area of the system,  f i is fluctuations in system frequency, Pmi
is output of synchronous machine, Pgi is position of valve, Pci is output of
the controller, PLi is load variations, Ti j is tie-line coefficient, Hi is synchronous
machine inertia, Di is damping coefficient of machine, Tgi is the governor time
constant, Tti is the turbine time constant, Ri is the speed drop, Ptie,i is the deviation
between the actual and the scheduled power flows.
Ptie,i Can be evaluated as:

N
Ptie,i = Ptie,act j − Ptie,sched j (12)
j=1, j=1

In the classical matrix form, the system changes can be written as:

ẋi (t) = Ai xi (t) + B i u i (t) + F i PLi (13)

where
 T
State variable matrix: xi (t) =  f i Pmi Pgi Ptie,i ,
366 V. Santoshi Raju and P. Venkatesh

Control input: u i = Pci ,

⎡ ⎤
− 2HD 1
2Hi
0 − 2H
1 ⎡ ⎤ ⎡ ⎤
⎢
i
⎥
i
0 − 1
⎢ 0 − T1ti 1
Tti
0 ⎥ ⎢ 0 ⎥ ⎢ 2Hi ⎥
⎢ ⎥ ⎢ ⎥ ⎢ 0 ⎥
Ai = ⎢ − Ri1Tgi 0 − T1gi 0 ⎥, B i = ⎢ 1 ⎥, and F i = ⎢ ⎥
⎢ ⎥ ⎣ Tgi ⎦ ⎣ 0 ⎦
⎣ N ⎦
2π Ti j 0 0 0 0 0
j=1, j=1

The transfer function for hydro turbine is

1 − Tti s
G hydro =
1 + 0.5Tti s

Now the LFC model of a multi area power system can be inscribed as:
  

ẋi (t) = Ai xi (t) + B i u i (t) + F i PLi (14)

where
 T
The state variable matrix: xi (t) =  f i Pmi Pgi Ptie,i ,
Control input: u i = Pci ,

⎡ ⎤
− 2H
D 1
0 − 2H
1
⎡ ⎤ ⎡ ⎤
⎢ i 2Hi i
⎥ 0 − 1
⎢ 2
− T2ti T2ti + 2
0 ⎥ ⎢− 2 ⎥ ⎢ 2Hi ⎥

⎢ Ri Tgi Tgi ⎥ 
⎢ Tgi ⎥ 
⎢ 0 ⎥
Ai = ⎢
⎢ − Ri1Tgi 0 − T1gi 0 ⎥ ⎥, B i = ⎢ ⎥ and F i = ⎢ ⎥
⎢ ⎥ ⎣ T1gi ⎦ ⎣ 0 ⎦
⎣ N
⎦
2π Ti j 0 0 0 0 0
j=1, j=1

The ultimate purpose of the LFC is to uphold the system frequency invariable,
i.e. the change in frequency  f i = 0. Equations (13) and (14) shows when there are
alteration in the system load, PLi , and scheduled tie-line power, Ptie,i ,  f i should
be driven to zero by regulating the generator control output Pci = PLi + Ptie,i
and therefore the alteration in the system, load frequency and the scheduled tie-line
power together dumped into the single parameter, i.e. extended disturbance. The
power system variations can be expressed as (Fig. 1):

ẋi (t) = Ai xi (t) + Bi u i (t) + Fi Pdi (15)

where
Optimal LFC Regulator for Frequency Regulation in Multi Area … 367

Fig. 1 Thermal unit with extended disturbance observer and SMC

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
D
− 2H 1 0
2Hi 0 0
⎢ i ⎥  ⎢
Ai = ⎢ −Tt1i −Tt2i ⎥ T ⎦, Bi = ⎣ Tg2i ⎥
⎥  ⎢
⎣ Tg1i ⎦, Bi = ⎣ g2i
1 1
⎦ and Pdi = PLi + Ptie,i
− R 1T 0 −T 1
Tgi Tgi
i gi gi

On the basis of the new LFC system proposed in Eq. (15) and the estimated
disturbance observer obtained from disturbance observer, a novel sliding mode is
proposed using system state variable transformation.

4.1 Disturbance Observer to Estimate Extended Disturbance

From Eq. (8), the state variables can be modelled as Eq. (16):


ẋi (t) = Ai xi (t) + Bi u i (t) + Fi d i + L yi − y i



(16)

 

yi = C x i (17)

where L = feedback matrix and C = output matrix [4].

A disturbance observer vector can predict the disturbance using the estimated
states as in Eq. (18):

 
 d i = βi + M xi
  
 
(18)
β̇i = −M Ai xi + Bi u + Fi d i + d i

where β̇i is a supplementary variable and M is a gain matrix constant [4].

368 V. Santoshi Raju and P. Venkatesh

Here, the first order derivative of disturbance is negligible, i.e. zero due to the
slow system load changes during the LFC operation.

4.2 Design of Sliding Surface

The main objective of the sliding mode control is to regulate the system and further
arrive at the sliding surface. Designing of sliding surface is completely dependent
on the control objective. The desired conditions of control objective is  f i = 0,
Pmi = Pdi , Pgi = Pdi and Pci = Pdi . In order to satisfy the requirements,
the new state variables are derived:
⎫
 fi =  fi − 0 ⎬
ηi = Pmi − Pdi (19)
⎭
ξi = Pgi − Pdi

By substituting (19) in (15), the power system dynamic equation can be rewritten
as:
      
 f˙i − 2H
Di 1
 fi 0
= i 2H i + ξi
η̇i Tg1i −Tt1i ηi Tt2i
   f  1 1 1
ξ̇i = − 1
0 i
− ξi + ui − Pdi (20)
Ri Tgi
ηi Tgi Tgi Tgi

The state variable ξi is directly proportional to the input variable u i . In order to
derive the sliding surface, the sliding variable can be selected as:
 
 fi
si = ξi − k , K ∈ R 1×2 (21)
ηi

When the system changes are only limited to the sliding surface si = 0, the
modified reduced model can be given as Eq. (22):
       
 f˙i − 2H
Di 1
0  fi
= i 2Hi + K (22)
η̇i Tg1i −Tt1i Tt2i ηi

The parameters of K can be calculated with propose methods i.e. Eigen value
assignment method [4].
Optimal LFC Regulator for Frequency Regulation in Multi Area … 369

4.3 The Control Law Based on Super-Twisting Algorithm

The super-twisting algorithm based LFC have 2 objectives: an Equivalent controller

to compensate the power system dynamics and the sliding mode control to compen-
sate the unmodelled dynamics and disturbance arising due to errors present in the
modelled parameter.
The equivalent controller can be given in Eq. (23):
   
   f 1
ui = − R1 0 + K + Tgi K (A11 + A12 K ) i + Tgi − − K A12 si − Tgi vi (23)
i ηi Tgi

Equation (20) is transformed to Eq. (24) as:

   
 f˙i  fi
= (A11 + A12 K ) + A12 si (24)
η̇i ηi

ṡi = vi + f˜( f i , ηi , t) (25)

where f˜( f i , ηi , t) = the system disturbance, unmodelled changes and system
uncertainties. The sliding mode controller based on the super-twisting algorithm is
given by Eq. (26):

t
v = −k1 ∅1 (s) − k2 ∅2 (s)dt (26)
0
1
∅1 (s) = |s| 2 sign(s) (27)

1
∅2 (s) = ∅1  (s)∅1 (s) = sign(s) (28)
2
where
⎧
⎨ 1, s > 0
sign(s) = 0, s = 0
⎩
−1, s < 0
 
 
If  f˜( f i , ηi , t) has a boundary condition ρ|s| 2 , where ρ > 0, the condition
1

for the globally asymptotic at origin si = 0 is given in Eqs. (29) and (30):

k1 > 2ρ (29)
370 V. Santoshi Raju and P. Venkatesh

Fig. 2 Frequency deviations of two areas using of PID controller

5ρ + 4ρ 2
k2 > k1 (30)
2k1

5 Results

Parameters of non-reheat turbine of area-1: Tt1 = 0.5 (s), Tg1 = 0.2 (s), H1 = 5
(s), D1 = 0.6 (pu/Hz), R1 = 0.05 (pu/Hz) and K I i = 0.3. Parameters of non-reheat
turbine of area-2: Tt1 = 0.6 (s), Tg1 = 0.3 (s), H1 = 4 (s), D1 = 0.3 (pu/Hz),
R1 = 0.0625 (pu/Hz) and K I i = 0.3.
Figure 2 shows the obtained graph when PID controller is used to tune the proposed
two area power system and Fig. 3 shows the obtained graph when FOPID controller
is used to tune the power system and it is showing best results in comparison with
PID.
Figure 4 shows the obtained graph when the TID controller is used to tune the
two area power system and Fig. 5 shows the obtained graph when SMC controller is
used to tune the proposed two area power system.

6 Conclusion and Future Scope

A second-order SMC algorithm with an additional extended disturbance observer for

a two area LFC scheme is proposed in this paper. For the proposed two area single unit
power system, the overshoot and undershoot of the SMC is less than the PID, FOPID
and TID controllers (shown in Table 1) which enables the more efficient operation of
LFC. Though the settling time of TID controller is nearby SMC but on comparing the
overall overshoots, undershoots and settling times of all the controllers, SMC yields
Optimal LFC Regulator for Frequency Regulation in Multi Area … 371

Fig. 3 Frequency deviations of two areas using FOPID controller

Fig. 4 Frequency deviations of two areas using TID controller

Fig. 5 Frequency deviations of two areas using SMC

372 V. Santoshi Raju and P. Venkatesh

Table 1 Controller parameters tuned using GWO algorithm

Controllers KP KT KI KD
F1 F2 F1 F2 F1 F2 F1 F2
PID 5.4062 25.5691 – – 41.2698 12.864 8.69693 5.06956
FOPID 2.0842 0.53467 – – 2.6327 6.1092 1.0024 1.4542
TID – – 5.4062 25.5691 41.2698 12.864 8.69693 5.06956

Table 2 Comparison of settling times, overshoots and undershoots for various controllers
Controllers Settling time (in s) Overshoot Undershoot
F1 F2 F1 F2 F1 F2
PID 48 50 0.0005 − 0.0014 0.00003 − 0.00007
FOPID 40 30 0.0019 − 0.0004 0 − 0.0034
TID 18 24 0.0003 − 0.0022 0.00001 − 0.00013
SMC 16 26 0 − 0.0014 0 − 0.000022

the efficient results. The proposed work is effective for maintaining the frequency
deviation of the power system to zero in a considerably less time and also reduces the
overshoots and undershoots with SMC compared to other controllers which enables
the efficient operation of the power system (Table 2).
Further, FACTS devices can be incorporated and also the proposed work can be
applied to the deregulated power system for more efficient operation.

References

1. Siti MW, Tungadio DH, Sun Y, Mbungu NT, Tiako R (2019) Optimal frequency deviations
control in microgrid interconnected systems
2. Daneshfar F, Bevrani H (2012) Multi objective of load frequency control using genetic
algorithm
3. Singh VP, Kishor N, Samuel P (2016) Distributed multi-agent system based load frequency
control for multi-area power system in smart grid
4. Liao K, Xu Y (2017) A robust load frequency control scheme for power systems based on
second-order sliding mode and extended disturbance observer
5. Mirjalili S, Mirjalili SM, Lewis A (2014) Grey wolf optimizer
6. Hossain MdM, Peng C (2021) Observer-based event triggering H1 LFC for multi-area power
systems under DoS attacks
7. Ali H, Madby G, Xu D (2021) A new robust controller for frequency stability of interconnected
hybrid microgrids considering non-inertia sources and uncertainties
8. Khodabakhshian A, Edrisi M (2007) A new robust PID load frequency control
9. Liu F, Li Y, Cao Y, Jinhua S, Wu M (2015) A two-layer active disturbance rejection controller
design for load frequency control of inter connected power system
10. Tan W (2009) Unified tuning of PID load frequency controller for power systems via IMC
11. Yousef HA, Al-Kharusi K, Albadi MH, Hosseinzadeh N (2013) Load frequency control of a
multi-area power system: an adaptive fuzzy logic approach
Optimal LFC Regulator for Frequency Regulation in Multi Area … 373

12. Bevrani H, Daneshmand PR, Babahajyani P, Mitani Y, Hiyama T (2013) Intelligent LFC
concerning high penetration of wind power: synthesis and real-time application
13. Chen C, Zhang K, Yuan K, Wang W (2017) Extended partial states observer based load
frequency control scheme design for multi-area power system considering wind energy
integration
14. Mohanty S, Subudhi B, Ray PK (2015) A new MPPT design using grey wolf optimization
technique for photovoltaic system under partial shading conditions
Optimal Power Flow Using Firefly
Algorithm with Solar Power

Thangella Aravind and Balusu Srinivasa Rao

Abstract One of the most intractable problems in power system networks is the
optimal power flow problem (OPF). The firefly algorithm (FA), among the most
popular meta-heuristic nature-inspired algorithms, is used to solve the OPF problem.
This research uses FA to solve the optimal power flow problem with the addition of
a solar energy system. The goal of this study is to reduce total fuel cost, minimize
L-index (voltage stability index) and minimizing real power loss. The effect of incor-
poration of renewable energy system into OPF problem is studied on 30-bus IEEE
test system. The proposed method has been implemented in MATLAB program, and
these results are compared with various algorithms available in the existing literature.

Keywords Firefly algorithm (FA) · Optimal powerflow (OPF) · Solar energy

system · Voltage stability index (L-index) · Transmission losses

1 Introduction

One of the very hard problems in power system networks is the optimal power flow
(OPF) problem. During the span of time, many researches came into existence in
OPF to reduce the optimization problems using different methods. In recent years,
the OPF is a major task in renewable energy sources [1]. OPF problem is the main
intention on three major conflicting objectives, i.e. minimization of generation cost,
transmission losses, L-index [2]. In 1962, the OPF is first discussed in Carpentier. The
power system network has to satisfy several constraints while maintaining generation
costs as low as in an electrical network. There are two types of system constraints in
a network: inequality and equality constraints [3]. An equality constraint is defined
as to maintain the power balance equations, and the various inequality constraints

T. Aravind (B) · B. S. Rao

V R Siddhartha Engineering College, Kanuru, Vijayawada, A.P. 520007, India
B. S. Rao
e-mail: balususrinu@vrsiddhartha.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 375
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_28
376 T. Aravind and B. S. Rao

of a power system network are required to maintain the system operating limits and
security limits.
Predictable and artificial intelligence (AI) these is the solution of OPF problem
methods. OPF is made up of a variety of universal techniques and has some drawbacks
[4], i.e. continuous-time, slow convergence, and qualitative features are very weak
in handling and operation is slow. Many authors are most preferred in artificial
intelligence method since to get the optimal solution in global or approximate global.
These approaches have a number of advantages, including the ability to deal with
a variety of qualitative constraints, a single execution to obtain a large number of
optimal solutions, the ability to solve multi-objective optimization problems, and
the ability to find a global optimum solution [5]. The firefly algorithm is employed
in this study to solve the multi-model optimization problem discovered by Xinshe
Yang’s [6]. It stands on the flashing behaviour of the bugs, including light emission,
light absorption and the mutual attraction. There are various types of meta-heuristic
algorithms that are differential evolution (DE) algorithm, artificial bee colony (ABC)
algorithm, particle swarm optimization (PSO), clonal selection (CS) algorithm which
are also similar to the proposed firefly algorithm [7]. FA is more useful for controlling
parameters and also local searching ability, robustness, fast convergence [8]. The
latest crowd intelligence gathering that utilizes firefly algorithm (FA) is proffered to
determine the solution of OPF problem.
Wang Yi-BO [9] this paper presents the under structure of analysing steady-state
characteristics of photovoltaic (PV) system connected to the power grid. Basically,
the PV system consists of power converters. A PV system is separated into three basic
modules: alternative current (AC) module, direct current (DC) module and inverter
module.
This chapter is structured into seven sections as follows. The mathematical
modelling of OPF problem formulation is presented in second section. Modelling
of solar energy system is discussed in Sect. 3. The concept of FA is explained in
fourth section. Section 5 discusses how to include the FA into OPF. In Sect. 6, FFA
results obtained with MATLAB program are discussed. In Sect. 7, valid conclusions
are discussed, and the last one is references.

2 Mathematical Problem Formulation of OPF

In any power system network, some objectives are reduced, and they met inequality
and equality constraints. The OPF is a disordered optimization problem. Below
equation represents the basic form of OPF problem.

Minimize : f (l, m)
 
g(l, m) ≤ 0
Subject to : (1)
h(l, m) = 0
Optimal Power Flow Using Firefly Algorithm with Solar Power 377

where m—Independent control variables; l—Dependent state variables; f (l,

m)—OPF objective function; g(l, m)—Inequality constraint; h(l, m)—Equality
constraints.

2.1 Dependent State Variables

These types of variables in a power network can be expressed by vector l as:

 
l = PG 1 , E L 1 ... E L NL , Q G 1 ... Q G NG , Dl1 ... Dlnl (2)

wherePG 1 —Slack bus generator real power; Q G i —Generator reactive power at ith
bus; E L p —Magnitude of bus voltage at pth bus (load bus); Dlq —Line loading of qth
line; NL—Total transmission lines; Nl—Total load buses.

2.2 Independent system Control Variables

In a network controlling, the power flow depends on the variables presented in the
below equation.
 
m = PG 2 ... PG NG , E G 1 ... E G NG , Q c1 ... Q cNC , T1... TNT (3)

where PG i —Generator real power at ith bus; E G m —Voltage magnitude of PV bus;

Q c j —shunt compensation at jth bus; Ti - ith branch transformer taps setting.

2.3 Constraints

Various types of constraints which are to be satisfied by this OPF problem are
discussed in the following section.

2.3.1 Equality Constraints

These constraints are intertwined with both active and reactive power, as seen by the
equations below.


NB
    
PG i − PDi − E i E j K ij cos δij + Bij sin δij = 0∀i ∈ NB (4)
j=1
378 T. Aravind and B. S. Rao


NB
    
Q G i − Q Di − E i E j K ij sin δij − Bij cos δij = 0∀i ∈ NB (5)
j=1

where E i − Voltage magnitude of bus i;δij —are the voltage angles between the
buses j and i; NB—Total no. of buses; PD —Load demand of active power; Q D —
Load demand of reactive power; K ij —Transfer conductance which is connected to
ith bus; Bij —Susceptance which is connected to jth bus.

2.3.2 Inequality Constraints

It is represented to maintain sustainable limits in a power system as shown in below

equations.
(a) Generator Constraints: These constraints apply to both real and reactive power
outputs, with the following upper and lower bounds limiting generator voltages:

E Gmin
l
≤ E G l ≤ E Gmax
l
∀l ∈ NG (6)

PGmin
n
≤ PG n ≤ PGmax
n
∀n ∈ NG (7)

G n ≤ Q G n ≤ Q G n ∀n ∈ NG
Q min max
(8)

(b) Transformer constraints: Minimum and maximum limits limited these

constraints in a transformer setting, expressed as follows.

Timin ≤ Ti ≤ Timax ∀i ∈ NG (9)

(c) Shunt compensator constraints: These constraints are illustrated in reactive

power injected at different buses and maintain upper and lower limits.

c j ≤ Q c j ≤ Q c j ∀ j ∈ NC
Q min max
(10)

(d) Security constraints:

E Lmin
p
≤ E L p ≤ E Lmax
p
∀ p ∈ NL (11)

Slq ≤ Slmax
q
∀q ∈ nl (12)

Equation (11) represents the voltage magnitudes at pth bus, and Eq. (12) represents
the transmission line loading at qth bus.
Optimal Power Flow Using Firefly Algorithm with Solar Power 379

2.3.3 Objective Functions

The following are the three major objective functions that were considered in this
study in order to find the solution of the OPF problem:
a. Minimize cost of generation: This aims to decrease the generation cost of
interrelated generation units. The stranded quadratic expression is given as
follows.

  
Ng
 
f PG k = αk + βk PG k + γk PG2 k $/hr (13)
k=1

where
 αk , βk , γk —Cost coefficients of the kth generator.
f PG k —Fuel cost function; PGk —Generator power output at kth bus.
N g—Total generators.
b. Voltage Stability index (L-index): A power system to maintain voltage of load
buses L-index is used to avoid the voltage fall down point. This can be attained
by minimization of L-index [10], expressed as shown in below equation.

L= min{E j j = 1, . . . FPQ } (14)

where FpQ —total load buses.

c. Minimization of transmission losses: In this objective, to decrease the real power
losses and it is denoted by PLoss .


NL
rk  2  
PLoss = E i + E 2j − 2E i E j cos δi − δ j (15)
r2
i=1 k
+ xk
2

where N L —Number of transmission lines;

rk —Resistance of kth transmission line;
E i ,E j —Voltage at ith and jth bus;
δi ,δ j —Angles at ith and jth bus.

3 Modelling of Solar Energy System

In the contemporary years, photovoltaic power generation is more developed in the

power system network which reduces the pollution and PV generation has more vital
social and economic advantages. One of the main boons of photovoltaic (PV) system
is that it directly converts the solar irradiance into electricity. The PV system gradually
improves the technology and reduces the cost and many countries adopted the PV
generation system in order to reduce the harmful emissions which are dangerous
380 T. Aravind and B. S. Rao

Fig. 1 Equivalent
transformation of AC circuit

for the environment. Commonly, PV system is integrated by power electronics

converters.
One form of renewable energy source is solar energy, when sunlight energy is
directly converted into the electricity using PV panels. When PV panels are made
up on mostly semiconductor materials since this is more gain of sunlight energy
comparison of insulator materials. For the calculation of AC circuit, output power
by using the parallel–series and star/delta transformation is shown in Fig. 1. The PV
panel power output is transformed from inverter, and output of the inverter can be
further transformed into grid as shown in below equations.

Ei Ei Eg  
Pi = 3E i cos ϕ12 + cos ϕ13 − cos θg − α − ϕ12 (16)
Z 12 Z 13 Z 12
Ei Ei Eg  
Q i = 3E i sin ϕ12 + sin ϕ13 + sin θg − α − ϕ12 . (17)
Z 12 Z 13 Z 12
Ei   Eg Eg
Pg = 3E g cos α − θg − ϕ12 − cos ϕ12 − cos(ϕ23 ) (18)
Z 12 Z 12 Z 23
Ei   Eg Eg
Q g = 3E g sin α − θg − ϕ12 + sin ϕ12 + sin(ϕ23 ) (19)
Z 12 Z 12 Z 23

4 Firefly Algorithm

Several innovative algorithms for solving engineering optimization problems have

been introduced in the last couple of decades. Among all these new algorithms, it has
been expressed that firefly algorithm (FA) is the most appropriately planned in dealing
with global optimization problem [6]. FA, which is based on the shining pattern and
social interactions of fireflies, was created in 2007 and 2008 by XinShe Yang at
Cambridge University, including light absorption, light emission and mutual attrac-
tiveness [8, 11]. For the flexibility of new meta-heuristic FA, three major idealized
rules are indicated [12–14].
(1) Generally, fireflies are unisexual; i.e. each firefly will be attracted to the other
firefly in the group despite the sex.
Optimal Power Flow Using Firefly Algorithm with Solar Power 381

(2) Attractiveness α brightness, i.e. any two shinning fireflies, the firefly that is
less luminous will approach the firefly that is brighter. As the distance between
them grows, the brightness’s appeal reduces, and vice versa. If there isn’t a
brighter firefly nearby, it will migrate at random.
(3) The brightness of a firefly algorithm will be resoluted from the landscape of
the objective function.
These three idealized principles are based on, and FA may be clarified in a step-
by-step approach that can be presented as the pseudo-code [15].

5 Firefly Algorithm for Solving OPF Problem

This algorithm is mainly considered two major issues: The first one is a divergence
in light intensity I, while the second is an expression of attraction β. Any brilliant
firefly in a specific point z can be chosen at random as:

1
I (z) ∝ (20)
f (z)

The firefly light intensity I is proportional to distance r. That is,

I (z) = Io e−γ r (21)

whereIo —Starting luminous intensity.

γ —Absorption ratio.
The light intensity observed by surrounding fireflies is related to the attraction of
fireflies; i.e. a firefly’s attractiveness can be calculated as:

β(r ) = βo e−γ m m ≥ 1 (22)

where
βo —Attractiveness at distance r = 0; M—Total fireflies.
The firefly i that is less brilliant goes towards the firefly j that is less luminous.
The updated position of firefly i can be represented as in Eq. (23):
2 
z i = z i + β0 e−γ rij z j − z i + α(rand − 0.5) (23)

with


d
rij = z i − z j  (z i,k − z j,k )2 (24)
k=1

where rij —Parting between the two fireflies j and i at locations z j and z i .
382 T. Aravind and B. S. Rao

α—Randomness parameter.

6 Results and Discussions

The propounded FA method has been practised on a standard 30-bus IEEE system
with a solar energy system for single-objective optimization problem. This test system
included 41 branches, 6 generator buses and twenty-four load buses, 4 transformers,
and 9 shunt compensations on various buses. The test system consists of six thermal
generators (TG) which are placed on the 1st (Slack), 2nd, 3rd, 4th, 5th and 6th buses.
Table 1 shows the minimum and maximum real power generating limits, and cost
coefficients of total generators. Table 2 lists the minimum and maximum voltage
magnitudes, transformer tap settings, and reactive power injections. The overall load
demand is 283.4 MW and 126.2MVAR. This manuscript includes three conflicting
objectives such as total cost, L-index and power loss for optimization. The proposed
FA is applied to find a solution to single-objective optimization with and without
solar energy system.
Case 1-Without solar energy system: Initially, without considering the solar
energy system, each objective function was considered separately for single-objective
optimization using the FA technique. Table 2 shows that the FA is successful in
decreasing total fuel cost, L-index and real power loss. Table 2 shows the optimal
settings for all control variables for 30-bus IEEE system without solar energy. Fig. 2
depicts the convergence plots of these objectives in the absence of a solar energy
system.
Case 2-With Solar energy system: In this part, the proposed FA is used to solve
a single-objective OPF problem with the three objectives mentioned above and the
incorporation of a solar energy system. At 7th bus of 30-bus IEEE system solar
generator is placed. The optimal values of all the control variables obtained using FA
when optimized separately with solar energy system are shown in Table 3. Figure 3
depicts the convergence curves of these objectives with a solar energy system.

Table 1 Cost coefficients

Generator αi βi γi PGmin
i
PGmax
i

TG1 37.5 200 0 0.5 2.0

TG2 175 175 0 0.2 0.8
TG3 83.4 325 0 0.1 0.35
TG4 250 300 0 0.1 0.3
TG5 625 100 0 0.15 0.5
TG6 250 300 0 0.12 0.4
Solar 0 160 0 0.1 0.2
Optimal Power Flow Using Firefly Algorithm with Solar Power 383

Table 2 Best solution with FA when optimized separately without solar system (Case-1)
Variables names Max Min Fuel cost L-index Loss
PG 1 (MW) 200 50 177.03 51.46 51.24
PG 2 (MW) 80 20 48.71 79.98 80.00
PG 3 (MW) 50 15 21.08 35.00 35.00
P4 (MW) 35 10 11.88 29.99 30.00
PG 5 (MW) 30 10 21.30 49.98 50.00
PG 6 (MW) 40 12 12.00 40.00 40.00
V1 (p.u.) 1.10 0.95 1.1000 1.1000 1.1000
V2 (p.u.) 1.10 0.95 1.0878 1.1000 1.0975
V3 (p.u.) 1.10 0.95 1.0693 1.1000 1.0867
V4 (p.u.) 1.10 0.95 1.0906 1.1000 1.0945
V5 (p.u.) 1.10 0.95 1.0618 1.0937 1.0798
V6 (p.u.) 1.10 0.95 1.1000 1.1000 1.1000
T1 1.10 0.90 1.0376 1.0485 0.9838
T2 1.10 0.90 0.9411 1.0289 1.0411
T3 1.10 0.90 0.9643 0.9982 0.9728
T4 1.10 0.90 0.9589 0.9548 0.9706
Q c10 (p.u.) 0.05 0.0 0.0023 0.0500 0.0006
Q c12 (p.u.) 0.05 0.0 0.0500 0.0500 0.0500
Q c15 (p.u.) 0.05 0.0 0.0500 0.0500 0.0500
Q c17 (p.u.) 0.05 0.0 0.0500 0.0500 0.0499
Q c20 (p.u.) 0.05 0.0 0.0500 0.0500 0.0500
Q c21 (p.u.) 0.05 0.0 0.0495 0.0500 0.0485
Q c23 (p.u.) 0.05 0.0 0.0258 0.0500 0.0216
Q c24 (p.u.) 0.05 0.0 0.0268 0.0500 0.0270
Q c29 (p.u.) 0.05 0.0 0.0244 0.0432 0.0232
Cost($/hr) – – 799.0345 967.2860 966.7782
L-index – – 0.1163 0.11012 0.1160
Loss(MW) – – 8.61 3.01 2.8467

The comparison of results both (without solar and with solar) by using the FA
method is shown in Tables 2 and 3. The overall cost is lowered from 799.0345$/hr
to 759.4226$/hr when a solar energy system. The L-index is slightly increased from
0.11012 to 0.11148 with solar energy system. Finally, with a solar energy system the
total power loss is reduced from 2.8467 to 2.4 MW.
Table 4 shows that the proposed FA results for case 1 best among all other tech-
niques currently available in the literature. However, the results obtained with incor-
poration of solar energy systems are not compared with the literature as there is no
similar work found for case 2.
384 T. Aravind and B. S. Rao

a) Cost b) L-Index c) Loss

Fig. 2 Convergence curves without solar system (Case-1)

7 Conclusion

In this paper, a current robust crowd intelligence built on FA with a solar energy
system to work out the OPF problem. The FA was effectively implemented to solve
the OPF problem to optimize the generation cost, L-index and active power loss.
The proposed method is tested on standard 30-bus IEEE system. The FA results
compared with and without solar energy system. The result analysis of the given
test system shows that the proposed FA method is well suitable for handling single-
objective OPF problems using solar power. The future scope of this research will be
a multi-objective OPF problem combining solar and wind power.
Optimal Power Flow Using Firefly Algorithm with Solar Power 385

Table 3 Best solution with FA when optimized separately with solar energy system (Case-2)
Variable names Max Min Fuel cost L-index Loss
PG 1 (MW) 200 50 167.05 107.78 31.03
PG 2 (MW) 80 20 46.12 42.70 79.84
PG 3 (MW) 50 15 15.26 34.70 35.00
P4 (MW) 35 10 10.11 29.75 30.00
PG 5 (MW) 30 10 20.49 17.81 50.00
PG 6 (MW) 40 12 12.00 38.79 39.94
PG 7 (MW) 50 0 20.00 16.96 19.98
V1 (p.u.) 1.1 0.95 1.100 1.0612 1.0373
V2 (p.u.) 1.1 0.95 1.0868 1.0504 1.0358
V3 (p.u.) 1.1 0.95 1.0660 1.0600 1.0239
V4 (p.u.) 1.1 0.95 1.0827 1.1000 0.9991
V5 (p.u.) 1.1 0.95 1.0592 1.0224 1.0187
V6 (p.u.) 1.1 0.90 1.0999 1.0808 1.0491
V7 (p.u.) 1.1 0.90 1.0655 1.0449 1.0228
T1 1.1 0.90 0.9493 1.0207 1.0226
T2 1.1 0.90 1.0573 0.9617 0.9311
T3 1.1 0.90 0.9661 0.9291 0.9695
T4 1.1 0.90 0.9506 0.9558 0.9601
Q c10 (p.u.) 0.05 0.0 0.0202 0.0499 0.0004
Q c12 (p.u.) 0.05 0.0 0.0472 0.0499 0.0540
Q c15 (p.u.) 0.05 0.0 0.0485 0.0499 0.0570
Q c17 (p.u.) 0.05 0.0 0.0498 0.0498 0.0611
Q c20 (p.u.) 0.05 0.0 0.0497 0.0500 0.0873
Q c21 (p.u.) 0.05 0.0 0.0495 0.0498 0.0040
Q c23 (p.u.) 0.05 0.0 0.0283 0.0495 0.0155
Q c24 (p.u.) 0.05 0.0 0.0102 0.0496 0.0273
Q c29 (p.u.) 0.05 0.0 0.0121 0.0500 0.0185
Fuel cost($/hr) – – 759.4226 818.7200 951.3025
L-index 0.1147 0.11148 0.1254
Loss(MW) – – 7.67 5.2 2.4
386 T. Aravind and B. S. Rao

a) Fuel cost b) L-Index c) Loss

Fig. 3 Convergence curves with solar system (Case-2)

Table 4 FA results compared to prior research for the 30-bus IEEE system
Objectives Algorithm Fuel cost L-index Loss
Cost objective FA 799.0345 0.1163 8.61
MSA [16] 800.5099 0.13833 9.0345
BSA [17] 799.0760 0.1273 8.6543
ARCCBO[18] 800.5159 0.1385 9.0255
SKH [19] 800.5141 0.1382 9.0282
DE [20] 799.0827 0.1277 8.63
GEM [21] 799.0463 0.1264 8.6257
L-index objective FA 967.2860 0.11012 3.01
SKH [19] 814.0100 0.1366 9.9056
GEM [21] 816.9095 0.1257 6.2313
DE [20] 915.2172 0.1243 3.626
Loss objective FA 966.7782 0.1160 2.8467
MSA [16] 967.6636 0.13832 3.0844
DSA [22] 967.6493 0.12604 3.0954
ARCBBO[18] 967.6605 0.1386 3.1009
APFPA [23] 965.6590 – 2.8463
GEM [21] 966.7473 0.1265 2.8863

References

1. Muller SC, Hager U, Rehtanz C (2014) A multiagent system for adaptive power flow control
in electrical transmission systems. IEEE Trans Ind Inf 10(4):2290–2299
2. Rao BS (2017) Application of adaptive clonal selection algorithm to solve multi-objective
optimal power flow with wind energy conversion systems. Int J Power Energy Conver 8(3):322–
342
3. Biswas PP et al (2018) Optimal power flow solutions using differential evolution algorithm
integrated with effective constraint handling techniques. Eng Appl Artif Intell 68:81–100
Optimal Power Flow Using Firefly Algorithm with Solar Power 387

4. Bouchekara HREH (2014) Optimal power flow using black-hole-based optimization approach.
Appl Soft Comput 24:879–888
5. Ponsich A, Jaimes AL, Coello Coello CA (2012) A survey on multiobjective evolutionary algo-
rithms for the solution of the portfolio optimization problem and other finance and economics
applications. IEEE Trans Evol Comput 17(3):321–344
6. Yang XS (2013) Firefly algorithm: recent advancements and application. Int J Swarm Intell
1:36–50
7. Mishra N, Pandit M (2013) Environmental/economic power dispatch problem using particle
swarm optimization. Int J Electron Comput Sci Eng (IJECSE) 2(2):512–519
8. Yang X-S (2009) Firefly algorithms for multimodal optimization. Stochastic Algor: Found
Appl SAGA 5792:169–178
9. Wang Y-B et al (2008) Steady-state model and power flow analysis of grid-connected
photovoltaic power system. 2008 IEEE international conference on industrial technology. IEEE
10. Tuan TQ, Fandino J, Hadjsaid N, Sabonnadiere JC, Vu H (1994) Emergency load shedding
to avoid risks of voltage instability using indicators. IEEE Trans Power Syst 9(1):341–351.
https://doi.org/10.1109/59.317592
11. Sarbazfard S, Jafarian A (2016) A hybrid algorithm based on firefly algorithm and differential
evolution for global optimization. Int J Adv Com Sci Appl 7(6):95–106
12. Lukasik S, Zak S (2009) Firefly algorithm for continuous constrained optimization tasks. In:
Nguyen NT, Kowalczyk R, Chen SM, eds. Proceedings of the international conference on
computer and computational intelligence (ICCCI ‘09), vol 5796. Springer, Wroclaw, Poland,
pp 97–106
13. Yang XS (2010) Firefly algorithm, Levy flights and global optimization. Research and
development in intelligent systems XXVI. Springer, London, UK, pp 209–218
14. Yang XS (2010) Firefly algorithm, stochastic test functions and design optimization. Int J
Bio-Inspired Comput 2(2):78–84
15. Subramanian R, Thanushkodi K (2013) An efficient firefly algorithm to solve economic dispatch
problems. Int J Soft Comp Eng (IJSCE) 2(1):52–55
16. Mohamed AAA, Mohamed YS, El-Gaafary AA, Hemeida AM (2017) Optimal power flow
using mothswarm algorithm. Electric Power Syst Res 142:190–206
17. Chaib AE et al (2016) Optimal power flow with emission and non-smooth cost functions using
backtracking search optimization algorithm. Int J Electr Power Energy Syst 81:64–77
18. Kumar AR, Premalatha L (2015) Optimal power flow for a deregulated power system using
adaptive realcoded biogeography-based optimization. Int J Electric Power Energy Syst 73:393–
399
19. Pulluri H, Naresh R, Sharma V (2018) A solution network based on stud krill herd algorithm
for optimal power flow problems. Soft Comput 22(1):159–176
20. Shaheen AM, El-Sehiemy RA, Farrag SM (2016) Solving multi-objective optimal power
flow problem via forced initialized differential evolution algorithm. IET Gen Trans Distrib
10(7):1634–1647
21. Bouchekara HREH, Chaib AE, Abido MA (2016) Multi-objective optimal power flow using a
fuzzy based grenade explosion method. Energy Syst 7(4):699–721
22. Abaci K, Yamacli V (2016) Differential search algorithm for solving multi-objective optimal
power flow problem. Int J Electr Power Energy Syst 79:1–10
23. Mahdad B, Srairi K (2016) Security constrained optimal power flow solution using new adaptive
partitioning flower pollination algorithm. Appl Soft Comput 46:501–522
Designing of Learning Environment
of an Intelligent Tutoring System
for Academic Learning Needs
of Learning-Disabled Learners Based
on Survey Report of Region-Specific
Target Group

Neelu Jyothi Ahuja, Monika Thapliyal, Anand Nayyar, and Adarsh Kumar

Abstract The twenty-first century is known for exponential growth in the techno-
logical as well as the education domain. However, conventional educational tools
are still reliable to understand the actual scenario of performance and efficiency of
the brain of young minds. If combined with technology this could play a vital role in
getting a clear perception about what they feel and how they improve the educational
methodology. Studies on dyslexia, dysgraphia and dyscalculia have shown that it
is very tough and sometimes impossible to identify these learners without the help
of professionals. Unlike physical challenges, the challenges from these disabilities
are not measurable in quantified terms. For this, perception-based studies play a
vital role. There are various studies, which suggest questionnaire-based survey, or
direct interviews with the target group are a more reliable source of information.
Also, regional-level data sometimes plays an important role to understand geograph-
ical and environmental impacts on the target group. Various reports were studied to
understand the similarities. Authors collected information as a pilot project through
a survey on 292 learners (learning-disabled and non-learning-disabled) in different
institutes. 142 of them were dealing with single or multiple learning disabilities.
Study helped in identifying the most affected learning domains and related multiple-
criteria affecting the learners. This eventually is implemented in the domain model

N. J. Ahuja · M. Thapliyal · A. Kumar

Department of Systemics, School of Computer Science, University of Petroleum and Energy
Studies, Dehradun, India
e-mail: neelu@ddn.upes.ac.in
A. Kumar
e-mail: adarsh.kumar@ddn.upes.ac.in
A. Nayyar (B)
Graduate School, Duy Tan University, Da Nang 550000, Vietnam
e-mail: anandnayyar@duytan.edu.vn
Faculty of Information Technology, Duy Tan University, Da Nang 550000, Vietnam

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 389
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_29
390 N. J. Ahuja et al.

of an intelligent tutoring system to develop the four learner-centric learning environ-

ments. Results show that practice-based learning environment was the most relatable
learning environment followed by visual-based learning environment.

Keywords Dyslexia · Dysgraphia · Dyscalculia · Intelligent tutoring system ·

Domain model · Learning disability · Survey · Learning environments

1 Introduction

Learning disabilities (LDs) refer to the condition in which learner repeats similar
errors and faces difficulty in academic learning skills related to reading, writing,
speaking and solving mathematics. The learner may have one or multiple prob-
lems. Although there is no problem with the intelligence of these learners, repeti-
tion in mistakes make them look like ‘problematic’ or ‘uninterested’ person. Some
characteristics features in these learners are slow learning speed, limited interest,
speech defect, lack of muscular coordination, adjustment problem and social anxiety.
Dyslexia, dysgraphia and dyscalculia are the most common of all these learning
disabilities. 5–15% of children deal with learning disability of some form. Dyslexia
is most common LD [1]. More than being learning-disabled, these learners are
‘differently abled’.
The cases of learning disability first get the attention of the world when a psychol-
ogist Samuel Kirk coined the term ‘Learning Disability’ in Chicago conference
in 1963. Until then, these learners were misunderstood as mentally challenged. In
the literature, learning disability (LD) is also mentioned as learning difficulty and
specific learning disability. The condition of intellectual disabilities like autism, down
syndrome and attention deficit hyperactivity disorder are very much different from
LD. The learning needs are very different in these two learning conditions. Intellec-
tual disabilities are characterized based on intelligent quotient (IQ), whereas learning
disabilities are characterized on problems in academic learning skills. LD learners
have normal to high IQ. Until recent times, it is hard to diagnose due to lack of
awareness among parents and teachers and lack of proper assessment tools [2].

1.1 Learning Disabilities: Dyslexia, Dysgraphia

and Dyscalculia

Dyslexia concerns to reading-related problems. Despite having normal vision and

normal IQ learner find it difficult to read alphabets, words, sentences or paragraphs.
Learner adds, omits, substitutes or reverse letters or words with similar structure.
Frequent mistakes make learner hesitate of reading in front of class or in stage. Since
learner reads incorrectly, he learns incorrectly, which leads to frequent mistakes
during writing. Learners lack in literacy skills and phonological awareness [3].
Designing of Learning Environment of an Intelligent Tutoring System … 391

Dysgraphia concerns to writing-related problems. Learner has weak grip over

pencils or has lack of muscular coordination. Handwriting is illegible therefore
despite knowing correct answer learner fails to express in written form. These learners
show positive results in verbal examinations. Some dysgraphia learner has problem
in expression of thoughts and is weak in sentence and word expressions. Some find
it hard to organize information of pictures, graphs, maps have visual confusion of
left–right align and directions and also observed having weak visual–spatial response
[4].
Dyscalculia concerns to math-related problems. Learners have difficulty in under-
standing the concept of basic math. He may find difficulty in some of the following
problems: counting, backward counting, basic calculation like addition, subtraction,
multiply, division, telling time, geometry and reasoning. Learner gets confused with
mixed sums and finds more difficulty in understanding word problems [5].
Sometimes these learning disabilities are a result of some other intellectual
disability or learning condition. For example, a learner with hyperactive disorder
may tend to show the dysgraphia symptoms due to the effect on sensory–motor
coordination. Also, learning-disabled learners have word-finding difficulty (WFD).
It means child knows the word but is not able to use it in conversation or find it diffi-
cult to identify when shown in pictures. A learner with stuttering or social anxiety
may show similar symptoms. Assessment through professionals or special educator
is required to understand the exact condition of learner. Research has shown that
special educator can diagnose very much accurately the condition of their student
during early age.

1.2 Domain Model of an Intelligent Tutoring System

An intelligent tutoring system is a software designed with the ability to assess the
learning needs of user. It reflects the developed rules and designed curriculum in
preferential learning styles of users while adapting as per their response and activity
during learning [6].
Out of four models of an ITS: student model, domain model, pedagogy model and
user-interface model, the domain model is responsible for the resources, assessments,
inventories, instruction plan and designed environment. A learner-centric approach
based on individual skill and competency improves the learning [7, 8].

2 Objectives of Paper

The objectives of the paper are the following:

i. Study of learners to get regional quantified data on dyslexia, dysgraphia and
dyscalculia to understand academic learning needs of learners.
392 N. J. Ahuja et al.

ii. To perform a survey to identify target group in the target agency/school,

understand their clinical condition and nature of learning disability, gauge
their academic performance in a traditional learning environment followed
by a thorough need assessment and analysis phase through written and verbal
examination.
iii. Developing learner-centric learning environments for domain model of an intel-
ligent tutoring system corresponding to affect learning domains and academic
learning needs of target learners.
Here in this paper, the next section will describe the literature review section. In
the following section, a brief study was conducted to identify regional-specific target
group and to study their academic learning needs. Then, in research methodology
section, a design of four learning environment for domain model of an intelligent
tutoring system is presented. The later part of paper consists of result, discussion and
conclusion.

3 Review of Literature

Learning is a dynamic process. It depends on various factors like academic level,

prerequisite knowledge, level of interest about topic and learning style of learner. A
child having learning disability has tendency to lose focus and interest in study. In
order to teach such learner effectively, proper assessment is required. A detailed test
in the form of written and verbal examination is necessary to understand the require-
ments and problems of learners. Following are some tests taken by professional to
get detailed report of learner: Group intelligence test, individual intelligence test,
achievement test, attainment test, teacher’s/special educator’s opinion.
Singh et al. [1] did a 5-year study in Indian students mainly from English medium
schools and concluded most of the children with LD have dyslexia, dysgraphia and
dyscalculia. Most children were reported to have more than one LD. 38.56% had
ADHD along with LD [1].
Karande et al. [2] India stated SpLD as invisible handicap and suggests early
intervention for all children dealing with LD. However, for proper diagnosis of LD a
team of specialists should include a paediatrician, counsellor, clinical psychologist,
special educator and child psychiatrist [2].
Gregely. K et al. [3] in their study concluded that a paediatrician can play a crucial
role in diagnosis and therapy of LD learners. They can help parents by providing a
detailed report of problems and solutions. Parents are confused about the learning
disabilities and their impact on studies and daily life of their child. A paediatrician
along with a special educator can be helpful in the most effective14 manner [3].
Nagyova [6] emphasizes the monitoring of student’s learning activities in various
learning environments to categorize students into studious and practical types. These
learners were then provided learning material with different learning approaches [6].
Designing of Learning Environment of an Intelligent Tutoring System … 393

Brandenburger et al. [9]in their study on the preference of data visualization during
online teaching found out that learners find it more attractive if the classic form of
data is presented. It helps in reducing distraction. Colour schemes were found not
influencing the learners perception [9].
Karampa and Paraskeva [10] gave the concept for smart learning environment by
blending information and communications technology and smart pedagogy. Thus,
created SLE would be portable, affordable [10].
Nair et al. [11] emphasized on the individual and interdisciplinary early interven-
tion with focus on skills of phoneme awareness, phonics, reading comprehension,
spelling, organization of information [11].
Taur et al. [12] performed six-psychomotor computerized test on 25 learners
with LD. Finger test, reaction test, sorting tests were performed. Weak attention and
motor coordinations were reported [13]. LD learners also found it difficult to adjust
to social norms. Social anxiety, weak adjustment, low confidence are common in
these learners [12, 14].
González-Valenzuela and Martín-Ruiz [15] discussed the effect of written and
verbal interventions in LD learners. They conclude that the early, systematic and
planned interventions improve the writing skills. Learners with dysgraphia can get
benefit when remedies are provided at an early age [15].
Beckmann and Minnaert [16] reviews the 23 publications on gifted children with
learning disabilities of some form. Among these literatures, most common charac-
teristic in this learner was frustration. In addition, a need for counselling was found
for these learner due to the dual nature in cognitive abilities [16].
Gorker [2014] in their work concluded that problem in one learning domain can
affect another learning domain. A learner with comorbid learning disorder will tend
to perform poor in word problems of math due to poor reading skills. Such learner
needs different guidance than others performing weak in math [17].
Asthana and Hazela [18] explored the role of machine learning in ITS to improve
the learning environment. They propose the idea of using supervised and unsuper-
vised learning to customize learning, automate assessments and provide feedbacks
[18].
Rose et.al [19] in their work for children with learning disabilities in rural and
urban areas of southern states of India emphasize on the benefits of inclusive educa-
tion. The learning space in rural areas requires a great attention, guidance, counselling
and professional support to fulfil the learning needs [19].
Kumar et al. [20] identified types of learners as acoustic, imagistic, intuitive and
active. This is later used in their ITS designed for university students about the study
of seismic waves to adapting as per learner’s preferential learning style [20].
Russell et al. [21] of UK found that learning-disabled people tend to have weaker
physical and mental health and found it difficult to explain their condition. Learning-
disabled people find it difficult to analyse information, making decisions, making
plans, logical sequencing, processing information at normal speed, solving prob-
lems, reasoning and visuospatial reasoning. Russell & team also found that mild and
moderately learning-disabled were even not registered. Hence, recognition of such
people is necessary to improve their health [21].
394 N. J. Ahuja et al.

Chavan and Rozatkar [22] of India pointed out the drastic difference in data
regarding disabled in India. Census 2001 has data of 2.13% people (21 million
people) with disabilities, and National Sample Survey Organization (NSSO) has
1.8% people with disabilities. Also, this sample included visual, hearing, locomotor
and mental disabilities. IQ tests are used to test these learners but IQ tests have
limitations and are not helpful in finding the exact learning needs of learners [22].
A similar study by Polat et al. [23] in Turkey stated the unavailability of diagnosis
tool and direct help to such learners. Their research was focused on learners with
dyslexia, dysgraphia and dyscalculia. Their research involves children, teachers and
parents. They concluded that there is a lack of information about LD in society and
learners needs help which can be provided through a web-assisted system that can
be used inside and outside classrooms [23]. Taymans and Kosaraju [24] in a study
in USA pointed out the need of assessment and help for college students and adults
dealing with learning disabilities for better job opportunities [24].
Mehta [25] points out the tremendous stress a learner with learning disability goes
through. Parents and teachers pressurize the child for better score all the time. This
results in low self-esteem and lack of self-confidence for such learners. Also, private
treatment session costs high. Central Board of Secondary Education (CBSE) has
given provisions of extra time in examinations, quota in higher studies and exemption
of third language to these learners. This is beneficial for the needy children and is a
good step but some highly ambitious parents are reported in misuse of certification
by getting it with unethical way for their intellectually disabled children [25].
Ariffin et al. [26] studied the impact of mobile application for learners with dyscal-
culia. Seven learners with dyscalculia participated in experiment. The application
designed for Malaysian kids was found helpful and effective [26].

4 Research Methodology

Learners were observed individually through questionnaire and direct interview in the
presence of their special educators. Participants were from various cities and villages
from in-and-around Dehradun District of Uttarakhand. Phone calls and mails were
sent to schools and education centres. Survey was conducted in online and offline
mode. Learners from LKG to class 7th in various schools of state board and Central
Board of School Education (CBSE) participated in the survey. The schools were
chosen out of 21 existing schools dealing with various forms of disabilities. A series
of descriptive interviews, questionnaires and observations were taken with these
learners and their educators or guardians. Target group identified was consisted of
292 learners during initial assessment.
In detailed assessment, it was found that 98 learners out of 292 were non-LD
learners. These non-LD learners had weak academic performance but they were not
among LD. 52 learners out of 292 learners were physically challenged or had some
other intellectual disability like autism and attention deficit hyperactivity disorder
Designing of Learning Environment of an Intelligent Tutoring System … 395

(ADHD). These learners are taken into group of others and not counted as LD
learners.
Questionnaires were created based on NCERT syllabus and books after discussion
with professionals. Five questionnaires were made for learners to measure academic
learning skills related to reading, writing and mathematics: (KG), (I–II), (III–IV),
(V–VI), (VII). No IQ test was taken. To determine the dyslexia-related problem,
verbal tests and interview were done. For dysgraphia and dyscalculia written tests
were given. Questionnaire involves the following types of questions:
1. Verbal questions (to check speech-related problems like pronunciations, phono-
logical awareness, rhyming, expression of thoughts and visual confusions).
2. Written questions (to check handwriting, writing skills, calculations, series and
patterns).
3. Inventories for guardian/special educator (to check social and personal skill of
learner).
Questionnaire were given according to the educational level and prerequisite
knowledge after discussion with special educator of learner. LD-specific cognitive
strengths are taken into consideration [27]. The questions are mapped corresponding
to the nine domains of cognitive ability:
Long-term memory, short-term memory, attention, concentration, orientation,
abstraction/rational thinking, language ability, visual perception and fluency [18].
Academic learning skills and cognitive strengths are taken by us cover these nine
domains of cognitive abilities.
Language ability and fluency are overlapping with problems of dyslexia as
academic learning skills (phonological awareness and literacy skills of dyslexia)
[28] and hence not considered as cognitive strength. Performance is checked on a
scale from 1 to 5, 1 being bad, 2 being below average, 3 being average 4 being good
and 5 being very good. Question unanswered is marked as 0 (zero). X is marked for
questions not applicable according to age limit of learner. Observation sheets were
made for learners (to be filled by project team based on verbal response of learners).
Dyslexia: To find out academic learning skills related to dyslexia, i.e.
literacy skills and phonological awareness following tests were taken—identifi-
cation of letters/words, phonological awareness, reading letters/words/non-sense
words/sentences, rapid naming of colour/objects and rhyming words.
Dysgraphia: To find out academic learning skills related to dysgraphia, i.e.
handwriting, sentence and word expression, visuospatial response following tests
were taken—dictation, rewrite the sentence, jumbled words/sentence, directions and
storytelling by picture.
Dyscalculia: To find out academic learning skills related to dyscalculia, i.e.
counting, basic calculation, reasoning following tests were taken—identification of
numbers, basic sums, pattern, forward/backward series and word problems.
All questions were as per educational level of learner and were suggested by their
special educators. If learners find it very difficult to perform test of their grade level,
a lower grade-level questionnaire was provided to them.
396 N. J. Ahuja et al.

5 Results

Result section presents the result for the research objectives. It comprises of two
sections:

5.1 Research Objective 1: Identification of Specific Target

Group Through Various Assessments

Table 2 shows the list of most visible academic learning problems in written test and
verbal test of learners. Figure 1 shows the associated problems observed by survey
team while the learner was performing the test (Table 1).
Academic learning skills related to dyslexia (DL) are shown in Fig. 1 as literacy
skills (LS), phonological awareness (PA) and reading (Rdng). LS consists of the
identification and reading of letters/words/sentences and spellings PA consists of
reading non-sense words, rhyming and Rdng involves fluency and Rapid Naming.
Academic learning skills related to dysgraphia (DG) are shown in Table 3 as
handwriting (HW), visuospatial response (VSR) and sentence and word expres-
sion (SWE). HW consists of writing ability of learner to form words/sentences
during dictation and rewriting/copying the sentences. VSR consists of organizing
and outlining information from pictures, expressing their thoughts and storytelling.
SWE includes arranging jumbled words, jumbled sentences, punctuations, etc.
Academic learning skills related to dyscalculia (DC) are shown in Table 1.3 as
counting numbers (CN), basic calculation (BC) and reasoning (Rsng). CN consists of
identification of numbers and ability to count. BC consists of solving basic problems
of addition/subtraction/multiplication/division as per academic level of learner. Rsng
involves series and pattern recognition.
In Table 3, problems from the literature were identified and observed during survey
in learners. The classification of Yes and No is based on the mod of observations. If
more than 50% learners with the subtype identified with the problem, then they are
marked as Yes.

Fig. 1 Academic learning Academic learning skills and the type of LD

skills of LD learners
10
8
6
4
2
0
DL DG DC DL + DG DG + DL+ DC DL + DG
DC + DC

LS PA Rdng HW VSR SWE CN BC Rsng

Table 1 Distribution of identified learner with learning disability
Learner DL DG DC DG + DC DL + DC DL + DG + DC Non-LD Other disability Total learners
No. of learners 36 30 21 14 16 07 98 52 292
Girls 20 12 08 04 09 02 42 33 137
Boys 16 18 13 10 07 05 56 19 155
Rural 13 19 07 09 10 05 54 38 164
Urban 23 11 14 05 06 02 44 14 128
Designing of Learning Environment of an Intelligent Tutoring System …
397
398 N. J. Ahuja et al.

Table 2 Mapping of academic learning problems

Learning disability Type of questions
Dyslexia related question (verbal and written) • Literacy skill (letter identification—word
reading—letter writing—listening
comprehension)
• Phonological awareness (rhyme—phoneme
replacement), rapid naming (picture naming)
• Fluency (semantic fluency-verbal fluency)
• Language processing
• Storytelling
Dysgraphia related questions • Motor ability (grip- control- muscle
coordination), dictation (alphabets/words)
• Handwriting (font size, legibility, font shape),
rewrite the sentence
• Expression of thoughts in written form
• Visuospatial response (left–right confusion,
direction problem)
• Word or sentence formation
Dyscalculia related questions • Number sense, Counting (number system,
place value), number operations
• Basic calculation (addition, subtraction,
multiplication, division),
• Word problems and formal written numeracy
• Organizing information
Affective related questions • School adjustment
• Social anxiety
• Frustration
• Self-confidence

5.2 Research Objective 2: Developing Learner-Centric

Learning Environments for Domain Model
of an Intelligent Tutoring System

Learner’s preferred learning style differ time to time and learning in more than
one way retain the knowledge. Same content in different learning environment in
domain model can help learners in learning gain and retaining the knowledge for
long time. This impact is proposed to be measure in the future work. The learning
model used in various platforms is based on: personality, information processing,
social interaction and multidimensional factors. The most renowned learning model
are Kolb model (active/reflective/observation/experience), Felder–Silverman Model
(active/reflective, sensory/intuitive, visual/verbal, sequential/global) and VARK
model (visual, aural, read, kinesthetic) [20].
Designing of Learning Environment of an Intelligent Tutoring System … 399

Table 3 Observation grade sheet (by survey team)

S no Problems DL DG DC
i Omits a word/letter Yes No Yes
ii Substitutes a word/letter Yes No Yes
iii Adds a word/letter Yes Yes No
iv Mispronounces a word/letter Yes Yes Yes
v Spelling aloud & blending known words Yes Yes Yes
vi Spelling aloud & blending unknown words No No No
vii Ignores punctuation No No Yes
viii Weak storytelling Yes Yes No
ix Too much time to respond No No Yes
x Stutter when attempts hard No No Yes
xi Extra loudness/softness in voice Yes No Yes
xii Reading too slow/too fast No No Yes
xiii Distance between books and eyes too near No Yes Yes
xiv Ask for help repeatedly No Yes No
xv Answers only when asked each question No No No
xvi Illegible handwriting Yes Yes No
xvii Weak grip on pen/pencil No Yes No
xviii Rapid naming Yes Yes Yes
xix Colour identification Yes Yes Yes

Based on cognitive ability and psychomotor skills, four types of learning envi-
ronment are considered in this project. These intelligent learning environments are
supposed to provide content according to most suited style.
1. Case-Based Learning Environment [CBLE]: With similar ability of a Reflective
learner of Kolbe’s learning model, CBLE focuses on good memory and attention
as learner’s strength. Cognitive weakness for these learners is that they give little
time in thinking as they are quick to respond. Taking time to review and revise
will help them in retaining knowledge later. This environment will help the
learners to be part of the topic and be involved in problem. This will improve
their imagination.
2. Video-Based Learning Environment [VBLE]: With similar ability with aural and
visual learner of Felder–Silverman model or visual/auditory learner of VARK
model, VBLE focus on the audio–verbal and visual perception of learner. These
learners learn best through pictures, diagrams, videos. Here, learner is mapped
with video-based learning environment to get content in video format using
demonstration method.
3. Game-Based Learning Environment [GBLE]: With similar ability of active
learner of Kolbe’s learning model and Felder–Silverman GBLE focuses on
rational and logical thinking ability of learner. Discussion, problem-solving
400 N. J. Ahuja et al.

Table 4 Distribution of learning environments among target group

Learning DL DG DC DL + DG DG + DC DL + DC DL + DG + Total
environment DC
GBLE 2 12 8 1 3 4 1 31
CBLE 9 0 5 4 0 1 1 20
PBLE 8 17 6 8 11 4 3 57
VBLE 17 1 2 5 0 7 2 34
Total 36 30 21 18 14 16 7 142

and working with other help these learners to learn. Here, learner is supposed
to get content as a game. This helps learner to be a part of problem and learn in
fun environment.
4. Practice-Based Learning Environment [PBLE]: With similar ability of kinaes-
thetic learner of various model PBLEs focuses on practice and repetition. The
concept is similar to learning by doing. Learner is provided practice sets. For
example, learner with dysgraphia has weak psychomotor skills [29]. Their prob-
lems need to be addressed by practice and repetition. Same practice is required
for the learner having mild LD (learner with accuracy but less than average
speed).
These above learning environments are designed as a part of study of ITS devel-
oped for LD. The above-stated four LEs were developed and during the pilot study
with 142 learners. Following results were observed. Table 4 shows the learners and
their corresponding learning environments.
(a) Out of 142 LD learners, 62% had single learning disability.
(b) Maximum 57 learners were identified with PBLE.
(c) Second most common LE is VBLE.
(d) GBLE and CBLE had very low difference of three learners only.
(e) Maximum learners with DL and DL + DC find VBLE more relatable.
(f) Maximum learners with DG, DL + DG, DG + DC and DL + DC identify
with PBLE.
(g) Maximum learner with DC and DL + DC identified with GBLE.
(h) CBLE did not identify with maximum learners but still it has second-highest
number of learners in DL and DC groups.

6 Discussion

A human tutor utilizes his/her intelligence, in order to investigate the learners, in terms
of their competency levels, learning needs, and delivers instruction in accordance with
this. Human tutor engages with the learner, to tutor as per the comfort level of the
learner and has capability to seamlessly, adjust the tutoring style, to suit the learners
[30]. In order to make an ITS adaptive, a content style on the basis of learner’s
Designing of Learning Environment of an Intelligent Tutoring System … 401

academic profile and cognitive abilities is required. A domain of ITS consisting of

learning material in different styles can address user problems in best-suited style.
Researches proved that learning in more than one style maximizes the learning.
In this study, the most visible following cognitive ability groups were taken into
consideration for the developing learning environment:
i. Memory and attention,
ii. Visual perception and
iii. Rational and logical thinking (processing)
And the following academic learning skills have been considered in the proposed
work:
i. Dyslexia – Literacy Skill, Phonological Awareness, Rhyming
ii. Dysgraphia – Handwriting, Sentence and Word formation, Visuospatial
Response
iii. Dyscalculia – Counting Numbers, Basic Calculation, Reasoning
Later the study, observations and series of tests and responses of target group
help us identify with the following four learning environments for strengthening of
learner-centric domain model. Table 5 shows the learning environment mapped with
their type of learners, cognitive strengths and learning models.
Figure 2 shows the overall distribution of learning environments. PBLE was
provided to 40% of learners as first choice. VBLE scored 24% and GBLE scored
22%. CBLE had least 14% of learners as first choice. Maximum learners identified
with PBLE. The various reasons for this are illegible handwriting, slow speed and
inaccuracy in similar questions. The time taken by LD learners was more than average
learners. This makes them suitable performer for practice-based learning environ-
ment. PBLE works on accuracy and speed of learners along with writing ability [29].
Minimum learners identified with CBLE. The reason could be that CBLE is provided

Table 5 Learning environment model

Learning environment Types of learner Learning model Cognitive strength
(based on)
Case-based learning Reflective learner Reflective observant Good memory and
environment (CBLE) (Kolbe’s learning attention
model)
Video-based learning Visual learner Input (aural /visual) Good audio–verbal and
environment (VBLE) (Felder–Silverman visual perception
model)
Game-based learning Active learner Processing Good rational and
environment (GBLE) (Felder–Silverman logical thinking
model)
Practice-based learning Kinaesthetic learner VARK model To improve:
environment (PBLE) 1. Psychomotor skill
2 Speed and accuracy
402 N. J. Ahuja et al.

Fig. 2 Distribution of
learning environments with Learning Environments
LD learners

22% GBLE
24%
CBLE
14% PBLE

40% VBLE

to learners with average memory and attention span. Most of LD learners had short
attention span and weak retention of knowledge.

7 Conclusion

The study explored various combinations of learning disability, intellectual disability

and their most relatable learning environments. We conclude that learners were found
with different academic learning problems, and the effect of same learning problem
has different intensities. Some important conclusions were:
• 62% of learner had single LD while rest LD learners have multiple LD.
• Most of LD learners could not perform questionnaire of their own grade levels
and were given questionnaire of one or two less grade level on recommendation
of their special educators.
• Mispronunciations were common problems in all learners with LD.
• Most LD learners had lack of phonological awareness.
• Some LD learners were able to identify the letters/words but were unable to read.
• During reading paragraph or sentences, LD learners became stressed and started
to stutter.
• Learners with symptom of dyslexia were noticed to speak in very low voice and
were not audible.
• Many of them show disinterest for solving a word problem.
• The created learning environment based on their performance of offline and online
mode match their problems and traditional solutions and preferences.
In further studies, we are interested in in-depth evaluation of these remedial envi-
ronments. The study on improvement on various learning disabilities due to these
environments can be measured and compared with other form of learning disability.
Also, the impact on different learning skills can be measured using modified tests.

Acknowledgements This work is carried out at the University of Petroleum and Energy Studies
(UPES), Dehradun, under grant bearing reference number SEED/TIDE/133/2016. The authors
Designing of Learning Environment of an Intelligent Tutoring System … 403

thankfully acknowledge the funding support received from Science for Equity Empowerment and
Development (SEED) Division, Department of Science and Technology (DST) for the project. The
authors thank the management of the University of Petroleum and Energy Studies for supporting
the work and granting permission to publish it.

References

1. Singh S, Sawani V, Deokate M, Panchal S, Subramanyam AA, Shah HR, Kamath RM (2017).
Specific learning disability: a 5 year study from India. Int J Contemp Pediatr 4(3):863. https://
doi.org/10.18203/2349-3291.ijcp20171687
2. Karande S, Sholapurwala R, Kulkarni M (2011) Managing specific learning disability in schools
in India. Indian Pediatr 48(7):515–520. https://doi.org/10.1007/s13312-011-0090-1
3. Gergely K, Lakos R (2013) Role of pediatricians in the diagnosis and therapy of dyslexia,
dysgraphia and dyscalculia. Orv Hetil 154(6):209–218. https://doi.org/10.1556/oh.2013.29526
4. Dutt S, Ahuja NJ (2020) A novel approach of handwriting analysis for dysgraphia type diag-
nosis. Int J Adv Sci Technol 29(3):11812. http://sersc.org/journals/index.php/IJAST/article/
view/29852
5. Reisman F, Severino L (2020) Dyscalculia assessment. In: Using creativity to address dyslexia,
dysgraphia, and dyscalculia. Routledge, pp 89–120. https://doi.org/10.4324/978100303831
3-10
6. Nagyová I (2017) E-learning environment as intelligent tutoring system. AIP Conf Proc
1863(1). https://doi.org/10.1063/1.4992244
7. Bisht A, Ahuja NJ (2020) Design and development of competency-based instructional model
for instruction delivery for learning disabled using case based reasoning. Int J Rec Technol
Eng 8(6):1847–1858. https://doi.org/10.35940/ijrte.f7998.038620
8. Zydney JM, Warner Z, Angelone L (2020) Learning through experience: using design based
research to redesign protocols for blended synchronous learning environments. Comput
Edu 143. https://doi.org/10.1016/j.compedu.2019.103678
9. Brandenburger J, Constapel M, Hellbrück H, Janneck M (2020) Analysis of types, positioning
and appearance of visualizations in online teaching environments to improve learning expe-
riences. In: Advances in Intelligent Systems and Computing, vol 963. Springer Verlag, pp
355–366. https://doi.org/10.1007/978-3-030-20135-7_35
10. Karampa V, Paraskeva F (2020) Smart learning environments: a blend of ICT achievements and
smart pedagogy for the world sustainable development. In Advances in intelligent systems and
computing, vol 1018. Springer Verlag, pp 482–488. https://doi.org/10.1007/978-3-030-25629-
6_75
11. Nair MKC, Prasad C, Unni J, Bhattacharya A, Kamath SS, Dalwai S (2017) Consensus state-
ment of the Indian Academy of Pediatrics on evaluation and management of learning disability.
Indian Pediatr 54(7):574–580. https://doi.org/10.1007/s13312-017-1071-9
12. Bandla S, Mandadi GD, Bhogaraju A (2017) Specific learning disabilities and psychiatric
comorbidities in school children in South India. Indian J Psychol Med 39(1):76–82. https://
doi.org/10.4103/0253-7176.198950
13. Taur S, Karande S, Saxena AA, Gogtay NJ, Thatte UM (2014) Use of computerized tests to
evaluate psychomotor performance in children with specific learning disabilities in comparison
to normal children. Indian J Med Res 140:644–648
14. Padhy SK, Goel S, Das SS, Sarkar S, Sharma V, Panigrahi M (2016) Prevalence and patterns
of learning disabilities in school children. Indian J Pediatr 83(4):300–306. https://doi.org/10.
1007/s12098-015-1862-8
15. González-Valenzuela MJ, Martín-Ruiz I (2020) Effects of early intervention on the writing
skills of pupils who present a risk of learning disabilities within spain’s early years and
404 N. J. Ahuja et al.

primary education system. Revista de Educacion 2020(388):85–106. https://doi.org/10.4438/

1988-592X-RE-2020-388-448
16. Beckmann E, Minnaert A (2018) Non-cognitive characteristics of gifted students with learning
disabilities: an in-depth systematic review. In: Frontiers in psychology. Frontiers Media S.A.
https://doi.org/10.3389/fpsyg.2018.00504
17. Görker I (2020) The prevalence and gender differences in specific learning disorder.
In: Learning disabilities—neurological bases, clinical features and strategies of intervention.
IntechOpen. https://doi.org/10.5772/intechopen.90214
18. Asthana P, Hazela B (2020) Applications of machine learning in improving learning environ-
ment. In: Intelligent systems reference library, vol 163. Springer Science and Business Media
Deutschland GmbH, pp 417–433. https://doi.org/10.1007/978-981-13-8759-3_16
19. Rose R, Narayan J, Matam S, Reddy Sambram P (2021) A comparison of provision and access
to inclusive education for children with disabilities in a metropolitan city and a rural district in
Telangana State, India. Edu Sci 11(3). https://doi.org/10.3390/educsci11030111
20. Kumar A, Singh N, Ahuja NJ (2017) Learning styles based adaptive intelligent tutoring systems:
document analysis of articles published between 2001 and 2016. Int J Cogn Res Sci Eng Edu
5(2):83–98. https://doi.org/10.5937/IJCRSEE1702083K
21. Russell AM, Bryant L, House A (2017) Identifying people with a learning disability: an
advanced search for general practice. Br J Gen Pract 67(665):e842–e850. https://doi.org/10.
3399/bjgp17X693461
22. Chavan B, Rozatkar A (2014) Intellectual disability in India: charity to right based. Indian J
Psychiatry. Medknow Publications. https://doi.org/10.4103/0019-5545.130477
23. Polat E, Adiguzel T, Akgun OE (2012) Adaptive web-assisted learning system for students with
specific learning disabilities: a needs analysis study. Kuram ve Uygulamada Egitim Bilimleri
12(SUPPL. 4):3243–3258
24. Taymans JM, Kosaraju S (2012) Introduction to the journal of learning disabilities special
issue: adults with learning disabilities in adult education. J Learn Disabil https://doi.org/10.
1177/0022219411426860
25. Mehta M (2011) Use and abuse of specific learning disability certificates. Ind Psychiatry J
20(2):77. https://doi.org/10.4103/0972-6748.102471
26. Ariffin MM, Halim FAA, Arshad NI, Mehat M, Hashim AS (2019) Calculic kids© mobile app:
The impact on educational effectiveness of dyscalculia children. Int J Innov Technol Explor
Eng 8(8):701–705
27. Fanning E (2005) Formatting a paper-based survey questionnaire: best practices. Prac Assess
Res Eval 10(12). Available online: http://pareonline.net/getvn.asp?v=10&n=12
28. Rao C, Sumathi TA, Midha R, Oberoi G, Kar B, Khan M, Singh NC (2021) Development
and standardization of the DALI-DAB (dyslexia assessment for languages of India—dyslexia
assessment battery). Ann Dyslexia. https://doi.org/10.1007/s11881-021-00227-z
29. O’Connor ES, Mahvi DM, Foley EF, Lund D, McDonald R (2010) Developing a practice-based
learning and improvement curriculum for an academic general surgery residency. J Am Coll
Surg 210(4):411–417. https://doi.org/10.1016/j.jamcollsurg.2010.01.017
30. Felder RM, Soloman BA (2017) Felder & Soloman: learning styles and strategies. Retrieved
from http://www4.ncsu.edu/unity/lockers/users/f/felder/public/ILSdir/styles.htm
Trusted 6G-Envisioned Dynamic
Spectrum Allocation: A Reference
Model, Layered Architecture,
and Use-Case

Pronaya Bhattacharya, Farnazbanu Patel, Vishaka Ralegankar,

Bhaumik Thakkar, Sudeep Tanwar, and Mohammad Abouhawwash

Abstract Spectrum allocation among multiple telecom providers is challenged with

a fair spectrum allocation process and collusion among multiple telecom parties
involved in spectrum bids. As spectrum licensing have shifted toward decentral-
ization, blockchain-based spectrum allocation can address the limitations through
a fair and trusted bidding process, and spectrum allocation which is transparent to
all bidders. Moreover, the spectrum allocation process has shifted to dynamic spec-
trum allocations, and thus in the coming future, owing to the inherent advantages of
sixth-generation (6G) networks, with high network reliability, user bandwidth, high-
precision, ultra-high reliability, and extreme flexibility, all presented as networked-
in-a-box service, dynamic spectrum allocation in 6G-envisioned networks is a reality
in near future. In light of these facts, we present in this paper, a comprehensive exam-
ination of the integration of a blockchain-based scheme for dynamic spectrum access
in 6G-envisioned communications. We present the layered reference architecture and
highlight a case study of future blockchain-based spectrum access for 6G-serviced
network access. The paper serves as starting foundation toward the build of effective
blockchain-based spectrum allocation schemes with effective networked serviced
applications.

P. Bhattacharya (B) · F. Patel · V. Ralegankar · B. Thakkar · S. Tanwar

Institute of Technology, Nirma University, Ahmedabad, Gujarat, India
e-mail: pronoya.bhattacharya@nirmauni.ac.in
F. Patel
e-mail: 19mcei04@nirmauni.ac.in
V. Ralegankar
e-mail: 20mced18@nirmauni.ac.in
B. Thakkar
e-mail: 20mcec17@nirmauni.ac.in
S. Tanwar
e-mail: sudeep.tanwar@nirmauni.ac.in
M. Abouhawwash
Durban University of Technology, Durban, South Africa
e-mail: abouhaww@msu.edu

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 405
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_30
406 P. Bhattacharya et al.

Keywords 6G · Blockchain · Spectrum sharing · Decentralization ·

Telecommunication industry

1 Introduction

Over the past decade, there has been a surge in telecommunications networks that
have surged the requirements of spectrum allocation bands among telecom service
providers. Currently, with a large number of users, fourth-generation (4G) long-term
evolution (LTE) networks are facing bottlenecks to service the growing demands.
By 2021, 4.5 billion mobile subscribers are registered globally. Figure 1 presents the
scenario. Owing to the shift, the telecom industries have shifted toward spectrum
licensing in the fifth generation (5G) bands. 5G offers effective service orchestration
through a combination of different band frequencies to increase the coverage range.
5G commercial networks are expected to operate in the 3.3–3.8 gigahertz (GHz)
range, with support of lower bands that include 1500 megahertz (MHz), 2.1 GHz,
and 2.3 GHz for poor connection areas [2]. Thus, 5G is expected to provide faster
and reliable network services that would support different verticals in smart cities, like
smart factories, smart and autonomous vehicles, and healthcare industries. However,
it also requires telecom providers to access higher-frequency bands to make the vision
a reality.
However, spectrum frequencies are limited resources, and thus, an effective shar-
ing mechanism is required. With the advent of a shift of network services at the

4.5

4
Smartphone users worldwide (in billion)

3.5

2.5

1.5

0.5

0
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Progressive Years

Fig. 1 Global increase of mobile users [1]

Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 407

edge [3], latency in networked applications has also become a prime requirement.
5G services like ultra-reliable low-latency communications (eMBB) and massive
machine-type communications (mMTC) offer an end-to-end latency of 5 ms and
high connection density of 1 million devices/km2 . However, with the rise of automa-
tion, and an increase in massive device-to-device (D2D) connectivity in Internet-
of-anything (IoE) ecosystems, networks would require extreme dense connections,
edge intelligence support, and high reliability.
Thus, researchers have shifted toward sixth-generation (6G) networks, that is
envisioned to support ultra-high data rates in the range of terahertz (THz) range,
radio latency of 100 µs, and connection density of 107 devices/km2 [5]. 6G ser-
vices can be easily stacked to support the spectrum access. Moreover, 6G fine-tunes
the edge computing requirements through artificial intelligence (AI)-enabled radio
access, and thus, industries have shifted toward investment in 6G projects [6]. 6G
supports effective features like omnipresent global coverage in space–air–ground–
water communication, at ultra-high reliability of 99.9999999 %. 6G is expected to
support verticals like holographic and 3D integrations as well [7]. Table 1 presents
an overview of the mobile communication shift from second generation (2G) com-
munication to 6G.
In terms of application viewpoint, 6G would support low-rate and long-distance
Internet-of-things (IoT) applications, process automation of cyber-physical systems
in the industry, digital twins, holography, AI support with the complex machine and
deep learning models, extended, virtual, and mixed reality applications, and auto-
matic vehicular networks. Owing to the AI-enabled radio, it resolves the issues of
fixed spectrum allocation in decentralized environments and covers for underutilized
spectrum limitations. However, static spectrum allocation is mostly centralized, and
thus, blockchain is a viable choice of fair spectrum allocations owing to the inher-
ent benefits of fairness, immutability, and chronological access [8]. Moreover, in
spectrum allocation, we consider a permissioned blockchain, where government,
telecom providers, and spectrum licensing stakeholders are allowed to participate in
the bidding process.

1.1 Research Contributions

Following are the research contributions of the paper.

• A reference model of 6G-envisioned blockchain-based spectrum allocation is pre-
sented, and a layered stack reference is proposed.
• A case application based on the proposed reference scheme is presented that
discusses a reputation-based scorecard registration of new user in the proposed
ecosystem.
 408

Table 1 An overview of mobile communications shift from 2G to 6G [4]

Generation 2G 3G 4G 5G 6G
Year 1990 2000 2010 2018 2030
Technology DBN-CD DBB PD (IPv4) LAN/WAN, Unified LAN/WAN/PAN, Unified Blockchain-based spectrum,
IPv4/IPv6, WLAN IPv4, WLAN, advanced artificial intelligence, quantum
OFDM communication, laser and
VLC, THz communication
Technique SISO SISO MIMO m-MIMO SM-MIMO, UM-MIMO
Frequency 850–1900 MHz 1.6–2.5 GHz 2–8 GHz 3–300 GHz 95 GHz–3 THz
Spectral efficiency 0.17 bps/Hz 1.3 bps/Hz 30 bps/Hz 120 bps/Hz 5–10 times that of 5G
Latency 300–1000 ms 100–500 ms < 15 ms < 100 ms < 10−100 µs
Connection density Limited 100/km2 1000/km2 1,000,000/km 2 10,000,000/km 2
Mobility – – 350 kmph 500 kmph 1000 kmph
P. Bhattacharya et al.
Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 409

1.2 Article Structure

This paper is divided into five sections. Section 2 presents the state-of-the-art schemes
that are proposed related to 6G and blockchain-based schemes. Section 3 presents
the layered reference model of 6G-envisioned blockchain-assisted dynamic spectrum
allocation, which is supported by a layered reference stack architecture in Sect. 4.
Section 5 presents the case-study of the proposed scheme, and finally Sect. 6 con-
cludes the paper.

2 State of the Art

In this section, we present the recent state-of-the-art schemes that integrate blockchain
and 6G in telecommunications. Saravanan et al. [9] proposed the integration of
blockchain for telecom providers to simplify their phone usage charging and billing
operations. Via blockchain, the third-party intermediaries are removed, and inconsis-
tencies in the management of large customer databases are simplified. The paper pro-
poses that blockchain ledger can manage user call records in an immutable manner,
and through smart contracts, roaming agreements between inter-telecom providers
are also managed, and balance transfers are automated. This reduces the overall trans-
actional fees of third-party payment gateways and improves the complexity of the
overall billing ecosystem. Xu et al. [10] proposed a resource management scheme for
spectrum allocation for mobile operators and presented a reference framework that
manages resources and sharing in 6G-IoE ecosystems. The authors proposed a net-
work slicing-based approach in 6G, and a slice-broker-based scheme to manage the
6G resource orchestration. The resource transfer is managed as transactional ledgers
in the blockchain. Zhou et al. [11] presented a privacy-preserved 5G human-to-human
(H2H), and machine-to-machine (M2M) scheme, where a cost-effective solution is
presented to optimally utilize the spectrum resources. The paper introduces a two-
phased scheme. In the first phase, H2H users and 5G-enabled base stations execute
a smart contract for transactional payments, and spectrum is released. The spectrum
is allocated to M2M devices, with an incentive-based design.
Zhang et al. [12] proposed a distributed citizens broadband radio access (CBRS)
spectrum sharing scheme to address the limitations of administrative costs, and
privacy-based attack scenarios by an adversary. The authors include a low-powered
consensus mechanism known as proof-of-strategy that finalizes the spectrum alloca-
tion, even in case of node failures. Patel et al. [14] proposed a 6G-based blockchain-
based spectrum allocation scheme between dynamic service operations in a cell-free
spectrum. The paper proposes a dynamic auction and bidding process of spectrum
allocation. Hewa et al. [13] proposed a survey that introduces blockchain potential in
6G verticals such as health care, Internet-of-vehicles, infotainment, augmented and
virtual reality, and M2M communication. The challenges of 6G and potential pitfalls
are identified, and blockchain-based solutions are proposed to allow distributed 6G
protocols and standards.
410 P. Bhattacharya et al.

Table 2 State-of-the-art approaches of integration of blockchain and 6G in telecommunications

References Year Major contributions Application domain
[9] 2017 Blockchain-based consumer 5G-based telecom service
balance transfers, carrier billing,
roaming settlements through
decentralized application
deployments
[10] 2020 The complexities of adapting 6G-envisioned spectrum sharing
blockchain for spectrum sharing in and access and blockchain for
6G are discussed license transfers
[11] 2020 H2H and M2M enabled spectrum Fair spectrum sharing through
sharing ecosystem, with blockchain-based 5G dynamic
blockchain-based transactional spectrum access
ledgers
[12] 2020 Optimal CBRS design service 6G-CBRS allocation scheme with
model through blockchain and trusted blockchain ledgers
6G-radio access model
[13] 2020 6G technical key performance 6G-benefits in IoT, industry 4.0,
indicators (KPIs) compared virtual reality, and autonomous
against 5G KPIs in terms of driving
specification, regulation,
standardization, and design in
different driving verticals
[14] 2021 In 6G ecosystems, 6G-based dynamic spectrum
blockchain-enabled trusted allocation and bidding process via
dynamic frequency spectrum blockchain-based auctions
allocation and auction scheme
among telecom providers
[5] 2021 6G-benefits and reference 6G-IoE standards, interfaces, and
architecture in telecom servicing, protocol regularization via
IoE, and inherent limitations, with blockchain
architectures of blockchain to
mitigate the limitations

Jiang et al. [5] proposed different 6G frontiers in different verticals of smart cities
and discussed the requirement of 6G to handle a high volume of data traffic. Potential
use cases and scenarios are discussed, and a tentative roadmap of 6G standardization
is presented. The details of the recent schemes, their contributions, and application
domains are discussed in Table 2.

3 A Reference Model of Blockchain-Based

6G-Spectrum Allocation

In this section, we present the founding concepts of blockchain and 6G services.

The section presents a reference architecture that discusses the potential benefits of
Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 411

blockchain in 6G applications that handle the issues of trust, privacy, and secure
transfer of resources among communicating entities. We start initially with the dis-
cussion of 6G emergence and then move toward the usage of blockchain to support
secured 6G services. The details are presented as follows.

3.1 The Emergence of 6G and Blockchain

in Communications

With the increase of communication networks, and stringent requirements of band-

width, latency, and availability of resources, to support applications like extended
reality, autonomous driving, Internet-of-drones, real-time sensing and control, an
eight-year program, termed as 6Genesis Flagship started with an estimated fund of
290 million dollars. The project started in 2018 by Finland, and soon researchers
worldwide started with the design of protocols and standards for 6G communication
networks. The key visions of 6G communication highlighted in Table 3.
Initially, started as cryptographic ledgers [15], blockchain gained prominence
owing to its inherent benefits of trust, immutability in block creation, and verifi-
cation and thus has become a driving force in different verticals like smart grids,
autonomous vehicles, and Internet-of-things [16]. To secure the 6G connectivity
perimeter, blockchain can mitigate security attacks like distributed denial-of-service,
impersonation, replay, and certificate-based attacks [13]. Thus, blockchains empower
decentralized cooperative applications and also ensure that data is exchanged by all
parties involved.

Table 3 Visions of 6G communication [4]

Key parameters 6G vision
Mobility (km/h) 1000
Peak spectral efficiency (b/s/Hz) 60
End-to-end latency (ms) 0.1
Reliability 10−9
Connection density (device/km2 ) 107
Area traffic capacity (Mbps/m2 ) 1000
Channel bandwidth (GHz) 100
Spectral efficiency (b/s/Hz) 3
Energy efficiency (Tb/J) 1
User data rate (Gbps) >10 Gb/s
Peak data rate >100 Gb/s
Receiver sensitivity < −130 dBm
Coverage >99%
End-to-end delay <1 ms
412 P. Bhattacharya et al.

3.2 A Proposed Reference Model

In this subsection, we present the reference model of blockchain-based 6G-spectrum

access. Figure 2 presents the details of the proposed model. In the proposed model,
we consider entities E = {E BS , E SMS , E AC , E SL , E TP , E SB , E SV , E SR }, where E BS
denotes the base stations (BSs) that are integrated with 6G services to support

Fig. 2 A reference model of

blockchain-based
6G-spectrum access
Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 413

dynamic spectrum access. E SMS denotes the spectrum management server, E AC

denotes the spectrum auctioneer, E SL denotes the spectrum leaser, E TP denotes
the telecom provider, E SB denotes the spectrum borrower, E SV denotes the spec-
trum validators, and E SR denotes the spectrum requester, respectively. Depending
on the specifics of spectrum resource allocations, the utilization of spectrum is only
done post the auction process. To maintain the regulations and control in spectrum
allocation, we consider E SRMS that denotes the spectrum radio monitoring server,
which is a government regulating body to manage and distribute the spectrum to
E TP , E RS , E BS . As the considered ecosystem is a multi-party decentralized system
with different stakeholders like industrial applications, spectrum auctioneers, and
buyers, borrowers, spectrum leases, and spectrum advertisers, we require trust in the
ecosystem. For the same, we consider a consortium-based permissioned approach,
where the transaction updates are shared only by registered stakeholders in the chain.
For the spectrum ledger, we consider the ledger L with the required fields, namely
{AO, AI, SMod, FA, WU }, where AO denotes the asset (spectrum resource) owner-
ship, AI denotes the asset meta-information, SMod is defined as the sharing model
(competitive or collaborative), FA denotes the frequency allocated, and WU denotes
the wallet information of the user. The transaction ledgers are maintained through
distributed offline storage (interplanetary file systems), where the ledger records are
accessible by the IPFS key and the private key of the user only. We hash the stored
record, H (R), and store the record R indexed with its hash-pair H (R). In the main
chain, we store H (R) only as a transaction, so that the record R may be retrieved
by a search of the hash in the chain. Moreover, effective consensus schemes are
required to be designed for E SV , so that their incentives are maximized. Validators
E SV are chosen based on a reputation score R so that they add the transactions in a
fair manner in the blockchain [17].
In the reference architecture, we consider servicing E BS that provides network
service to user sets U = {U1 , U2 , . . . , Un }. We consider a cell-based 6G grant spec-
trum access scheme, and two regions, R1 , and R2 , respectively. Any nth user in region
R1 is mapped to E BS through a mapping M1 : Un → BS1n , and similarly, any user
Un in R2 is mapped to BS2n , through mapping M2 : Un → BS2n . The user requests
R = {R1 , R2 , . . . , Rn } are collected region-wise and send to E SMS through directed
6G uplink frequency f u . At E SMS , the collected requests R are serviced as digital
assets, and spectrum allocation requirements are advertised in the network, termed
as spectrum advertisements. E SMS handles the function of intelligent spectrum sens-
ing through AI models and maintains spectrum access historical ledger entries by
corresponding E BS . The spectrum auction A(S) is initiated at E SMS depending on
the base network requirements sent by E BS . For the same, a list of freely available
frequency F( f r ) is maintained, which is collected through network entities like IoE
networks, satellites, free users, vehicular networks, and others. The finalization and
broadcast of available spectrum bands are termed spectrum leasing. For the leasing
process, an auction strategy is set up between spectrum bidders and spectrum borrow-
ers, with the peer-profit optimization strategy. The auction can be modeled through
a cooperative game-theoretic approach, to maximize the incentives of both bidders
and borrowers, through the designated set of auctioneers. Once the spectrum auction
414 P. Bhattacharya et al.

process is over, the spectrum grant is maintained as transactional ledgers in IPFS

and meta-information are chronologically recorded in the blockchain. To spectrum
transaction information is maintained in consortium blockchain, and the available
usage and regulations are reflected all authorized nodes in the chain by E SRMS . This
ensures the transparency of spectrum allocation to all E TP and mitigates the possible
collusion among malicious bidder nodes.

4 The Layered Reference Stack of Blockchain-Based

6G-Spectrum Allocation Scheme

In this section, we propose the layered stack model of the proposed reference archi-
tecture that handles the issues of static spectrum allocation. Figure 3 presents the
details. We consider a four-layered scheme, and the details are presented as follows.

Fig. 3 Spectrum allocation using blockchain

Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 415

4.1 Layer 0: The Spectrum Layer

At Layer 0, we assume the spectrum details are present, which is a cluster of fre-
quency ranges R( f ), and consist of electromagnetic waves. Through R( f ), different
communication devices such as TV, radio, mobile to send wireless messages across
a certain distance d. The details of available spectrum bands are managed by E SMS ,
and the allocation of bands to different servicing E BS , of different E TP is leveraged
through a spectrum validator E SV . The spectrum band is mainly divided into three
regions as follows.
• Licensed: In the licensed band, a chunk of the radio spectrum is assigned to E SMS ,
or E SRMS , and is licensed as asset ownership by AO. Any user has to send a
spectrum access request to AO, and the spectrum grant is defined for a definite
time period T . Here, the access request is placed by E TP , so they buy the licensed
frequency ranges from F( f r ), for a given price, and allocate frequencies to E BS
through the servicing downlink Fd .
• Unlicensed: In this band, the available frequencies can be used by any user, and
normal users also have access to the unlicensed spectrum. This type of spectrum
does not involve a specific type of permission from either E SMS , or E SRMS . The
applications of the unlicensed spectrum are IEEE 802.11 access, TV white spaces,
and wireless personal area networks, like IEEE 802.15.x.
• Shared: In this band, the frequencies are shared among different users, and each
user utilizes a chunk of the frequency band. This type of paradigms helps the users
and devices to completely utilize the spectrum band.

4.2 Layer 1: The Authority Layer

The shared spectrum suffers from a lot of obstacles. Generally, the practice involves
the centralization of shared spectrum management (by CBRS) [18]. Here, an inter-
mediary is needed to manage the complete flow of control in the shared environment.
The centralized systems suffer from various issues such as lack of adaptability, over-
burdening on the central authority, one-sided communication, and biases in decision
making [14]. These issues lead to poor utilization of resources and a less secure sys-
tem. In the proposed scheme, the authority layer validates the authority of the users
in the ecosystem. For consensus, we consider a modified version of the Proof-of-
Authority (PoA) consensus mechanism. The primary PoA works by allowing nodes
to create initial blocks that have demonstrated their authority. Any new user in PoA
has to prove the identity to get access to the spectrum. Once the identity authentica-
tion gets done, a scorecard is generated for the user. This process is iterated for each
user in the network to prove the genuineness of the identity of the users.
416 P. Bhattacharya et al.

4.3 Layer 2: The Contract Layer

To break the tie created by a centralized environment, the user needs a mechanism
that can automate the flow of taking decisions in a very honest manner. For that, we
use smart contacts. Smart contracts are self-executing code without any third-party
(such as humans) interactions. In the proposed model, smart contracts ensure the
storage of authorized user data published on IPFS, and meta-information stored in
distributed ledgers. The access of IPFS is restricted through identity authorization
and IPFS key.

4.4 Layer 3: The Decentralization Layer

At Layer 3, we consider the distributed blockchain ledger. New blocks are added
only after E SV validates the transaction entries. Every authorized user has a copy
of ledger L, and L gets updated once the state of IPFS changes, to reflect new
contracts executed in the network. Through 6G, ease of access and scalability of
node communication are improved.

5 Case Study: A Scorecard and Reputation-Based

Spectrum Allocation

In this section, we propose a case study that presents the usage of the shared spectrum.
Figure 4 presents the details. The shared spectrum can be allocated to the user using
the integrated technology discussed in Sect. 4.
In the use-case, we consider entity A1 that wishes to access the joint spectrum
for communication purposes. A1 first registers himself in the network and has to
undergo the PoA consensus where E SRMS , or E SMS validates A1 identity to all users.
Then, A1 is granted access to spectrum resources. This whole registration process
is automated via a DApp that executes a smart contract at the back-end between E U
and E SMS and publishes the transactional state to IPFS. Also, other E U ledgers are
updated with the new entry in their ledgers. Here, A1 is presented with a scorecard,
and based on future transactions performed by A1 , the reputation score increases,
and the access-grant time of shared spectrum is reserved for A1 also increases. This
reward-based technique ensures the authenticity of the users is managed in real-time
through 6G service sets.
Trusted 6G-Envisioned Dynamic Spectrum Allocation: A Reference … 417

Fig. 4 Spectrum allocation using blockchain

6 Conclusion

The spectrum allocation process among competitive telecom providers and users is a
complex problem. The problem is further intensified in decentralized environments
owing to the issues of trust, alterations, and collusion-based attacks. Thus, in this
paper, we have presented a reference model for blockchain-assisted dynamic spec-
trum access at the backdrop of 6G-envisioned communications. Through blockchain,
a trusted chronology is maintained among distributed telecom stakeholders, and
provenance is established. Owing to the high influx of network traffic, and users, 5G
services would face bottlenecks in the near future. Due to this, we considered a 6G
service set that provides intelligent and real-time network orchestration to users in
the proposed ecosystem. A reference model is presented, and a supportive layered
stack model is also proposed. Then, we present a reputation-based scorecard for reg-
istration of new users in the ecosystem that ensures the genuineness and transparency
via PoA consensus in the spectrum allocation ecosystem.
As part of the future scope, the authors would investigate a deep reinforcement
learning framework that manages the reputation of a user in the ecosystem and
also would propose a cooperative game-theoretic approach to model and maximize
incentives of the auction process.
418 P. Bhattacharya et al.

References

1. Statista (2021) Number of smartphone users worldwide from 2012 to 2023. https://www.
statista.com/statistics/330695/number-of-smartphone-users-worldwide/, May 2021
2. Gorla P, Chamola V, Hassija V, Ansari N (2021) Blockchain based framework for modeling
and evaluating 5g spectrum sharing. IEEE Network 35(2):229–235
3. Bhattacharya P, Tanwar S, Shah R, Ladha A (2020) Mobile edge computing-enabled blockchain
framework—a survey. In: Singh PK, Kar AK, Singh Y, Kolekar MH, Tanwar S (eds) Proceed-
ings of ICRIC 2019. Springer International Publishing, Cham, pp 797–809
4. Alsharif MH, Kelechi AH, Albreem MA, Chaudhry SA, Zia MS, Kim S (2020) Sixth gen-
eration (6g) wireless networks: vision, research activities, challenges and potential solutions.
Symmetry 12(4)
5. Jiang W, Han B, Habibi MA, Schotten HD (2021) The road towards 6g: a comprehensive
survey. IEEE Open J Commun Soc
6. Huang T, Yang W, Wu J, Ma J, Zhang X, Zhang D (2019) A survey on green 6g network:
architecture and technologies. IEEE Access 7:175758–175768
7. David K, Elmirghani J, Haas H, You X-H (2019) Defining 6g: challenges and opportunities
[from the guest editors]. IEEE Veh Technol Mag 14(3):14–16
8. Patel SB, Bhattacharya P, Tanwar S, Kumar N (2021) Kirti: a blockchain-based credit recom-
mender system for financial institutions. IEEE Trans Network Sci Eng 8(2):1044–1054
9. Saravanan M, Behera S, Iyer V (2017) Smart contracts in mobile telecom networks. In:
2017 23RD Annual international conference in advanced computing and communications
(ADCOM). IEEE, pp 27–33
10. Xu H, Klaine PV, Onireti O, Cao B, Imran M, Zhang L (2020) Blockchain-enabled resource
management and sharing for 6g communications. Digit Commun Networks 6(3):261–269
11. Zhou Z, Chen X, Zhang Y, Mumtaz S (2020) Blockchain-empowered secure spectrum sharing
for 5g heterogeneous networks. IEEE Network 34(1):24–31
12. Zhang H, Leng S, Chai H (2020) A blockchain enhanced dynamic spectrum sharing model
based on proof-of-strategy. In: ICC 2020-2020 IEEE International conference on communica-
tions (ICC). IEEE, pp 1–6
13. Hewa T, Gür G, Kalla A, Ylianttila M, Bracken A, Liyanage M (2020) The role of blockchain
in 6g: challenges, opportunities and research directions. In: 2020 2nd 6G Wireless summit (6G
SUMMIT). IEEE, pp 1–5
14. Patel F, Bhattacharya P, Tanwar S, Gupta R, Kumar N, Guizani M (2021) Block6tel: blockchain-
based spectrum allocation scheme in 6g-envisioned communications. In: 2021 International
wireless communications and mobile computing (IWCMC), pp 1823–1828
15. Srivastava A, Bhattacharya P, Singh A, Mathur A, Pradesh U, Pradesh U (2018) A systematic
review on evolution of blockchain generations. Int J Inf Technol Electr Eng 7(6):1–8
16. Bhattacharya P, Tanwar S, Bodkhe U, Kumar A, Kumar N (2021) Evblocks: a blockchain-based
secure energy trading scheme for electric vehicles underlying 5g-v2x ecosystems. Wirel Pers
Commun, pp 1–41
17. Shyamsukha S, Bhattacharya P, Patel F, Tanwar S, Gupta R, Pricop E (2021) PoRF: proof-of-
reputation-based consensus scheme for fair transaction ordering. In: 2021 13th International
conference on electronics, computers and artificial intelligence (ECAI), pp 1–6
18. Zafaruddin SM, Bistritz I, Leshem A, Niyato D (2019) Distributed learning for channel allo-
cation over a shared spectrum. IEEE J Sel Areas Commun 37:2337–2349
Experimental Evaluation of QFT Adders
on IBM QX Hardware

Simran Jakhodia, Divyanshu Singh, and Babita Jajodia

Abstract Quantum Fourier transform (QFT) offers a versatile way to perform mod-
ular and non-modular quantum arithmetics and provide compact quantum circuits.
QFT is the quantum implementation of the classical discrete Fourier transform (DFT)
over the vector of amplitudes of a quantum state (wavefunction). In this paper, the
authors experimentally evaluate the performance of QFT adders on various IBM
Quantum Experience (QX) hardware against Quantum Information Science Kit
(Qiskit) Simulator. The authors have constructed quantum circuits to represent one-
qubit and two-qubit QFT adders using Qiskit and run on IBM QASM Simulator and
eight IBM real quantum devices: YorkTown, Santiago, Athens, Quito, Belem, Lima,
Manila and Melbourne. Experimental analysis of one-qubit QFT adder shows that
IBM Lima and IBM Athens provide better accuracy results of 80.2% and 80.125%,
respectively, in comparison to other five-qubit and fifteen-qubit quantum hardware.
Experimental results of two-qubit QFT adder demonstrate that IBM Athens and
IBM Manila provide better accuracy results of 25.575% and 17.52%, respectively,
in comparison to other five-qubit and fifteen-qubit quantum hardware.

Keywords IBM · Quantum computing · Quantum adder · Quantum gates ·

Quantum information science kit (Qiskit) · Quantum Fourier transform (QFT) ·
Inverse quantum Fourier transform (IQFT)

S. Jakhodia · B. Jajodia (B)

Indian Institute of Information Technology Guwahati, Bongora, Guwahati, Assam, India
e-mail: babita@iiitg.ac.in
S. Jakhodia
e-mail: simran.jakhodia@iiitg.ac.in
D. Singh
Gautam Buddha University, Greater Noida, Uttar Pradesh, India
e-mail: 19bhp011@gbu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 419
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_31
420 S. Jakhodia et al.

1 Introduction

Quantum computing along with its inherent quantum principles of superposition and
entanglement are believed to bring revolutionary changes overcoming the limita-
tions of classical computing systems [1]. Constant advancements are being made,
focussing on developments of higher-qubit real quantum systems.
Researchers from academia and research companies long predicted that future
developments in scientific computing would be enabled by the advent of software-
programmable quantum computers and solve real-world problems that are NP-
hardness (non-deterministic polynomial-time hardness) and not so easily solvable
by classical computers.
Quantum addition is the fundamental of quantum arithmetics, and its use cases
are in quantum cryptography, quantum modular arithmetics and multiplications. It
forms the basis of a majority of high-computational arithmetic computations. Shaik
and Rangaswamy [2] explore implementation of quantum addition circuits using
conventional gate-based approach [1, 3], but at the cost of higher number of qubits.
Beauregard [4] proposed a generalized circuit of Quantum Shor Algorithm [5] using
quantum addition circuit depicted by Draper [6]. Here, in [6], quantum Fourier trans-
form (QFT) [7], with exponential speed-up played an important part in implemen-
tation of quantum addition circuits [8]. In brief, QFT transforms the state of qubit
from computational basis to Fourier basis [9]. Therefore, in this work, the authors
have put an attempt to propose n-qubit QFT addition circuits with a minimum of
(2n + 1) qubits only, taking an example of one-qubit and two-qubit quantum addition
for easy illustration. An experimental evaluation study of one-qubit and two-qubit
QFT adders [7] will be discussed by its execution on real IBM quantum devices.
Later, this work infers and derives accuracy rates of quantum circuits on various
IBM real quantum systems and draw conclusions based on the performance. These
contributions are made so that this work help other researchers to improvise upon
this using real quantum computers to obtain better results.
The paper is organized as follows: Sect. 2 provides a brief background to QFT and
inverse QFT (IQFT), arbitrary single-qubit and basic quantum gates that will be used
in the development of QFT adders [10]. Section 3 discusses the proposed work on
QFT adder taking an example of one-qubit and two-qubit quantum addition for better
illustration. Section 4 discusses the experimental evaluation study of executing an
illustrative one-qubit and two-qubit QFT adder on various IBM quantum experience
(IBM QX) [11]. This is followed by conclusions and future works in Sect. 5.
Experimental Evaluation of QFT Adders on IBM QX Hardware 421

2 Background

2.1 QFT and IQFT [1]

Quantum Fourier Transform (QFT) [7, 9] is the fundamental building block of various
quantum algorithms such as for computing discrete logarithms, Shor’s algorithm for
factoring which is used in quantum cryptography, the quantum phase estimation
algorithm which is used for estimating eigen values of a unitary operator. QFT is the
quantum implementation of the discrete Fourier transform (DFT) over the amplitudes
of a wave function [1]. DFT acts on vectors (x0 , x1 , x2 , . . . , x N −1 ) and maps it to the
vector (y0 , y1 , y2 , . . . , y N −1 ) as

N −1
1 
yk = √ xn ω−kn
N , (1)
N n=0

where k = 0, 1, 2, . . . , N − 1 and ω = e2πi/N . Similarly, the QFT operator can be

defined as
N −1
1  i.2π xk
QFT |x = √ e N |k (2)
N k=0

considering a N − 1 dimensional system with states |x from the computational basis
|0 , |1 , . . . , |N − 1. The QFT helps to encode a number x in the relative phases
of the states of a uniform superposition consisting in the sum of all the states |k in
the computational basis, with the amplitude of √1N . It can also be stated that QFT
transforms states in the computational (Z ) basis to the Fourier basis. Applying QFT
on |0 and |1 in Z basis, the obtained states are |+ and |−, respectively, in X basis.
The IQFT can also be defined as

N −1
1  −i.2π xk
IQFT |k = √ e N |x . (3)
N x=0

The quantum circuit implementation of QFT and IQFT consists of Hadamard (H )

gates and controlled U1 gates [12]. The Hadamard (H ) gate when applied to a qubit
in a circuit it puts it in to a superposition of states such that when it is measured,
it could be 0 or 1 with equal probability. The controlled U1 (CU1 ) gate implements
rotation angle around the Z -axis (phase) of the target qubit if the control qubit is
1. The matrix representations of basic quantum gates (H , X and CU1 ) are given in
Sect. 2.3.
422 S. Jakhodia et al.

2.2 Arbitrary Single-Qubit Quantum Gates

An arbitrary single-qubit quantum gate, U3 (θ, φ, λ) can be implemented by setting

three parameters, θ , φ and λ, and the matrix representation is
 
cos( θ ) e−iλ sin( θ2 )
U3 (θ, φ, λ) = iφ 2 θ i(λ+φ) (4)
e sin( 2 ) e cos( θ2 )

Here, the three constraints for θ , φ and λ are 0 ≤ θ ≤ π , 0 ≤ φ < 2π and 0 ≤ λ <
2π , respectively. This is the most general form of a single-qubit unitary.
U2 and U1 are two additional single-qubit gates and are special cases of U3 gates.
The matrix representations of U2 and U1 are
π   
1 1 eiλ
U2 (φ, λ) = U3 , φ, λ = √ iφ i(λ+φ) (5)
2 2 e e
 
1 0
U1 (λ) = U3 (0, 0, λ) = (6)
0 eiλ

2.3 Quantum Gates

The matrix representations of various quantum gates using the computational basis
of |0 and |1 used in the development of QFT adders are:
H Gate (which puts qubits in superposition state)
 
1 1 1
H = U2 (0, π ) = √ H (7)
2 1 −1

X Gate (which flips the state of qubits)

 
01
X = U3 (π, 0, π ) = X (8)
10

Controlled U1 (CU1 ) gate

⎛ ⎞
1 0 0 0
⎜0 1 0 0⎟
CU1 = ⎜
⎝0
⎟ (9)
0 1 0⎠
0 0 0 eiφ
Experimental Evaluation of QFT Adders on IBM QX Hardware 423

3 Proposed Work on QFT Adder Circuits

This section illustrates the design and implementation of n-qubit QFT adders using
(2n + 1) qubits only taking two examples of one-qubit and two-qubit additions.

3.1 One-Qubit QFT Adder

Figure 1 illustrates the quantum implementation of one-qubit QFT adder. Here, a0

and b0 are the two inputs; the output qubits are a1 and a0 that can be considered as
sum qubits, respectively. Table 1 shows the truth table of one-qubit QFT adder for
every possible combinations of inputs.

3.2 Two-Qubit QFT Adder

Figure 2 shows the quantum implementation of two-qubit QFT adder and Table 2
shows the truth table for every possible combinations of inputs. Here, a0 , a1 and
b0 , b1 are the two inputs to adder and a2 , a1 , a0 is the output of the circuit (Fig. 2).

a0 : • H • H •
π π −π
2 2
a1 : H • • • H
π
2
b0 : • •

result : /2
0 1

Fig. 1 Circuit diagram of one-qubit QFT adder

Table 1 Truth table of one-qubit QFT adder

Inputs Outputs
A(a0 ) B(b0 ) A(a1 a0 )
a0 b0 a1 a0
0 0 0 0
0 1 0 1
1 0 0 1
1 1 1 0
424 S. Jakhodia et al.

a0 : • • H • H • •
π π π −π −π
4 2 2 4
a1 : • H • • • • H •
π π π −π
2 2 2
a2 : H • • • • • • H
π π
2 4
b0 : • • •

b1 : • •

result : /3
0 1 2

Fig. 2 Circuit diagram of two-qubit QFT adder

Table 2 Truth table of two-qubit QFT adder

Inputs Outputs
A(a0 a1 ) B(b0 b1 ) A(a2 a1 a0 )
a1 a0 b1 b0 a2 a1 a0
0 0 0 0 0 0 0
0 0 0 1 0 0 1
0 0 1 0 0 1 0
0 0 1 1 0 1 1
0 1 0 0 0 0 1
0 1 0 1 0 1 0
0 1 1 0 0 1 1
0 1 1 1 1 0 0
1 0 0 0 0 1 0
1 0 0 1 0 1 1
1 0 1 0 1 0 0
1 0 1 1 1 0 1
1 1 0 0 0 1 1
1 1 0 1 1 0 0
1 1 1 0 1 0 1
1 1 1 1 1 1 0

4 Experiments on Quantum Additions Using One-Qubit

and Two-Qubit QFT Adders

Experiments were conducted on one-qubit and two-qubit QFT adders (Figs. 1 and
2) on IBM Quantum Assembly Language (QASM) Simulator [11] and eight IBM
real quantum devices: YorkTown, Santiago, Athens, Quito, Belem, Lima, Manila and
Melbourne. The IBM Melbourne system supports fifteen qubits, whereas the rest of
the seven quantum hardware supports only five qubits. The authors have considered
1024 shots in each run and a total of ten runs on IBM devices. Results from each
Experimental Evaluation of QFT Adders on IBM QX Hardware 425

run are used to compute the probabilities of all the possible states and determine the
accuracy of quantum circuits on eight IBM Quantum Systems.
The accuracy rate can be calculated by taking the summation of all probabilities
of all expected output states divided by the number of all possible input states [13].
Mathematically, the accuracy rate can be calculated as
N +1
1 
2N
2 2
(i, j) (i, j)
accuracy_rate = pideal × preal_hw (10)
22N j=1 i=1

where pideal and preal_hw are the probabilities of the ideal simulator and that of the
real quantum hardware, respectively, for N-qubit adders; i and j represents the
possible input states and output states respectively.
The experimental analysis of accuracy rate on one-qubit and two-qubit QFT adder
circuits over IBM real quantum systems are as follows:

4.1 Experimental Analysis of One-Qubit QFT Adder

The authors have created all possible inputs states using X gates, tested the output and
then analysed the results by plotting histogram. Figures 3, 4, 5, 6, 7, 8, 9, 10 and 11
show the histograms of the results obtained by experimental evaluation of one-qubit
QFT adder (Fig. 1) for all nine IBM quantum systems. The accuracy rate is calculated
by comparing the results of histogram with the truth table (Table 1) according to
expected outputs. Table 3 gives the accuracy rate of one-qubit QFT adder on different
IBM quantum systems. The QASM simulator [14] is the ideal simulator and provides
100% accuracy (Table 3). From Table 3, it can be concluded that IBM Lima and
IBM Athens provide better accuracy results of 80.2% and 80.125%, respectively,
amongst available five-qubit and fifteen-qubit quantum hardware systems not taking
into consideration the QASM simulator (ideal simulator). Note: The text colour in
Table 3 are referred the same as the colour of histograms.

4.2 Experimental Analysis of Two-Qubit QFT Adder

Figures 12, 13, 14, 15, 16, 17, 18, 19 and 20 show the results obtained by experimental
evaluation of two-qubit QFT Adder (Fig. 2) for available IBM quantum systems. The
probabilities are calculated for the possible output states with reference to input states
as per the truth table (Table 2). Table 4 provides accuracy results of two-qubit QFT
adders. From Table 4, it can be clearly stated that IBM Athens performs the best
with accuracy rate of 25.575% amongst the presently available systems not taking
into consideration the QASM simulator [14] (ideal simulator) that provides 100%
accuracy.
426 S. Jakhodia et al.

1.00
1 1.00

1.00
1.00

0.75

Probability
0.5

0.00
0.25
0.00
0.00
0.00
0.00
0.00
0
0.00
0.00
0.00
0+0
0.00 11
0+1 0.00
0.00 10
1+0
01
1+1 00
Inputs Outputs

Fig. 3 Experimental results of one-qubit QFT adder on IBM quantum device: IBM QASM (32
qubits) [15]

0.75 0.730

0.541 0.621
0.5
Probability

0.457
0.120 0.055 0.148
0.25 0.177
0.142 0.257
0.095

0.169
0.100

0
0.103 0.117
0+0 0.169

0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 4 Experimental results of one-qubit QFT adder on IBM quantum device: IBM YorkTown (5
qubits) [15]

0.693

0.653
0.554
0.621
0.5
Probability

0.065
0.25 0.146 0.096 0.169 0.195
0.078
0.199
0.139
0.087

0 0.163

0+0 0.064
0.077
0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 5 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Santiago (5
qubits) [15]
Experimental Evaluation of QFT Adders on IBM QX Hardware 427

1
0.907

0.786
0.75
0.784

Probability
0.728
0.5

0.032
0.25 0.034 0.068

0.026 0.048 0.085

0.068 0.092
0 0.098

0+0 0.062 0.090

0+1 0.091 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 6 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Athens (5
qubits) [15]

0.746
0.75

0.719
0.604
Probability

0.5 0.616

0.058
0.25 0.098
0.083
0.099 0.108
0.205 0.076
0.122
0.046
0 0.159
0.142
0+0
0.120
0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 7 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Quito (5 qubits)
[15]

0.774
0.75
0.686
0.721

0.634
Probability

0.5

0.048
0.25 0.090
0.085
0.088 0.136
0.082
0.147
0.054 0.063
0
0.207
0.090 0.096
0+0

0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 8 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Belem (5
qubits) [15]
428 S. Jakhodia et al.

1
0.871

0.785
0.75
0.778
Probability
0.774

0.5

0.032
0.25 0.097
0.044
0.053
0.043 0.070
0.079
0 0.075 0.062

0+0 0.072
0.064
0+1 0.099 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 9 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Lima (5 qubits)
[15]

0.861

0.75 0.731
Probability

0.743
0.721
0.5

0.036
0.25 0.052 0.070
0.051 0.075 0.115

0.123 0.063
0 0.045

0+0 0.078
0.164 0.070
0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 10 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Manila (5
qubits) [15]

0.575

0.5

0.289
0.342 0.412
Probability

0.235 0.134 0.397 0.339

0.239
0.25 0.146

0.133
0.174

0.146
0.175

0 0.143

0+0 0.121

0+1 11
10
1+0
01
1+1 00
Inputs Outputs

Fig. 11 Experimental results of one-qubit QFT adder on IBM quantum device: IBM Melbourne
(15 qubits) [15]
Experimental Evaluation of QFT Adders on IBM QX Hardware 429

Table 3 Accuracy rate of one-qubit QFT adder on real IBM quantum devices [15]
Sl. No. IBM Quantum devices Accuracy rate (in %)
1. IBM QASM Simulator (32 qubits) 100
2. IBM YorkTown (5 qubits) 58.725
3. IBM Santiago (5 qubits) 63.025
4. IBM Athens (5 qubits) 80.125
5. IBM Quito (5 qubits) 67.125
6. IBM Belem (5 qubits) 70.375
7. IBM Lima (5 qubits) 80.2
8. IBM Manila (5 qubits) 76.4
9. IBM Melbourne (15 qubits) 43.15

1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

00+00
0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00
00+01
0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00
00+10
0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00
00+11
0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00
01+00
0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00
01+01
0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00
01+10
Inputs

0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00

01+11
0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00
10+00
0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00
10+01
0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00
10+10
0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00
10+11
0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00
11+00
0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00
11+01
0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00
11+10
0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 12 Experimental results of two-qubit QFT adder on IBM quantum device: IBM QASM (32
qubits) [15]

An another differentiating factor of two-qubit QFT adder is that it requires only

a minimum of five qubits and can run on any eight IBM systems readily avail-
able, whereas the conventional gate-based quantum adders requires a minimum of
seven qubits and one ancilla qubit for computation possible for IBM Melbourne (e.g.
ibmq_16_Melbourne) server only.
430 S. Jakhodia et al.

0.177 0.312 0.210 0.139 0.035 0.053 0.042 0.032

00+00
0.196 0.159 0.338 0.129 0.037 0.039 0.074 0.027
00+01
0.261 0.179 0.159 0.199 0.061 0.046 0.050 0.046
00+10
0.206 0.239 0.167 0.211 0.034 0.047 0.032 0.063
00+11
0.250 0.142 0.245 0.196 0.053 0.021 0.049 0.044
01+00
0.231 0.245 0.052 0.039 0.173 0.177 0.041 0.042
01+01
0.200 0.175 0.270 0.156 0.044 0.039 0.079 0.037
01+10
Inputs

0.233 0.242 0.040 0.049 0.149 0.196 0.049 0.041

01+11
0.162 0.223 0.036 0.053 0.163 0.291 0.026 0.046
10+00
0.361 0.146 0.071 0.031 0.219 0.100 0.044 0.027
10+01
0.222 0.237 0.034 0.051 0.189 0.190 0.045 0.031
10+10
0.183 0.168 0.237 0.252 0.039 0.031 0.042 0.048
10+11
0.219 0.169 0.303 0.156 0.048 0.027 0.050 0.028
11+00
0.231 0.206 0.046 0.051 0.201 0.192 0.028 0.044
11+01
0.280 0.141 0.242 0.167 0.045 0.035 0.059 0.031
11+10
0.229 0.032 0.229 0.044 0.232 0.046 0.166 0.021
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 13 Experimental results of two-qubit QFT adder on IBM quantum device: IBM YorkTown (5
qubits) [15]

0.119 0.111 0.119 0.109 0.137 0.131 0.150 0.123

00+00
0.174 0.154 0.115 0.106 0.091 0.100 0.115 0.145
00+01
0.144 0.178 0.141 0.154 0.125 0.100 0.079 0.080
00+10
0.148 0.177 0.154 0.126 0.113 0.091 0.083 0.107
00+11
0.135 0.118 0.101 0.188 0.039 0.053 0.064 0.302
01+00
0.129 0.224 0.234 0.147 0.079 0.055 0.055 0.077
01+01
0.135 0.128 0.153 0.143 0.137 0.102 0.106 0.097
01+10
Inputs

0.146 0.100 0.155 0.097 0.125 0.099 0.169 0.110

01+11
0.104 0.190 0.259 0.157 0.054 0.031 0.073 0.131
10+00
0.071 0.086 0.102 0.213 0.085 0.085 0.145 0.214
10+01
0.153 0.050 0.198 0.068 0.285 0.064 0.142 0.039
10+10
0.060 0.121 0.103 0.214 0.139 0.188 0.094 0.082
10+11
0.129 0.075 0.223 0.092 0.091 0.063 0.229 0.099
11+00
0.069 0.180 0.090 0.189 0.079 0.149 0.077 0.166
11+01
0.238 0.082 0.141 0.059 0.112 0.061 0.189 0.118
11+10
0.100 0.102 0.084 0.083 0.105 0.236 0.132 0.158
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 14 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Santiago (5
qubits) [15]
Experimental Evaluation of QFT Adders on IBM QX Hardware 431

0.253 0.129 0.118 0.061 0.105 0.094 0.130 0.110

00+00
0.150 0.203 0.166 0.133 0.115 0.089 0.066 0.070
00+01
0.175 0.099 0.266 0.062 0.143 0.072 0.105 0.078
00+10
0.066 0.073 0.122 0.183 0.116 0.128 0.157 0.154
00+11
0.167 0.272 0.072 0.050 0.076 0.167 0.082 0.113
01+00
0.070 0.097 0.207 0.102 0.131 0.133 0.184 0.077
01+01
0.121 0.158 0.150 0.224 0.099 0.091 0.065 0.092
01+10
Inputs

0.120 0.080 0.125 0.108 0.265 0.114 0.117 0.070

01+11
0.200 0.098 0.223 0.102 0.138 0.061 0.117 0.062
10+00
0.127 0.078 0.135 0.238 0.104 0.096 0.094 0.129
10+01
0.114 0.061 0.146 0.108 0.345 0.104 0.074 0.048
10+10
0.065 0.093 0.054 0.127 0.115 0.322 0.105 0.118
10+11
0.062 0.128 0.123 0.269 0.130 0.117 0.072 0.099
11+00
0.202 0.068 0.054 0.074 0.256 0.147 0.115 0.083
11+01
0.047 0.122 0.086 0.109 0.130 0.358 0.095 0.053
11+10
0.133 0.105 0.151 0.056 0.102 0.132 0.208 0.113
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 15 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Athens (5
qubits) [15]

0.198 0.179 0.099 0.108 0.127 0.120 0.078 0.091

00+00
0.146 0.120 0.136 0.136 0.129 0.106 0.128 0.099
00+01
0.159 0.119 0.162 0.090 0.139 0.102 0.150 0.079
00+10
0.178 0.088 0.139 0.093 0.156 0.106 0.140 0.101
00+11
0.171 0.154 0.096 0.088 0.142 0.136 0.116 0.098
01+00
0.160 0.149 0.159 0.087 0.104 0.091 0.141 0.108
01+01
0.126 0.135 0.132 0.131 0.132 0.119 0.119 0.106
01+10
Inputs

0.155 0.140 0.150 0.109 0.127 0.104 0.115 0.110

01+11
0.173 0.120 0.166 0.100 0.124 0.102 0.125 0.091
10+00
0.096 0.137 0.164 0.160 0.082 0.098 0.114 0.149
10+01
0.189 0.129 0.129 0.124 0.122 0.100 0.112 0.095
10+10
0.165 0.111 0.132 0.127 0.140 0.107 0.134 0.084
10+11
0.138 0.166 0.104 0.107 0.092 0.122 0.126 0.145
11+00
0.128 0.150 0.137 0.131 0.160 0.109 0.092 0.093
11+01
0.169 0.122 0.119 0.095 0.187 0.104 0.120 0.084
11+10
0.157 0.133 0.146 0.135 0.123 0.103 0.109 0.094
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 16 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Quito (5
qubits) [15]
432 S. Jakhodia et al.

0.247 0.141 0.248 0.068 0.098 0.087 0.066 0.045

00+00
0.131 0.109 0.153 0.133 0.143 0.101 0.145 0.086
00+01
0.203 0.067 0.144 0.071 0.309 0.043 0.122 0.041
00+10
0.434 0.057 0.085 0.042 0.115 0.063 0.126 0.078
00+11
0.124 0.191 0.137 0.172 0.104 0.107 0.084 0.080
01+00
0.122 0.084 0.137 0.191 0.149 0.111 0.092 0.113
01+01
0.165 0.249 0.118 0.075 0.092 0.114 0.090 0.097
01+10
Inputs

0.190 0.098 0.114 0.062 0.161 0.091 0.167 0.116

01+11
0.136 0.078 0.190 0.128 0.150 0.096 0.148 0.073
10+00
0.129 0.078 0.152 0.129 0.174 0.085 0.152 0.101
10+01
0.145 0.102 0.127 0.079 0.151 0.128 0.180 0.089
10+10
0.186 0.155 0.116 0.104 0.106 0.093 0.114 0.125
10+11
0.160 0.122 0.086 0.114 0.162 0.143 0.102 0.111
11+00
0.118 0.165 0.162 0.108 0.074 0.087 0.104 0.101
11+01
0.176 0.127 0.282 0.079 0.114 0.047 0.115 0.060
11+10
0.169 0.088 0.191 0.090 0.150 0.088 0.115 0.108
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 17 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Belem (5
qubits) [15]

0.193 0.108 0.198 0.078 0.153 0.085 0.124 0.060

00+00
0.194 0.170 0.138 0.159 0.111 0.089 0.082 0.057
00+01
0.244 0.117 0.159 0.158 0.115 0.052 0.079 0.075
00+10
0.124 0.123 0.095 0.173 0.178 0.125 0.071 0.111
00+11
0.223 0.218 0.092 0.082 0.123 0.121 0.085 0.057
01+00
0.177 0.146 0.165 0.114 0.116 0.109 0.109 0.063
01+01
0.215 0.158 0.133 0.129 0.141 0.112 0.059 0.054
01+10
Inputs

0.185 0.111 0.147 0.088 0.170 0.103 0.116 0.080

01+11
0.170 0.045 0.201 0.047 0.227 0.047 0.211 0.053
10+00
0.133 0.053 0.196 0.088 0.267 0.079 0.138 0.047
10+01
0.146 0.124 0.146 0.087 0.133 0.125 0.155 0.084
10+10
0.195 0.155 0.116 0.095 0.150 0.146 0.091 0.051
10+11
0.189 0.161 0.156 0.129 0.102 0.100 0.085 0.078
11+00
0.220 0.116 0.126 0.055 0.107 0.088 0.179 0.109
11+01
0.185 0.261 0.122 0.105 0.062 0.059 0.080 0.126
11+10
0.141 0.094 0.167 0.095 0.119 0.084 0.176 0.125
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 18 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Lima (5
qubits) [15]
Experimental Evaluation of QFT Adders on IBM QX Hardware 433

0.249 0.116 0.154 0.068 0.127 0.071 0.116 0.098

00+00
0.260 0.198 0.130 0.100 0.077 0.050 0.057 0.129
00+01
0.085 0.033 0.216 0.081 0.354 0.066 0.115 0.049
00+10
0.119 0.112 0.051 0.080 0.177 0.240 0.130 0.091
00+11
0.086 0.248 0.175 0.128 0.099 0.161 0.054 0.050
01+00
0.182 0.076 0.158 0.110 0.319 0.052 0.060 0.043
01+01
0.103 0.098 0.057 0.077 0.104 0.172 0.238 0.152
01+10
Inputs

0.190 0.222 0.049 0.048 0.103 0.157 0.091 0.141

01+11
0.109 0.088 0.314 0.121 0.166 0.047 0.086 0.068
10+00
0.090 0.076 0.055 0.174 0.174 0.136 0.086 0.210
10+01
0.072 0.091 0.144 0.150 0.207 0.142 0.102 0.093
10+10
0.111 0.100 0.127 0.045 0.081 0.122 0.269 0.146
10+11
0.138 0.107 0.084 0.128 0.110 0.104 0.098 0.230
11+00
0.130 0.063 0.147 0.053 0.138 0.094 0.310 0.065
11+01
0.137 0.191 0.066 0.060 0.166 0.150 0.150 0.144
11+10
0.123 0.072 0.128 0.065 0.080 0.072 0.241 0.218
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 19 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Manila (5
qubits) [15]

0.171 0.118 0.160 0.104 0.149 0.188 0.100 0.079

00+00
0.131 0.131 0.123 0.110 0.156 0.139 0.108 0.102
00+01
0.147 0.100 0.160 0.114 0.126 0.103 0.162 0.088
00+10
0.141 0.116 0.152 0.128 0.126 0.120 0.135 0.082
00+11
0.160 0.126 0.146 0.116 0.137 0.110 0.097 0.108
01+00
0.196 0.124 0.127 0.093 0.149 0.096 0.121 0.094
01+01
0.159 0.140 0.146 0.098 0.145 0.081 0.119 0.113
01+10
Inputs

0.152 0.111 0.139 0.124 0.157 0.118 0.129 0.069

01+11
0.155 0.100 0.154 0.096 0.136 0.100 0.146 0.114
10+00
0.143 0.123 0.165 0.095 0.121 0.105 0.134 0.114
10+01
0.150 0.113 0.141 0.102 0.145 0.104 0.136 0.110
10+10
0.169 0.128 0.134 0.104 0.146 0.116 0.116 0.087
10+11
0.127 0.100 0.144 0.111 0.144 0.135 0.134 0.106
11+00
0.180 0.129 0.145 0.111 0.143 0.095 0.113 0.085
11+01
0.151 0.090 0.138 0.101 0.152 0.097 0.162 0.109
11+10
0.176 0.093 0.185 0.086 0.170 0.077 0.134 0.080
11+11

000 001 010 011 100 101 110 111

Outputs

Fig. 20 Experimental results of two-qubit QFT adder on IBM quantum device: IBM Melbourne
(15 qubits) [15]
434 S. Jakhodia et al.

Table 4 Accuracy rate of two-qubit QFT adder on real IBM quantum devices [15]
Sl. No. IBM Quantum devices Accuracy rate (in %)
1. IBM QASM Simulator (32 qubits) 100
2. IBM YorkTown (5 qubits) 12.8125
3. IBM Santiago (5 qubits) 15.43125
4. IBM Athens (5 qubits) 25.575
5. IBM Quito (5 qubits) 13.62
6. IBM Belem (5 qubits) 12.62
7. IBM Lima (5 qubits) 15.1
8. IBM Manila (5 qubits) 17.52
9. IBM Melbourne (15 qubits) 13.08125

5 Conclusions

This paper discussed an experimental evaluation of the performance of the nine

IBM QX hardware (YorkTown, Santiago, Athens, Quito, Belem, Lima, Manila and
Melbourne) for quantum simulations of one-qubit and two-qubit QFT adder circuits.
Quantum circuits were developed to represent quantum additions using Qiskit for
simulating quantum computation and considered 1024 shots in each of the ten runs
performed on IBM devices. The authors have considered accuracy rate as a metric
to comparison the performance of various real quantum devices.
Future works include development of quantum circuits for higher-qubit QFT
adders on real hardware. Moreover, error mitigation techniques to manage various
types of hardware noise (readout error, gate error and state preparation error) and
improve the simulation performance of quantum additions will also be investigated.

References

1. Nielsen MA, Chuang IL (2011) Quantum computation and quantum information, 10th edn.
Cambridge University Press, USA
2. Shaik EH, Rangaswamy N (2020) Implementation of quantum gates based logic circuits using
IBM Qiskit. In: 2020 5th International conference on computing, communication and security
(ICCCS), pp 1–6
3. Sohel MA, Zia N, Ali MA, Zia N (2020) Quantum computing based implementation of full
adder. In: 2020 IEEE International conference for innovation in technology (INOCON), pp
1–4
4. Beauregard S (2003) Circuit for Shor’s algorithm using 2n+3 qubits. Quantum Inf Comput
3(2):175–185
5. Shor PW (1997) Polynomial-time algorithms for prime factorization and discrete logarithms
on a quantum computer. SIAM J Comput 26(5):1484–1509
6. Draper TG (2000) Addition on a quantum computer. Available at https://arxiv.org/pdf/quant-
ph/0008033.pdf
Experimental Evaluation of QFT Adders on IBM QX Hardware 435

7. Asaka R, Sakai K, Yahagi R (2020) Quantum circuit for the Fast Fourier Transform. Quantum
Inf Process 19(8)
8. Methachawalit W, Chongstitvatana P (2020) Adder circuit on IBM universal quantum com-
puters. In: 2020 17th International conference on electrical engineering/electronics, computer,
telecommunications and information technology (ECTI-CON), pp 92–95
9. Zhou SS, Loke T, Izaac JA, Wang JB (2017) Quantum Fourier Transform in computational
basis. Quantum Inf Process 16(3)
10. Ruiz-Perez L, Garcia-Escartin JC (2017) Quantum arithmetic with the Quantum Fourier Trans-
form. Quantum Inf Process 16(6)
11. IBM quantum experience. Available at https://quantum-computing.ibm.com/. Accessed on 5
June 2021
12. Sutor RS (2019) Dancing with qubits, 1st edn. Packt Publishing
13. Singh D, Jakhodia S, Jajodia B (2021) Experimental evaluation of adder circuits on IBM
QX Hardware. In: International conference on inventive computation technologies (ICICIT).
Springer, Berlin, pp 1–13
14. Cross AW, Bishop LS, Smolin JA, Gambetta JM (2017) Open quantum assembly language.
Available at https://arxiv.org/pdf/1707.03429.pdf
15. IBM quantum backends. Available at https://quantum-computing.ibm.com/docs/manage/
backends/. Accessed on 5 June 2021
Wireless Networks and Internet
of Everything (IoE)
A Microcontroller Based Hardware
Implementation to Detect and Protect
a Transformer Against Various Faults

Shafqat Nabi Mughal and Mustansir Nizam

Abstract The objective of this paper is to provide a mechanism to detect and protect
a transformer against various kinds of faults. The protection scheme for a transformer
is highly desirable as the continuity of power supply is directly associated with it. A
reliable transformer protection scheme is desirable for safe and economically viable
operation in the power system. In order to protect a transformer against different
kinds of faults many methods are in vogue, however, each of them is suitable for a
particular kind of fault and is very complex in nature. This paper describes a design
and execution of microcontroller-based system for protecting a transformer. In this
research work, a microcontroller is used to detect electrical faults and send signal to a
relay to disconnect the transformer from the system. At the time of fault occurrence,
it generates a signal to the control room providing information pertaining to position
of fault. The proposed methodology has been hardware implemented as well.

Keywords Transformer · Microcontroller · GSM · Relay · Temperature sensor

1 Introduction

An electric power system is a network containing different electrical instruments used

in the generation, transmission, and distribution of electric power. In order to meet
the specific needs of different types of consumers, special devices that change the
voltage levels are required. Some of these devices comprise of transformers, capacitor
banks, protection devices, etc. Protection of power system against various kinds of
fault is necessity for better power quality. The occurrence of fault is dangerous for
both electric power consumers and the electric network itself. The main function of
a power system is consumer should receive a stable supply of electric power and
also make sure that the electrical instrument does not get damaged. The transformer
is a heart of electrical power system. It is the most important instrument to transfer

S. N. Mughal (B) · M. Nizam

Department of Electrical Engineering, SoET, BGSB University, Rajouri, Jammu and Kashmir,
India
e-mail: snmughal@bgsbu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 439
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_32
440 S. N. Mughal and M. Nizam

electrical energy from one circuit to another. A number of transformer protection

schemes are available in the literature such as microprocessor-based relay systems,
differential protection system, etc. However, in this research work, we are designing
a mechanism using microcontroller and Global System Mobile (GSM) technology
along with a step-down transformer. Here the system voltage is decreased from 220 to
12 V and an automotive localization system is developed using GSM-SMS services.
This system communicates a message to predefined numbers about the status of
different electrical parameters such as current, temperature, voltage, power etc.
Transformer being heart of the electric power system, hence its protection against
various faults is very important. These are used in everyday life for different purposes
[1, 2]. They are also used to isolate two circuits electrically [3, 4]. They are used to
increase or decrease the alternating voltages in the various protective systems as well
[5, 6]. They are used to decrease the voltage from high level to low level for safety
[7]. They are used as rectifier also [8, 9]. They are used in power supplies, voltage
regulators, voltage stabilizers, etc. [10].
With the increase in the demand of power, transformers become overloaded. The
overloading of transformers arises owing to the increase in voltage/current levels
than the specified ratings of a transformer. Due to overloading, a large amount of
heat gets generated resulting in the breakdown of the insulation system leading
to reduction in the life expectancy of a transformer. So a control mechanism is
needed to overcome this overloading [11, 12]. Nowadays, microcontroller-based
relay protection is used for overloading protection, however, due to harmonics present
it becomes problematic to differentiate between different kinds of faults [13]. Earlier
solid state relays were used and now a days there is a transition towards digital relays
[14]. In the proposed method, we have provided microcontroller-based protection
against overload and overheating. Moreover using GSM technology an SMS can
also be generated to the control room indicating the nature and location of fault as
well.

2 Problem Formulation

The safety of transformers is critical since they are critical components of the elec-
trical power grid. Transformers can be damaged by the overcurrent phenomena.
When a transformer is damaged, the electrical supply to customers is disrupted.
Blackouts wreak havoc on the economy and throw customers’ work schedules off.
This illustrates why this system is required and how it may assist in mitigating the
impact of a transformer malfunction. To function at maximum capacity, electrical
transformers require the coordination of many systems as well as routine main-
tenance. A failure can be caused by a variety of sources, including mechanical or
electrical failures, lightning, or even human mistake. Fortunately, a lengthy history of
failure documentation and increased maintenance standards can assist in preventing
or detecting failures before serious harm occurs. Insulation deterioration is a typical
problem with older transformers. As a transformer warms up, the high temperatures
A Microcontroller Based Hardware Implementation … 441

Fig. 1 Types of transformer

faults Earth Faults

Core Faults

Inter turn
Transformer Faults
Faults Phase to
Phase Faults

Tank Faults

External
Faults

can cause the insulating system to degrade over time. If the transformer is repeatedly
overloaded, the system will deteriorate faster. A compromised insulating system is
the end outcome of minor, gradual improvements in loading capacity over time.
Overloading leads to overheating and, eventually, thermal damage, which occurs
through insulating fractures. Line surges, which are a major source of transformer
failure, are a prevalent cause of electrical failure. Electrical failure can be caused by
voltage spikes, switching surges, and line failures, to name a few. High-quality surge
protection and frequent maintenance of the existing surge protection system are the
key ways to avoid this Fig. 1 shows various types of transformer faults. These occur
due to issues inside the transformer as well as due to external causes. Each type of
fault causes interruption to power supply. Hence it becomes necessary to avoid them
by providing suitable protections. Table 1 shows different types of faults and their
protection mechanisms.
Modern power system demands highly accurate and efficient techniques/methods
to locate and isolate the faulty section from the healthy section. This involves
sensor-based daily monitoring operations for fast response. Since a typical micro-
processor relay does not provide speedy response, so the research is shifted
towards microcontroller-based relays. With the fast response of microcontroller-
based systems, we have proposed a microcontroller-based setup to detect and isolate
the fault in the transformer.

3 Proposed Model

The block diagram representation of our proposed setup is shown in Fig. 2. The main
goal of this project is to create an automated overcurrent relay that employs a micro-
controller to read transformer currents and isolates the transformer from the power
442 S. N. Mughal and M. Nizam

Table 1 Types of fault and

Type of fault Type of protection
its protection scheme
Phase-phase fault (primary Differential relays and
side) overcurrent protection
Phase-earth fault (primary Differential relays and
side) overcurrent protection
Phase-phase fault (secondary Differential relay protection
side)
Phase-earth fault (secondary Differential relay and earth
side) fault protection
Inter-turn faults Differential relay, and
Buchholz relay
Faults in the core Differential relay and
Buchholz relay
Faults in the oil tank Differential relay, earth relay
and Buchholz relay
Over flux Over flux protection
Over heating Thermal relays

Fig. 2 Proposed setup for transformer protection

supply automatically in the event of a malfunction. As a result, the key control element
in this system is the programmable Interrupt Controller (PIC microcontroller). This
system’s design is broken down into the following sections: Various components that
are required in the design process of the hardware model are as follows.
• Transformer
It is a device that is required to increase/decrease the voltage level of the system.
In various stages of the power system network, a differently rated transformer is
A Microcontroller Based Hardware Implementation … 443

required. In our proposed setup we are using a stepdown transformer to change

voltage level from 230 to 12 V.
• Electrical Load
A load is something that consumes power. There are several types of loads like
domestic, industrial, municipal, etc. Each of them are an amalgam of resistive,
inductive, and capacitive types.
• Current transformer
A current transformer (CT) is a special device used to measure current which
is of higher magnitude. It is used for the protection of ammeter for measuring
current in high power line. In our setup, the current rating of the device varies
from 20A to 20 mA.
• Microcontroller
A microcontroller PIC16f877a is used in the design setup to provide multiple
options of use as it can accommodate many operations at one time owing to its
variety of inputs and output pin configurations and flash memory technology. This
controller being simple is easy to implement and provides a reliable setup as well.
• Relay Driver Circuit
It is an integrated circuit comprising of NPN transistor in the design setup
to control the operation of the relay. This is important to isolate the faulty
section/transformer from the power supply.
• Relay
A relay is a switch that causes the operation of the circuit breaker in the event
of fault. It has a pickup current setting that initiates the operation of the circuit
breaker if higher than predetermined value.
• Temperature Sensor
In order to sense the change in the temperature, a temperature sensor is required
which is basically a thermocouple. Here in our setup LM35 is chosen to sense the
varying temperatures.
• LCD Display
All the readings which are needed to be displayed for proper action are show-
cased using liquid crystal display panels. They provide information about various
voltages, currents, etc.
• GSM
The word GSM stands for global system for mobile communication. It has
spread its use in all walks of life. Here in our system, we are using it to generate
and transmit the signal in the form of SMS to the end user in the event of fault in
the transformer.
• Power Supply
A power supply is basically an energy source to make things work. In our setup
5 V supply was required which was generated using step-down transformer and
rectifier/filter circuits, etc.
444 S. N. Mughal and M. Nizam

4 Hardware Implementation and Working

The hardware setup of the proposed design is shown in Fig. 3. It comprises various
components as discussed in the Sect. 3.The design was fabricated in the laboratory.
The model was designed such that the system can be switched ON/OFF using a
Personal Computer through GSM. By using all the described components above, a
hardware setup was designed that disconnects the transformer from the power supply
in the event of fault like overvoltage and overheating, etc. This setup will prevent
the cascading effect of the damage of fault in the power system. The two cases are
described as follows:
(a) Effect of temperature on transformer:
The heat produced during transformer operation raises the temperature of the trans-
former’s interior structures. More efficient transformers have a smaller temperature
increase, whereas less efficient transformers have a bigger temperature increase. A
more efficient transformer produces less waste heat in the first place, but the temper-
ature of the transformer rises as a function of both the amount of heat produced and
the amount of heat removed. Make sure that a device with a minimal temperature
rise isn’t equally inefficient since it relies on fans to remove the surplus heat.
Dry-type transformers come in three temperature ranges: 80, 115, and 150 °C.
Standard rises of 55 and 65 °C are available for liquid-filled transformers. These
figures assume a maximum ambient temperature of 40 °C. For example, an 80C rise
dry transformer operating at full-rated load in a 40C ambient environment will have

Fig. 3 Hardware implementation of the model

A Microcontroller Based Hardware Implementation … 445

an average winding temperature of 120 °C. (Temperatures within the transformer may
be greater than usual at so-called hot spots.). The 80C rise unit offers greater space
for an occasional overload than the 150 °C rise unit, because most dry transformers
employ the same insulation on their windings (usually rated at 220 °C), regardless
of the design temperature rise.
Figure 4 shows the flowchart demonstrating the protection against rise in temper-
ature. In the proposed setup the preset value of temperature is set to 40 °C. If the
temperature becomes greater than this predefined value, relay operates isolating a
transformer from the power supply thereby protecting it.
(b) Effect of over-voltage on transformer:
When electrical equipment is subjected to excessive voltage, the insulation becomes
stressed, and the equipment is more likely to be damaged. Surges can induce sparking
and flashing between phase and ground at the weakest point in the network, as well as
the breakdown of gaseous, solid, and liquid insulation, and the failure of transformers
and spinning machinery.
The iron losses (hysteresis and eddy current losses) in a transformer are affected
by changes in voltage and/or frequency. Hysteresis loss (Ph ) and eddy current loss
(Pe ) vary according to the following relationships as shown below as long as the flux

Start

Microprocessor

Relay

Display ADC Value on LCD

If temperature is
greater than preset
Relay will operate
value

Display on LCD &

Message Through GSM

Fig. 4 Flowchart of circuit for temperature [15]

446 S. N. Mughal and M. Nizam

changes are sinusoidal with respect to the line.

Ph ∝ f (∅maz )z (1)

Pe ∝ f 2 (∅maz )2 (2)

where x varies from 1.5 to 2.5 depending upon the iron grade used.
If the frequency and voltage of the transformer are changed in the same proportion,
the flux density will stay unchanged and the no-load current will similarly appear
to be unaffected. The transformer may be safely used at a frequency lower than the
rated frequency, with a correspondingly lower voltage. Iron losses will be decreased
in this circumstance. The core losses may become unacceptable if the transformer is
operated with increased voltage and frequency in the same proportion.
When the frequency is increased while the supply voltage remains fixed, the
hysteresis loss is reduced while the eddy current losses are unaltered. At higher
frequencies, some voltage rise may be allowed, but how much depends on the relative
degree of hysteresis and eddy current losses, as well as the quality of iron employed
in the transformer core.
A working methodology during overvoltage scenario is depicted in the form of
flow chart as shown in Fig. 5.If the voltage becomes greater than the preset value
(300 V), relay operates and hence transformer gets protected.

5 Results and Discussion

After completing the hardware design, the following findings were obtained. Tables 2
and 3 illustrates the condition of the relay, LED, and buzzer with respect to the change
in current. The primary side circulating current of the transformer increases owing
to the decrease in load resistance. The decrease in resistance could be due to the
breakdown of insulation resulting in shorts circuits etc. Two cases are discussed in
this research work.
Case. I Effect of Temperature Rise
The results obtained for this effect are given in Table 2. The reference
temperature was kept at 40 °C.

Case. II Effect of Voltage Rise.

The results obtained for this effect are given in Table 3. The reference voltage
was kept at 240 V.

It has been discovered that the sensor output increases owing to the increase in the
transformer circulating current. The table above demonstrates this. When the sensor
output voltage was increased.
A Microcontroller Based Hardware Implementation … 447

Fig. 5 Flowchart of circuit

for overvoltage [15] START

Microprocessor

Relay

Display ADC Value on

LCD

If voltage is
Relay will
greater than
operate
preset value

Display on LCD &

Message Through GSM

Table 2 Effect of temperature rise and state of system

Loading Temperature Sensor output Relay state Buzzer state LED LCD
(°C) (V)
Low 25 0.05 OFF OFF Green –
Normal 30 0.95 OFF OFF Green –
Medium 35 1.89 OFF OFF Green –
High 45 2.5 ON ON Red Temperature
fault
Very high 50 3 ON ON Red Temperature
fault

When the transformer current supersedes the rated value kept, the system, as illus-
trated in Tables 2 and 3 activates the LCD display. The LCD shows “OVERCUR-
RENT FAULT” and the buzzer sounds to inform the system operator about a fault in
a specific transformer. To disconnect the transformer from the electricity system, a
448 S. N. Mughal and M. Nizam

Table 3 Effect of voltage rise and state of system

Loading Voltage (V) Sensor output Relay state Buzzer state LED LCD
(V)
Low 150 0.05 OFF OFF Green –
Normal 200 0.95 OFF OFF Green –
Medium 220 1.89 OFF OFF Green –
High 300 2.5 ON ON Red Overvoltage
fault
Very high 320 3 ON ON Red Over voltage
fault

signal is also transmitted to the relay. The procedure of transformer separation from
the system is carried out in this type using a relay and a contactor.
The system performs the role of an over-current relay, which detects overcurrent
faults automatically and utilizes a microcontroller to keep the fault record, sometimes
isolation of the transformer from the afflicted zone, and communicate the fault occur-
rence to the system operator. The benefit of the proposed methods over conventional
over current relay is that after the fault is cleared, it automatically recloses.

6 Conclusion

To protect transformers against failures, many relaying principles have been devised
and deployed. Over current, over flux, and overheating relays protect transformers
against overloads and other externally imposed situations. Transformers are protected
against internal faults by differential relays. By doing this research work, a novel
method of fault detection and protection for important devices like Transformer
has been established. This method will be able to protect costly equipment like
transformer etc. This method will not only detect the fault but also initiate the required
action to isolate transformer from the power supply. The system works by sending
an SMS to the end user who can switch off the system remotely.

References

1. Bashi SM, Marium N, Rafa A (2007) Power transformer protection using microcontroller-based
relay. J Appl Sci 7(12):1602–1607
2. Thiyagarajan V, Palanivel TG (2010) An efficient monitoring of substations using microcon-
troller based monitoring systems. Int J Res Rev Appl Sci 4(1):63–68
3. Benmouyal G, Guzman A, Zocholl S, Altuve HJ (2002) A current-based solution for trans-
former differential protection. II. Relay description and evaluation. IEEE Trans Power Delivery
17(4):886–893
A Microcontroller Based Hardware Implementation … 449

4. Ahamed SS (2009) The role of zigbee technology in future data communication system. IEEE
J Theor Appl Inf Technol 5(2):129–135
5. Thote PB, Daigavane MB, Bawane NG (2015) MATLAB simulink of based digital protection
of transformer. Int J Res Eng Technol 3(2):484–487
6. Jardini JA, Brittes JL, Magrini LC, Bini MA, Yasuoka J (2005) Power transformer temperature
evaluation for overloading conditions. IEEE Trans Power Delivery 20(1):179–184
7. Grainger JJ (2003) Power system analysis. Tata McGraw-Hill Publishing Co ltd.
8. Mao H (2010) Research of wireless monitoring system in power distribution transformer station
based on GPRS. In: The 2nd international conference on computer and automation engineering
(ICCAE), vol 5, pp 386–389
9. Nath S, Chakrabarti A, Mukhopadhyay AK (2006) A microcomputer-based unique digital fault
diagnosis scheme of radial transformer feeders. IEEE Trans Power Delivery 21(4):1824–1829
10. Perera N, Rajapakse AD (2012) Design and hardware implementation of a modular transient
directional protection scheme using current signals. IET Gen Trans Distrib 6(6):554–562
11. Mughal SN, Thaker R (2018) A novel approach for harmonic analysis using MATLAB. Int J
Adv Eng Res Dev Emerg Trends Innov Electron Commun Eng 5(1)
12. Malik H, Mughal SN, Sood YR, Jarial RK (2011) Application and implementation of artificial
intelligence in electrical system. In: International conference on advances in computing and
communication (ICACC-2011). NIT Hamirpur (Himachal Pradesh)
13. Wadhwa CL (2005) Electrical power systems. New Age International (P) Ltd., New Delhi, pp
656–657
14. Malik H, Azeem A, Mughal SN, Singh S (2011) Artificial intelligence techniques for incip-
ient fault diagnosis and condition assessment in transformers. In: International conference on
emerging trends in engineering (ICETE-11). Karnataka, pp 5–8
15. Salunkhe K, Shaikh S, Kunure BS (2016) Transformer protection using microcontroller based
relay & monitoring using GSM technology. Int Eng Res J (IERJ) 2(2):813–817
Fuzzy—KNN-Assisted Vehicular
Localization for Bluetooth and Wi-Fi
Scenario

Shubhangi Verma, O. P. Singh, Sachin Kumar, and Sumita Mishra

Abstract The exponential expansion of intelligent and smart communication,

infrastructure, and other frontier technology has restructured many domains of
day-to-day life and has given birth to new opportunities for managing cities more
effectively and holistically. Technologies such as high-speed internet, 5G mobile
networks, the internet of things, and big data play an increasingly important role
in transition towards “smart cities”. Technological innovations such as autonomous
vehicles among them have taken the lead. Number of automobiles companies are
offering different variants of autonomous vehicles with innovative solutions for issues
related to safety, connectivity, transportation, and more. In terms of data transfer with
optimal connection and speed, VANET technology is around here but with changing
trends in technology limitations have surfaced. The work proposed performs vehic-
ular localization employing fuzzy systems, for localization the proposed algorithm
uses k-nearest algorithm with fuzzy decision system. System suggested employs two
distance estimating techniques—Euclidian distance and physical distance between
test nodes and reference nodes. RSSI values were evaluated for BLE and Wi-Fi
scenarios for approximately 100 different vehicle positions in a 2 × 2 km simulated
environment. BLE due range constraints would require a setup with V2V and V2I
network. Simulated MATLAB environment was used for testing results. RSSI values
were then fed to Neural Network to estimate MSE and R values for both scenarios.
Experiment results demonstrate the effectiveness and merit of the two proposed
scenarios, and it was found, localization through Bluetooth outperforms slightly as
compared to Wi-Fi.

Keywords VANET · Fuzzy systems · RSSI · Localization · K-NN

S. Verma · O. P. Singh · S. Kumar (B) · S. Mishra

Amity University, Lucknow Campus, Lucknow, India
e-mail: skumar3@lko.amity.edu
O. P. Singh
e-mail: opsingh@amity.edu
S. Mishra
e-mail: smishra3@lko.amity.edu

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 451
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_33
452 S. Verma et al.

1 Introduction

Vehicular communication systems or VANET are networks in which roadside units

and vehicles are the nodes that communicate with each other and share informa-
tion like traffic status and safety information. Figure 1 depicts the vehicle to vehicle
(V2V) and vehicle to infrastructure (V2I) communication in VANET architecture.
Sharing of information in real-time communications between infrastructure and vehi-
cles has become possible owing to innovations in wireless communications. It is
important for one to find solutions based on issues with vehicular communication
like accidents, visibility, and safe environment. Other important applications include
monitoring systems, routing problems, traffic information for vehicular networks,
and vehicle localization systems. With an increase in vehicle density, it has become
extremely important to improve vehicle’s navigation and localization systems. Due
to the random and fast mobility of vehicles and weak signals, it is difficult to get accu-
rate results of location. Hence various localization techniques have been developed
and are under research.
The focus is on widely integrating with other hybrid technologies to improve the
localization system on parameters of accuracy, connectivity, timing constraints, and
other factors associated with VANETs. A recent survey was published on VANET
localization techniques [1] where the authors discussed about the improvements
in the localization and classified the VANET localization use case techniques on
the basis of 1.Non-Cooperative Localization techniques 2.Cooperative Localization
techniques as shown in Fig. 2. The cooperative technique is further classified as
GPS and non-GPS based-techniques. The various techniques use filters based on
the application and for non-cooperative techniques various communication based on
cellular towers, zigbee, or other communication protocols is used.

Fig. 1 VANET architecture

Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 453

Fig. 2 VANET localization techniques

Table 1, presents a comparison of related technologies employed for entity-to-

entity communications. Bluetooth Low Energy (BLE) offers several advantages as
compared to Bluetooth classic (BC) for vehicle to vehicle and vehicle to infrastructure
communications. BLE offers lower latency and with less bit rate a range beyond 100m
is possible, further BLE for exchanging data does not require devices to be paired.
Wi-Fi on the other hand offers significantly higher data rate not necessarily required
for V2V applications.
Wi-Fi has extremely high energy requirements which is a major disadvantage.
DSRC, has been specifically designed for Vehicle-to-vehicle communication and
it provides a better performance compared to all existing ones employed. One of

Table 1 Related V2V technologies [3]

BC BLE WiFi DSRC
Band 2.4 2.4 2.4 / 5 5.9
Channel BW 1 2 20/40 10
Number of Ch. 79 40 11 7
Modulation GFSK GFSK OFDM OFDM
Max. bit rate (Mbps) 3 1 600 27
Max. power (mW) 100 10 100 2000
Approx. range (m) 10 100 100 1000
Approx. latency (ms) 100 6 50 1
454 S. Verma et al.

Fuzzy Fuzzy
Input Output
Crisp Crisp
Input Inference Output
Fuzzification De-fuzzification
Engine

Fuzzy Rule Base......

Fig. 3 Fuzzy logic system

the major disadvantages of DSRC is that it is still not widely deployed. The work
presented compares BLE and Wi-Fi technology for V2V localization.
Work survey based on location-based routing protocols, [2] discussed about issues
and their solutions in the VANET communications. Here the authors presented issues
and challenges associated with factors including local optimum problems, accurate
positioning, broadcast overhead problem, and optimum routing selection. Hence
location-based routing protocols can be one of the best solutions as this technique
does not rely on pre-destination routes.
Another interesting work on fuzzy logic in wireless localization [4] explains the
approach as a feasible and attractive to overcome some uncertainties and challenges
in wireless localization systems. The authors reviewed various fuzzy-related method-
ologies and techniques. They also evaluated the techniques based on four factors, i.e.,
environmental metrics, system metrics, fuzzy system, and positioning metrics. They
classified fuzzy-based localization methods in two ways, i.e., fuzzy system will be
incorporated as core estimation process in positioning system and another method
is where fuzzy system will assist the overall positioning system. The fuzzy system
can be further classified as fuzzy clustering techniques, fuzzy optimization method,
fuzzy arithmetic and analysis, and hybrid-based approach.
The fuzzy inference system has three stages which includes Fuzzification—where
the fuzzy sets using the linguistic variables are constructed, then Fuzzy rule base
and Defuzzification—which produce a non-fuzzy output. The fuzzy rules work on
the pre-defined rule regarding the problem statement as shown in Fig. 3. Paper is
organized as follows section II addresses related work, section III presents proposed
methodology in detail. Simulation and results are discussed in section IV and finally,
section V concludes the paper with discussion about future direction.

2 Related Work

Jiyao An et al explained about the vehicle location problem in urban road conditions
using the GPS, WiFi, and Cellular Network (CN) positioning systems. The fuzzy-
based hybrid location algorithm is developed with some advantages like enhancing
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 455

the positioning features by new fuzzy-weighting location mechanism and also can
adjust by using fuzzy Kalman filtering method. The experiment results are evaluated
to show effectiveness and merit of the proposed approach [5].
Lina Altoaimy et al proposed the localization method based on fuzzy logic and
location information of neighbors. The objective of the proposed methodology is
to estimate the vehicle’s location by considering the neighbor’s vehicle location
information. In order to achieve accurate localization, the authors modeled vehicles’
weights with fuzzy logic system, by utilizing the distance and heading information to
obtain the weight values. They also considered the concept of centroid localization.
The results obtained from the simulation are promising and shows the effectiveness
of the proposed methodology in different traffic densities [6].
Sudhir Kumar et al discussed about the distributed and range-free algorithm to
track vehicles. Connectivity with audible anchor nodes is used to track vehicles for
range-free. The method is described as low cost and saves the cost of spending on
GPS systems in the cars. The results of the proposed system in a simulation show
the improved performance when compared with the other conventional methods [7].
Azat Rozyyev et al presented the results of the study where they combined KNN
with fuzzy logic technique which improved the accuracy of KNN. In comparison with
other localization techniques in terms of high accuracy, multilateration, and fuzzy
logic indoor positioning system fuzzy KNN got better results in terms of accuracy
and algorithm complexity [8].
Juan Cota-Ruiz et al proposed a hop-weighted scheme which can be useful for
distributed multi-hop localization based on distance. This localization approach
generates spatial locations and estimates position of unknown nodes and then
computes local functions in order to minimize distance errors. The proposed method-
ology achieves root mean square error values which are less than the radio range of
the unknown nodes with less number of iterations [9].
Kaur. A et al proposed another weighted centroid DV-Hop algorithm. This
methodology uses weights that use different factors like number of anchors, radius,
and to determine location of nearest anchor of unknown node. With the help of
simulation and theoretical analysis the proposed methodology y outperforms the
traditional algorithm of DV-hop in terms of power consumption and localization
error [10].
Kang S et al presented the algorithm based on a target localization using an
approximated error covariance matrix based weighted least squares solution, which
integrates the received signal strength and angle of arrival data for wireless network
communication. The proposed methodology does not need any prerequisite knowl-
edge of the target true position and noise variance. Simulations show the superior
performance of the method [11].
Shi X et al presented the HPWB algorithm in which unknown nodes are divided
into three categories according to their number of neighboring anchors. For category
one, weighted method based on RSSI was proposed. For category two and three
the exclusion method was presented. The simulation results show that the proposed
method HPWB got high precision, in comparison with the traditional bounding box
method and weighted centroid algorithm [12].
456 S. Verma et al.

RSSI Values

Euclidean Distance Physical Distance

Calculation Calculation

EDi PDi
Location Estimation
Wedi Wpdi (x,y)
Weight Calculation Weight Calculation
Wi

Select Nearest KNN Fuzzy Controller for Recalculation of

points KNN points Weights

Fig. 4 Proposed architecture

3 Methodology

3.1 Architecture

Paper presents a methodology for estimating location of vehicles in a vehic-

ular network, the technique parallelly evaluates Euclidean and Physical distance,
performs weight calculation to minimize error then selects k nearest neighbors with
Fuzzy logic recalculates the weight for accurate localization. Figure 4 depicts the
architecture of the proposed methodology.

3.2 RSSI Values

There are many radio propagation models known for wireless communications that
predict signal-strength loss with distance—path loss. These models are basically used
in those simulations simulating WSN. When a practical approach is required there are
two main variables measured when defining received-power and link-quality: RSSI
stands for Received Signal Strength Indicator, It is the measured power of a received
radio signal. It is implemented and widely used in 802.11 standards. Received power
can be calculated from RSSI. The other one is LQI stands for Link Quality Indicator.
LQI estimates how easily the received signal can be modulated when considering
noise in the channel. Practical examples of RSSI and LQI values can be summarized
as:
1. Weak signal and noise presence will give low RSSI and low LQI.
2. Weak signal with noise absence will give low RSSI and high LQI.
3. Strong signal and without noise gives high RSSI and high LQI.
4. Strong signal in a noisy environment will return high RSSI and low LQI.
5. And finally, strong noise my give high RSSI and low LQI.
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 457

RSSI is estimated calculated by the given Formula 1:

RSSI = Pt − Pl(d) (1)

Where Pt is signal transmission power and Pl(d) is path loss when the distance is d.
The unit for both power factors is dBm.
 
d
Pl(d) = Pl(d0) + 10n log +X (2)
d0

Here ‘X’ is the random number of the Gaussian distribution and ‘d’ is distance
between receiver and transmitter and ‘n’ is weakness factor of the environment.
Therefore,

RSSI = Pγ + G γ − Pl(d) (3)

Where Pγ is the wireless signal power and

G γ is gain of transmitting antenna.

3.3 Distance Calculation

The two distances employed for estimation here is Euclidean distance and the
physical distance based on the RSSI values.
Euclidean distance is calculated using Formula 4.


 n  2
EDi = 
j
RSSj − RSSi , i = 1, 2, 3, . . . N (4)
j=1

The weights are calculated using Formula 5.

1
WEDi = k
EDi
1
(5)
i=1 EDi

Physical distance is calculated using Formula 6.


n
j
PDi = PDi
j=1 (6)
j j
where PDi =d −j
di , i = 1, 2, 3, ...N

RSS(d0)−RSSj RSS(d0)−RSSij
j
Where d j = d0 10 10n di = d0 10 10n
458 S. Verma et al.

The weights are calculated using Formula 7.

1
WPDi = k
PDi
1
(7)
i=1 PDi

3.4 K-NN Points

K-nearest neighbors have been widely used to address localization issues, however,
traditional KNN uses only Euclidean distance to measures the similarity between
two vectors.
The K-NN algorithm working can be summarized as:
Step 1: Select the number K of the neighbors.
Step 2: Calculate the Euclidean distance of K number of neighbors.
Step 3: Take the K nearest neighbors as per the calculated Euclidean distance.
Step 4: Among these k neighbors, count the number of the data points in each
category.
Step 5: Assign the new data points to that category for which the number of the
neighbor is maximum.

3.5 Fuzzy Controller and Weights Calculation

Inserting the model in the fuzzy controller and get the common k-points and then
based on the weighted formula we can get location points.
Recalculation of weights can be done using Formula 8.

Wpdi + Wedi Wpdi.Wedi

wi = k
wi = k
(8)
i=1 (Wpdi + Wedi) i=1 (Wpdi.Wedi)

Location can be estimated using Formula 9 (Figs 5, 6 and 7).


k
x̂, ŷ = wi(xi, yi) (9)
i=1

Root mean square value is estimated employing following relationship:

2 2
δ= x̂ − x + ŷ − y
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 459

Fig. 5 Input variable as distance

Fig. 6 Output variable as weights

Fig. 7 Rules evaluation

1
n
ME = δi
n i=1
460 S. Verma et al.

 n
1 
RMSE =  δ2
n i=1 i

Where δ represents mean error, ME represents cumulative mean error and RMSE
represents root mean square error [13–18].

3.6 Algorithm

Algorithm estimates RSSI values based on reference nodes and test nodes, i.e., target
node: (x, y), simulation of the mathematical equations will perform the following
steps.
Step 1. Start.
Step 2. Measuring RSSI values from reference nodes to target nodes.
Step 3. Calculation of distance between target and reference nodes: Euclidean
distance and Physical distance.
Step 4. Finding the common nearest neighbors based on KNN algorithm.
Step 5. The common points inserted into fuzzy controller and weights are
calculated.
Step 6. According to the weights location of target node is estimated.
Step 7. MSE and R are estimated through NN tool.
Step 8. Stop.

4 Result and Discussion

To test the proposed algorithm, the experiment was carried out in simulation environ-
ment in MATLAB 2019. Simulation model was used to generate target and reference
nodes for the area (2 × 2 km) for two different scenarios in an approximated simulated
environment. Two different wireless techniques are used for the different scenarios
Bluetooth and Wi-Fi and based on their RSSI values are calculated to estimate the
position of target nodes. We compared the two wireless techniques based on the eval-
uation factors mentioned in section 3.6. The sampling rate of the RSSI acquisition
was set as 2 s for both reference points and test points. Fig. 8 shows the distribution of
nodes for the first case scenario and Fig. 9 depicts the second scenario respectively.
The weights calculated in FIS and the reference coordinates of k NN are used to
calculate the coordinates of the target, RSSI was estimated employing reference and
test nodes, approximately 100 vehicle locations were selected for both Bluetooth and
Wi-Fi scenarios for RSSI estimation.
The accuracy of RSSI values of each model was evaluated using Neural Network
model. Figs. 10 and 11 depicts the scatter plot of RSSI estimate for Bluetooth and
Wi-Fi scenarios.
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 461

Fig. 8 Scenario 1—Bluetooth

Fig. 9 Scenario 2—WiFi

462 S. Verma et al.

Fig. 10 RSSI estimates RSSI estimate Bluetooth

scenario1—Bluetooth 100

80

60

40

20

0
0 2 4 6 8 10 12

Fig. 11 RSSI estimates RSSI estimate Wi-Fi

scenario1—Wi-Fi 60
50
40
30
20
10
0
0 2 4 6 8 10 12

Mean Squared Error is the average squared difference between outputs and targets.
Lower values are better. Regression R Values measure the correlation between outputs
and targets. An R value of 1 means a close relationship, 0 a random relationship. Zero
means no error. The Mean Square Values (MSE) and R values estimated for Scenario
1 and 2 with x and y coordinates of vehicle location are presented in Table 2 [19–24].
MSE values obtained suggested increase in error in scenario 2 as compared to
scenario 1 for the same vehicle communication environment. Further, Figs 12 and 13
depict best validation test for both scenarios. Figure 12 presents that best validation
performance for scenario 1: which is 55.37 at Epoch 7 and Fig. 13 presents that best
validation performance for scenario 2: Wi-Fi which is 59.58 at Epoch 3.

Table 2 MSE and R values for scenario 1: Bluetooth and scenario 2: Wi-Fi
MSE and R values Scenario 1: Bluetooth Scenario 2: Wi-Fi
Samples MSE R MSE R
Training 73 33.29196e-0 6.10186e-1 24.03158e-0 7.24842e-1
Validation 16 55.36769e-0 4.65164e-1 59.58184e-0 4.32347e-1
Testing 16 41.50336e-0 3.30810e-1 49.95423e-0 6.01166e-1
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 463

Fig. 12 Present that best validation performance for scenario 1: Bluetooth

Fig. 13 Presents that best validation performance for scenario 2: Wi-Fi

464 S. Verma et al.

Methodology proposed compares vehicular localization for approximately 100

different target points for a simulated environment, in which Bluetooth localization
outperforms Wi-Fi.

5 Conclusion

The work proposed performs vehicular localization employing fuzzy systems, for
localization the proposed algorithm uses k-nearest algorithm with fuzzy deci-
sion system to form the cluster of vehicles under investigation. System suggested
employs two distances estimating techniques—Euclidian distance and physical
distance between test nodes and reference nodes. Simulation was implemented using
MATLAB. RSSI values were evaluated for BLE and Wi-Fi scenarios for approxi-
mately 100 different vehicle positions in a 2 × 2 km simulated environment. BLE due
range constraints would require a setup with V2V and V2I network. Simulated envi-
ronment was used for testing results. RSSI values were then fed to Neural Network
to estimate MSE and R values for both the scenarios. MSE values obtained suggested
increase in error in scenario 2 as compared to scenario 1 for same vehicle commu-
nication environment. Validation performance for scenario 1: Bluetooth comes as
epoch 7 which is 55.37 and for scenario 2: Wi-Fi comes as epoch 3 which is 59.58.
From the results, it can be concluded that localization through Bluetooth outperforms
slightly as compared to Wi-Fi owing to positional losses.

References

1. Günay FB, Öztürk E, Çavdar T, Sinan Hanay Y, KhanAR (2020) Vehicular Ad Hoc Network
(VANET) localization techniques: a survey. In: Computational methods in engineering
2. Srivastava A, Prakash A, Tripathi R (2020) Location based routing protocols in VANET: issues
and existing solutions. Elsevier
3. Kelarestaghi KB, Foruhandeh M et al (2019) Survey on vehicular Ad Hoc networks and its
access technologies security vulnerabilities and countermeasures. arXiv e-prints
4. Alakhras M, Oussalah M, Hussein M (2020) A survey of fuzzy logic in wireless localization.
EURASIP J Wireless Commun Netw
5. An J, Yu Y, Tang J, Zhan J (2019) Fuzzy-based hybrid location algorithm for vehicle position
in VANETs via fuzzy kalman filtering approach. Adv Fuzzy Syst, Hindawi
6. Altoaimy L, Mahgoub I (2014) Fuzzy logic based localization for vehicular Ad Hoc networks.
In: IEEE symposium on computational intelligence in vehicles and transportation systems
(CIVTS)
7. Kumar S, Kislay K, Singh MK, Hegde RM (2014) A range-free tracking algorithm in vehicular
Ad-Hoc Networks. In: Twentieth national conference on communications (NCC)
8. Rozyyev A, Hasbullah H, Subhan F (2012) Combined K-nearest neighbors and fuzzy logic
indoor localization technique for wireless sensor network. Res J Inf Technol
9. Cota-Ruiz J, Gonzalez-Landaeta R, Diaz-Roman JD, Mederos-Madrazo B, Sifuentes E (2019)
A weighted and distributed algorithm for multi-hop localization. Int J Distrib Sens Netw
10. Kaur A, Kumar P, Gupta GP (2017) A weighted centroid localization algorithm for randomly
deployed wireless sensor networks. J King Saud Univ Comput Inf Sci
Fuzzy—KNN-Assisted Vehicular Localization for Bluetooth … 465

11. Kang S, Kim T, Chung W (2020) Hybrid RSS/AOA localization using approximated weighted
least square in wireless sensor networks. Sensors 20(4):1159
12. Shi X, Zhang L (2013) High-precision weighted Bounding Box localization algorithm for
Wireless Sensor Network. In: IEEE third international conference on information science and
technology (ICIST)
13. Altoaimy L, Mahgoub I (2016) Mobility data verification for vehicle localization in vehicular
Ad Hoc Networks. IEEE Wireless Commun Netw Conf
14. Altoaimy L, Mahgoub I (2014) OWL: optimized weighted localization for vehicular ad hoc
networks. In: International conference on connected vehicles and expo (ICCVE)
15. Altoaimy L, Mahgoub I (2016) Mobility data verification for vehicle localization in Vehicular
Ad Hoc Networks. IEEE Wireless Commun Netw Conf
16. Sun L et al (2012) An RSU-assisted localization method in non-GPS highway traffic with dead
reckoning and V2R communications. In: 2nd international conference on consumer electronics,
communications and networks (CECNet), pp 149–152
17. Khattab A, et al (2015) High accuracy GPS-free vehicle localization framework via an INS-
assisted single RSU. Int J Distrib Sens Netw. https://doi.org/10.1155/2015/795036
18. Qafzezi E, Bylykbashi K, Ikeda M, Matsuo K, Barolli L (2020) Coordination and management
of cloud, fog and edge resources in SDN-VANETs using fuzzy logic: a comparison study
for two fuzzy-based systems. Internet of Things 11:100169. https://doi.org/10.1016/j.iot.2020.
100169
19. Balta M, Özçeli̇k I (2020) A 3-stage fuzzy-decision tree model for traffic signal optimization
in urban city via a SDN based VANET architecture. Future Gen Comput Syst 104:142–158.
https://doi.org/10.1016/j.future.2019.10.020
20. Ghaffari A (2019) Hybrid opportunistic and position-based routing protocol in vehicular ad
hoc networks. J Ambient Intell Human Comput. https://doi.org/10.1007/s12652-019-01316-z
21. Purkait R, Tripathi S (2019) Fuzzy logic based multi-criteria intelligent forward routing in
VANET. Wireless Personal Communications
22. Hu X, Ma L, Ding Y, Xu J, Li Y, Ma S (2019) Fuzzy logic-based geographic routing protocol
for dynamic wireless sensor networks. Sensors 19(1):196. https://doi.org/10.3390/s19010196
23. Peng X, Chen R, Yu K et al (2020) An improved weighted K-nearest neighbor algorithm for
indoor localization. Electronics 9:2117. https://doi.org/10.3390/electronics9122117
24. Khanna A, Rodrigues JJPC, Gupta N, Swaroop A, Gupta D (2020) Local mutual exclusion
algorithm using fuzzy logic for flying Ad hoc networks. Comput Commun 156:101–111
A Proposal of Iot Based
Evapotranspiration Irrigation System
for Water Conservation

Simrat Walia and Jyotsna Sengupta

Abstract Water conservation is a prime concern in arid and semi-arid regions. The
farming sector consumes a lot of water in irrigation and other activities. IoT (Internet
of Things) and WSN (Wireless Sensor Network) are extensively used in precision
agriculture to control fundamental activities like Irrigation Scheduling, Data Acqui-
sition, Data Storage and Analysis. These systems involve sensors inputs to be evalu-
ated against some threshold values and decisions are taken. The proposed automated
irrigation system involves monitoring environmental parameters and controlling the
irrigation process remotely, involves short-range communication from sensors to the
base station and long-range communication from the base station to the internet.
The focus of the proposed system is to conserve water and other resources. The
system incorporates use of sensor data, evapotranspiration of region and precipita-
tion information from online weather forecast. The hardware components of system
are microcontroller, 5v DC motor, L293 motor driver module, resistive soil moisture
sensor and breadboard. The soil moisture sensor continuously records the moisture
of the soil and transmits readings to the microcontroller. The DC submersible motor
connected to the microcontroller through L293 motor driver module, supply water
when soil moisture level is below the threshold. The system generates irrigation
schedules are based on evapotranspiration of the region and sensor feedback.

Keywords IoT (Internet of Things) · Automated · Irrigation · Evapotranspiration ·

GPRS · ZigBee

1 Introduction

Water is a fundamental resource for life and having importance in many sectors
such as agriculture, industry, recreation and residential. The freshwater resources are

S. Walia (B)
Department of CSE, Baba Banda Singh Bahadur Engineering College, Fatehgarh Sahib, India
J. Sengupta
Department of Computer Science, Punjabi University, Patiala, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 467
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_34
468 S. Walia and J. Sengupta

groundwater and surface water. These resources are insufficient as demand for fresh-
water increase with population growth. In human civilization, agriculture consumes
approximately seventy percent of freshwater [1]. The continuous population growth
requires more crop production to meet increasing food requirements of the popula-
tion. The effective utilization of water in agriculture is need of hour owing to limited
resources, climate change and depletion of ground water resources. Water conser-
vation is a major concern in developing countries where economies are primarily
driven by agriculture. The Indian irrigation mainly dependent on groundwater, facing
several challenges like over exploitation, poor water supply infrastructure and deple-
tion of groundwater resources. The country has a huge area of 39 mha irrigated with
groundwater, followed by China with 19 mha and USA with 17 mha. The country
has a requirement to conserve current water resources to meet urban, industrial and
agricultural needs of present and future [1].
Farmers are an integral part of society as 50% of the workforce in India depend on
agriculture for their livelihood. There are several challenges from routine activities to
harvest in this sector. The current era is of IT and varied technology provides solutions
for these challenges encompassing sensors, autonomous vehicles, automatic devices,
web services, mobile devices and IoT. In modern age, automation of the irrigation
processes using sensors and machine learning techniques is an eminent idea for
research.

1.1 Motivation

The IoT and WSN technologies link real-world objects to the internet through wire-
less and wired technologies. These objects are on the network can interact and
exchange the data aiming at enhancement in service, cost cutting in procedures
and saving manpower. Farmers have started to use computer and internet to orga-
nize data, monitor crop cultivation and to automate routine tasks. IoT is a dynamic
field to due rapid advancements in technology of sensors, communication and micro-
controllers. Therefore, this is an eminent area for research and gaining more atten-
tion from the researchers. There are many automated irrigation systems based on
evapotranspiration and sensors readings based on crop and geography of location.
Punjab is an agricultural state where those crops are grown which consumes a
lot of water. As a result, the water resources like surface water and groundwater are
not properly utilized. The automation of irrigation system using IOT techniques is
proposed in this paper to conserve groundwater according to the climate of state.

1.2 Literature Review

Irrigation is an accurate supply of water to plants for optimum plant growth in terms
of quantity and frequency. The plant growth and yield are influenced by the amount
A Proposal of Iot Based Evapotranspiration Irrigation System … 469

of water supply as the excess water harms the crops by making roots inefficient in
taking nutrients from the soil while lower supply causes slow seed germination [2].
The different ways to supply water are flood, drip and sprinkler irrigation. Many
of these procedures are manual and involve human labor to ON/OFF supply based
on estimation. The various pitfalls of these methods are: less supply, more supply
and late supply. These issues can be addressed in automated irrigation system. The
automation of irrigation systems is categorized as open loop and closed loop based
on control. The various inputs like length of watering, volume of water, start time,
end time and watering periods are provided by the operator of the system in open
loop. In a closed loop, a computer program recommends time and length of irrigation
[3].
Gutierrez et al. [4] proposed automated drip irrigation in Canada that saved 90%
more water as compare to manual irrigation. The system had two components such
as WSU (Wireless Sensor Units) to obtain sensor information and a WIU (Wireless
Information Unit) to transmit that information for analysis. The WSU consist of an RF
transceiver, sensors and a microcontroller. It was a wireless network of soil moisture
and temperature sensors situated in the roots of plants. The WIU had a GPRS module
that used public mobile network to transmit the sensor observations to a web server.
Giusti and Libelli [2] proposed a fuzzy decision support system based on soil water
predictions to improve a web based irrigation system. The system took the decisions
based on climate data (rain, temperature and solar radiation) and agricultural data
(soil composition, crop characteristics and site location).
Ghosh et al. [5] proposed an Arduino based automated irrigation system with
remote control from an Android smartphone. The system had user interface which
displays sensor data on smartphone and irrigation frequency and volume were
recorded through the android app. The results showed that the automated fields
consumed less water as compared to the manually irrigated fields. Hu and Shao [6]
provided a remote irrigation system with a cloud platform. The various parame-
ters such as soil temperature, humidity and CO2 concentration considered to take
the irrigation decisions. Bandara et al. [7] presented an artificial sprinkler irrigation
mechanism for green roof vegetation in Sri Lanka. It had predicted evapotranspiration
to determine crop water need.
Sivagami et al. [8] proposed an automatic irrigation system in a greenhouse for
both drip and sprinkler, which calculated the irrigation requirements based on soil
moisture and environmental conditions. Kwok and Sun [9] developed a machine
learning based irrigation system that utilized information like crop water require-
ments and sensor inputs. The system has two main modules: plant recognizing mobile
app to detect plant type by taking a photograph and an Arduino based irrigation
controller. Aydin et al [10] presented a technique to receive sensor data withy WeMos
D1 Mini microcontroller and sent it to MQTT server. Murthy et al. [11] proposed an
Irrigation Control that predicted the Maximum Allowable Runtime (MAR) for each
sprinkler zone with weather data acquired from imminent weather station and soil
surface runoff. An et al. [12] presented a sensor based automated irrigation system by
monitoring substrate temperature. The performance of system measured by consid-
ering the total number of leaves, the length and width of the leaves, chlorophyll
470 S. Walia and J. Sengupta

content in leaves and roots length. Munir et al [13] presented an approach where
decision dependent on two things KNN based ontology and sensor data.

1.3 Objective

The literature review presents many methods to automate irrigation systems like
IoT based systems, fuzzy decision support system, evapotranspiration based irriga-
tion scheduling and machine learning based irrigation system. Following a review
of the pros and cons of previous studies, the objective of this paper is to propose
an automated system using IoT that will use sensor inputs, evapotranspiration and
precipitation information from internet to schedule irrigation and to take decisions
about irrigation.
The paper consists of five main sections. The second section presents the back-
ground of the paper. Third section describes the proposed system. The fourth section
consists of results, fifth section contains conclusion and sixth section provides future
scope.

2 Background

This section presents main activities and a common framework of automated irriga-
tion system. The main activities of system are sensor data collection, data transfer,
data storage and analysis. The Fig. 1 describes the common framework of the system
having sensor nodes, base station and server. The base station collects data from

Fig. 1 Common framework of automated irrigation system

A Proposal of Iot Based Evapotranspiration Irrigation System … 471

sensor nodes and then transmits the obtained information to the server. The server
side has a remote user or web application to take the decisions based on analysis of
collected sensor data.
The automated system corresponds to various factors such as sensors used, number
and types of sensor nodes, communication, data storage and power sources. The main
activities and hardware components are discussed as follows:

2.1 Data Acquisition

On a base station, the data is collected from in-field sensor nodes made up from
various sensors such as soil moisture sensor, humidity and temperature sensor. The
base station uses appropriate internet technologies to send it to a cloud network or web
server. Gutierrez et al. [4] employed the use of web application for data collection
and analysis. Ghosh et al. [5] proposed the use of a cloud computing platform to
store and analyze data. Murthy et al. [11] used MQTT protocol to transfer zone wise
sensor data on web server implemented on AWS platform.

2.2 Data Storage and Analysis

After data collection, the various values obtained from sensors are analyzed against
some threshold values and then decisions are taken accordingly. Lenka and Moha-
patra [14] developed a hybrid system for crop-specific irrigation based on a feed
forward neural network and fuzzy logic-based decision support system (DSS). The
fuzzy logic-based DSS predicts soil moisture, which is delivered to the farmer with
SMS notification. The real-time agriculture data becomes the input and training
data to the feedforward neural network, which trained to take the decisions in fully
automated irrigation system.

2.3 Sensors Used

Sensors are used to collect readings of soil temperature, soil moisture, soil pH
and humidity, which are important for irrigation scheduling. These parameters are
determined by the type of crop and irrigation technique.
472 S. Walia and J. Sengupta

2.4 Microcontroller

A microcontroller is an essential component of an automated irrigation system that

collects data from sensor nodes, converts that data into digital form and then trans-
mits that data to cloud or web server. The various characteristics, such as low
cost, versatility to connect with sensor nodes and low power consumption taken
into account while selecting a microcontroller. Gutierrez et al. [4] used single chip
(PIC24FJ64GB004) 16-bit microcontroller programmed in C compiler to transfer
data.

2.5 Communication Protocols

The automated irrigation system needs two types of communication technolo-

gies: short-range communication from sensors to the base station and long-range
communication from the base station to the internet. The Internet Protocol Version
6 (IPv6), ZigBee and Near Field Communication (NFC) are used for low range
private area network communication while GPRS and Wi-Fi are technologies used
for communicating over the internet.

2.6 Power Sources

The power requirements of sensor nodes and base station are met by recharge-
able batteries, dry cell batteries and solar power. The low power consumptions are
always preferred to design a modest system, which can be used in real life situations.
Gutierrez et al. [4] used solar powered photovoltaic cell to supply power to a wireless
sensor unit (WSU) and rechargeable battery to supply power to wireless information
unit (WIU).

3 Proposed System

In this section, the system is proposed to automate irrigation processes using IoT
infrastructure and sensor network by considering the influencing factors such as
weather, surface runoffs, deep percolation and precipitation. The system takes advan-
tages of feedforward and feedback control mechanisms. The feedforward control
is focused on crop evapotranspiration to determine crop water requirements and
feedback control takes sensor readings to inquire about soil moisture in the fields.
The evapotranspiration, sensor feedback and precipitation information from online
weather forecasts are used to produce irrigation schedules. Figure 2 illustrates
A Proposal of Iot Based Evapotranspiration Irrigation System … 473

Fig. 2 Schematic diagram of proposed system

schematic diagram of proposed system where evapotranspiration, precipitation and

sensor readings are used as input parameters, and the outputs is irrigation schedules.
The system involves use of historical weather data, online weather forecast and
sensor feedback in various modules. The various modules of proposed system are
discussed as follows:

3.1 Evapotranspiration Prediction

The crop water requirements are determined from crop evapotranspiration (ETc ) that
in turn calculated with reference crop evapotranspiration (ETo ). The ETo is estimating
the evapotranspiration on the reference hypothetical crop of grass with height 0.12 m .
It depends on many factors such as humidity, temperature, atmospheric pressure, solar
radiation and wind speed. The present method is to predict ETo of upcoming days
from historical weather data of the region involves machine learning. The dataset is in
JSON format having hourly values of previous five years. The training part consists
of calculating reference evapotranspiration is calculated with the Hargreaves Samani
empirical method that uses the air temperature as main parameter. The equation of
method as follows:

ETo = α ∗ (Tmean + 17.8)(Tmax − Tmin )1/2 Ra (1)

where ETo is the reference evapotranspiration [mm day−1 ] is a measure of maximum

removal water from the surface with the processes of evaporation and transpiration
by the atmosphere. T mean is the mean daily temperature [°C] as T mean = (T max +
474 S. Walia and J. Sengupta

T min )/ 2. α is empirical constant having value 0.0023 and Ra is extraterrestrial solar

radiation [15].
The reference evapotranspiration is used to calculate the crop evapotranspiration.
The crop evapotranspiration is calculated from Eq. 2.

ETc = K c ∗ ETo (2)

ETc is crop evapotranspiration. K c is the crop coefficient of evapotranspiration

depends upon the type of crop and geographical location [15]. The system predicts
ETo of upcoming days with an artificial neural network which trained and tested on
past weather data [16].

3.2 Precipitation Information from Online Weather Forecast

It is important factor to decide timings, number of irrigation cycles, duration and

interval between successive irrigation. Punjab has semi-arid climate and having deep
water tables. It requires regular irrigation at small intervals to minimize the effect
of percolation and evaporation. The system collects weather forecast information of
next five days from internet.

3.3 Data Collection from Sensors

The hardware of proposed system consists of sensor nodes having soil moisture and
soil temperature sensors connected to ESP32 microcontroller. The ESP 32 has 32-
bit LX6 microprocessor that can operate at 160 or 240 MHz having Bluetooth and
Wi-Fi modules for data transfer act as base station to send data on Thingspeak cloud
platform. The sensor data is collected over cloud platform with Wi-Fi module for
analysis and decision making.

3.4 Data Analysis

It involves designing a system capable of generating irrigation schedules for crops

based on evapotranspiration, weather forecasts and sensor inputs. The irrigation time
and duration will be decided by the system. The steps of data analysis as follows:
1. Collect historical weather data and precipitation information of selected region.
2. Maintain the information in a dataset.
3. Using the Hargreaves Samani empirical method, calculate the crop evapotran-
spiration in dataset [15].
A Proposal of Iot Based Evapotranspiration Irrigation System … 475

4. Train and Test the model for prediction of evapotranspiration using Artificial
Neural Network [17].
5. Make a prediction of potential evapotranspiration for upcoming days.
6. Find out the crop irrigation requirements from potential evapotranspiration.
7. Classify the day as rainy, cloudy, partly cloudy or sunshine based on the amount
of precipitation.
8. The intervals between sensor readings and the duration of irrigation cycles are
determined by the type of day.
9. Make decisions using a rule-driven structure based on sensor inputs and
potential crop evapotranspiration.
10. Supply water according to need and generate irrigation schedule.
This system addresses irrigation water requirements in open fields and experi-
mental test will be conducted using flood irrigation in a small garden in Fatehgarh
Sahib, Indian Punjab.

4 Results

The results of various modules are presented this section. The hardware unit for
sensor data capturing and watering system is proposed. The main components of
system are Microcontroller, 5v DC Motor, L293 Motor Driver Module, Resistive Soil
Moisture Sensor and Breadboard. The soil moisture sensor continuously monitors
the moisture content of the soil and transmits readings to the microcontroller every
minute. The device is attached to the DC submersible motor through the L293 motor
driver module. When the moisture content is less than the threshold value motor
pumps water.
Figure 3 shows moisture data collected over a serial port which are represented
in graphical form in Figs. 4 and 5. The soil moisture sensor data retrieved in minute
(Time interval of 3 min) and second (Time interval of 3 s) intervals are plotted as
a graph in Figs. 4 and 5 respectively. The sensor reading was obtained with various

Fig. 3 Snapshot of soil moisture data from serial port

476 S. Walia and J. Sengupta

Fig. 4 Data retrieved from soil moisture sensor in time interval in minutes

Fig. 5 Data retrieved from soil moisture sensor in time interval in seconds

soil moisture conditions such as highly, moderately and low. The maximum and
minimum values of moisture content of soil in minute interval are 41.06 and 16.62%
respectively. The maximum and minimum values of soil moisture in second interval
are 68.9 and 0.29% respectively.
The weather forecast module retrieves weather forecast data from internet
through API. The data has various parameters like city_name, latitude, longitude,
temp_minimum, temp_maximum, wind speed, wind degree (direction), cloudiness
and rainfall. Figure 6 displays maximum and minimum temperatures in Celsius

Fig. 6 Minimum and maximum temperature from online weather forecast

A Proposal of Iot Based Evapotranspiration Irrigation System … 477

Fig. 7 Cloudiness percentage in online weather forecast

against the date and time received from online weather forecast. Figure 7 displays
the cloudiness in percentage against the date and time received from online weather
forecast.
Characteristics of Dataset to Predict ET0: The dataset mentioned in Sect. 3 for
prediction of ET0 has the following characteristics:
1. The dataset contains Historical Weather data of Punjab.
2. It includes parameters: city_name, temp_minimum, temp_maximum, pressure,
humidity, cloudiness, weather_description (Rainy, Clear, Foggy etc.), rain_fall.
3. The data set is hourly data the period of 2016 to sept 2020.
4. The value of ET0 in data is manually calculated with Hargreaves Samani
formula.
5. The dataset is divided into testing and training data.

5 Conclusion

Automated irrigation systems can save energy and human labor while also
contributing significantly to water conservation. In this paper, various irrigation tech-
niques available from literature have been explored in based techniques of hardware
deployment, sensors usage and background of study. The fundamental activities and
common infrastructure are presented in the consecutive section. By considering the
benefits and limitations of various techniques, an automated irrigation system based
on irrigation scheduling is proposed to obtain robust high performance system that
utilizes precipitation forecast, evapotranspiration and sensor information. The main
challenge for automated irrigation system is high deployment cost, power sources
and effected by uncertainties and disturbances such as climate change, weather uncer-
tainties, type of soil and salinization. In order to minimize these factors and improve
the response of the system are areas for future work.
478 S. Walia and J. Sengupta

6 Future Scope

The few advancements for automated irrigation system are listed as follows:
1. Cost: The cost involves in wireless sensor network set up and maintenance of
those networks. To minimize the deployment and maintenance cost will increase
scope of applications.
2. Control approaches in the presence of model uncertainties and disturbances:
These factors can interfere in system performance. To minimize these factors and
improve the response characteristics of the system are areas for development.
3. User Friendliness: These users of these systems are farmers, therefore user
friendly interface and closed loop operations thrust areas.

References

1. Water Ethics and the World Commission on the Ethics of Scientific Knowledge and
Technology, & Appelgren, B. (2004). Water in agriculture (5)
2. Giusti E, Marsili-Libelli S (2014) A fuzzy decision support system for irrigation and water
conservation in agriculture. Environ Model Softw 63:73–86
3. Obota ME, Inyama HC (2013) Soil moisture based irrigation control system for rice cropping
using wireless sensor network. Int J Eng Sci (IJES) 2(3):37–43
4. Gutierrez J, Villa-Medina JF, Nieto-Garibay A, Porta-Gandara MA (2014) Automated irrigation
system using a wireless sensor network and GPRS Module. IEEE Trans Instrum Meas 63:166–
176
5. Ghosh S, Sayyed S, Wani K, Mhatre M, Hingoliwala HA (2017) Smart irrigation: a smart drip
irrigation system using cloud, android and data mining. In: Proceedings of IEEE International
conference on advances in electronics, communication and computer technology, ICAECCT,
pp 236–239
6. Hu F, Shao L (2017) Design of remote irrigation system in farmland based on the cloud platform.
In: Proc. 29th Chinese Control Decis. Conf. CCDC 2017, pp 1125–1129
7. Bandara AGN, Balasooriya BMAN, Bandara HGIW, Buddhasiri KS, Muthugala MAVJ,
Jayasekara AGBP, Chandima DP (2017) Smart irrigation controlling system for green roofs
based on predicted evapotranspiration. In: Proceedings of electrical engineering conference
(EECon), pp 31–36
8. Sivagami A, Hareeshvare U, Maheshwar S, Venkatachalapathy VSK (2018) Automated
irrigation system for greenhouse monitoring. J Instit Eng (India) 99(2):183–191
9. Kwok J, Sun Y (2018) A smart IoT-based irrigation system with automated plant recognition
using deep learning. In: ACM international conference proceeding series, pp 87–91
10. Aydin O, Kandemir CA, Kirat U (2019) An artificial intelligence and Internet of things
based automated irrigation. In: International conference on computer applications in food and
agriculture, Konya
11. Murthy A, Green C, Stoleru R, Bhunia S, Swanson C, Chaspari T (2020) Machine learning-
based irrigation control optimization. In: BuildSys 2019—proceedings ACM international
conference syst. Energy-Efficient Buildings, Cities, Transport, pp 213–222
12. An SK, Lee HB, Kim J, Kim KS (2021) Efficient water management for cymbidium grown
in coir dust using a soil moisture sensor-based automated irrigation system. Agronomy
11(41):2021
13. Munir MS, Bajwa IS, Ashraf A, Anwar W, Rashid R (2021) Intelligent and smart irrigation
system using edge computing and IoT. Wiley Complex. https://doi.org/10.1155/2021/6691571
A Proposal of Iot Based Evapotranspiration Irrigation System … 479

14. Lenka SK, Mohapatra AG (2016) Neuro-fuzzy-based smart DSS for crop specific irrigation
control and SMS notification generation for precision agriculture. Int J Conver Computing
2(1):3–22
15. Cruz-Blanco M, Lorite IJ, Santos C (2014) An innovative remote sensing based reference
evapotranspiration method to support irrigation water management under semi-arid conditions.
Agric Water Manag 131:135–145
16. Adeyemi O, Grove I, Peets S, Domun Y, Norton T (2018) Dynamic neural network modelling of
soil moisture content for predictive irrigation scheduling. Sensors (Switzerland) 18(10):3408.
https://doi.org/10.3390/s18103408
17. Zanetti SS, Sousa EF, Oliveira VP, Almeida FT, Bernardo S (2007) Estimating evapotranspi-
ration using artificial neural network and minimum climatological data. J Irrig Drainage Eng
133(2):83–89
A Comprehensive Study
of “etcd”—An Open-Source Distributed
Key-Value Store with Relevant
Distributed Databases

Husen Saifibhai Nalawala, Jaymin Shah, Smita Agrawal, and Parita Oza

Abstract Distributed systems overcome various limitations of a centralized system

and offer several advantages like high performance, increased availability, and exten-
sibility at a low cost. Their rise is unprecedented, and it is only going to increase
hereon. But for a distributed system to be effective, it needs a consistent data store,
which would store all its necessary metadata and also serve as a single source of truth
for the entire system. These data changes need to be stored and to be communicated
quickly in a consistent manner across all the nodes in the cluster. It should have
fault tolerant capability and should be able to handle failures without any manual
intervention. One such open-source data store is etcd. In this paper, we presented
a comprehensive analysis of etcd along with its internal working and characteris-
tics. Comparative analysis of etcd with other relevant data stores is also presented in
the paper. The paper also sheds light on the application domains and research areas
related to etcd.

Keywords Distributed systems · Key-value store · etcd · Kubernetes · High

availability

1 Introduction

A database is a collection of data in a well-defined structure and should serve orga-

nizational purposes. Databases can be classified based on different factors like type

H. S. Nalawala · J. Shah · S. Agrawal (B) · P. Oza

Computer Science and Engineering Department, Nirma University, Gujarat, India
e-mail: smita.agrawal@nirmauni.ac.in
H. S. Nalawala
e-mail: 19mca005@nirmauni.ac.in
J. Shah
e-mail: 19mca035@nirmauni.ac.in
P. Oza
e-mail: parita.prajapati@nirmauni.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 481
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_35
482 H. S. Nalawala et al.

of content, application area, and underlying data structure. Some examples include
operational databases, data warehouses, cloud database, distributed database, graph
database, document-oriented database, spatial database, and temporal database. A
key-value store is one such database which contains a list of records, and each record
in turn contains different fields with data. Every record is identified by a unique
key which is used for storing and retrieving the data. Distributed system is a system
where the components are located on different computers attached through a network.
Components interact with one another towards a common goal, for solving a large
computational problem [1]. Kubernetes is a famous open-source container orches-
tration platform of distributed systems. It helps to manage containerized workloads
as they scale, by taking care of configuration, deployment, load balancing, service
discovery, and health monitoring across all the clusters. In order to achieve coordina-
tion amongst all its clusters and pods, Kubernetes needs a single source of information
which stores the current state of the entire system. To fulfil this need, etcd comes
into play. etcd serves as a single source of information and stores all the critical
information that are needed to keep Kubernetes or other similar distributed systems
running.
This paper provides a detailed review on etcd, which includes its characteristics,
internal working, and security mechanisms. This paper also compares etcd with other
open-source databases like ZooKeeper and Consul which serve the same purpose and
are open-source in nature as well. We discuss the pros and cons of each database,
which would ultimately help in deciding the appropriate tool as the situation demands.
Next, we draw our attention back to etcd and provide application domains where etcd
played a pivotal role.
The rest of the paper is structured as follows: Sect. 2 introduces etcd, its character-
istics, and security mechanisms, which is followed by a comparative study between
etcd and its contemporaries in Sect. 3. In Sect. 4, we present the research opportunities
and application domains of etcd and finally conclude the paper in Sect. 5.

2 Overview of etcd

etcd is an open-source, key-value database which holds critical information required

for proper functioning of distributed systems in a reliable manner. Fig. 1 represents
an etcd interface and its key-value database. The name etcd is derived from the “/etc.”
directory in the UNIX system, where all the configuration files are stored. “d” here
in etcd stands for “distributed” [2].
etcd can have a wide range of applications, but its most notable use case is as a
backbone to Kubernetes cluster. To ensure data consistency across all the distributed
nodes, Raft algorithm [4] is used. Managing the state of each node in a distributed
cluster is the most challenging task too. During state management, problems like race
conditions and network partitions may arise which need special care. In essence, a
distributed system needs a reliable coordinator who tracks the changes in each node
and communicates the same to the entire cluster in a timely and reliable manner [4]. It
A Comprehensive Study of “etcd—An Open-Source … 483

Fig. 1 etcd Key-value database [3]

stores information like data cluster configuration, state of the cluster, node existence
in the cluster, service state (running or not), pods state (running or not), and node
state (running or not) at any given point of time [5]. Other notable adopters of etcd
technology include ROOK—the open-source distributed storage system manager,
CoreDNS—an open-source DNS server, Openstack—which provides a set of soft-
ware components for cloud, and M3—which serves as Uber’s metric platform [2].
etcd is written in Go programming language, and its backbone is Cloud Native
Computing Foundation. Go has been around for quite some time and thus has excel-
lent community support behind it [2]. Some characteristics of etcd which make it
desirable for managing distributed workload are [4]:
• Highly Available—It manages hardware failures and network partitions easily
and is designed in such a way that there is no single point of failure.
• Fully Replicated—Each node in the etcd cluster can access the entire data store.
• Fast—Can perform up to 10,000 writes per second.
• Consistent—Returns the latest data value during each read operation.
• Secure—It supports automatic TSL and optional SSL for security. Additionally,
it is recommended to implement role-based access and apply the principle of least
privilege to limit the access to secure configuration data.
• Simple—Be it a simple Web application or a highly distributed system like
Kubernetes, any application can read–write using HTTP/JSON tools.

2.1 Working of etcd

This section discusses working of etcd and its functionality. The working of etcd is
presented using three aspects that are RAFT Consensus Algorithm, high availability
systems, and etcd clusters which graphically represent in Fig. 2.
484 H. S. Nalawala et al.

Fig. 2 etcd stages [1]

RAFT Consensus Algorithm

As discussed earlier, etcd is a highly available database system which uses RAFT
algorithm to give uptime of 99.99% [6–8]. etcd uses RAFT Consensus Algorithm
for communicating between etcd machines. There are three stages of a server in the
RAFT cluster—leader, follower, and candidate. By this, all the machines are fully
replicated, if the leader node goes down, then other machines or the follower nodes
can become candidate, and then, one of them is elected as leader node. Any candidate
receiving a majority vote becomes a leader. If the leader finds another leader or a
server with a higher term id, it becomes a follower again. Leader election is one of
the main aspects of RAFT. There are three scenarios which can occur during this
process mentioned as follows.
a. Winning the Election:
• Each node votes for only one candidate in a single term. The leader requires
a vote from the majority of nodes to win.
b. Receiving an < Append Entries Message > :
• If the term is > = to the current term of that node, recognize the other node
as the leader node. Transferal to the follower state.
c. No one is elected split votes (TIE):
• Each node times out and starts a new election. Timeout period is randomized.
This reduces the chances of split votes happening again.

High Availability Systems

Computing environments configured to provide nearly full-time availability are
known as high availability systems. This type of system typically has redundant
hardware and software that makes the system available regardless of failures. Well-
designed high availability systems circumvent having single points-of-failure. Any
A Comprehensive Study of “etcd—An Open-Source … 485

hardware or software component that can fail has a redundant component of the same
type. When failure occurs, the failover process moves processing performed by the
failed component to the backup component. This process remasters system wide
resources, recovers partial or failed transactions, and restores the system to normal,
preferably within a matter of microseconds. The more transparent that failover is to
users, the higher the availability of the system.
etcd clusters
etcd is an open-source distributed key-value database used to hold and manage the
critical information that distributed systems need to keep running. Most notably, it
manages the configuration data, state data, and metadata for Kubernetes, the popular
container orchestration platform. Starting an etcd cluster statically requires that each
member knows another in the cluster as shown in Fig. 3. In a number of cases, the IPs
of the cluster members may be unknown ahead of time. In this case, the etcd cluster
can be bootstrapped with the help of a discovery service [9, 10]. In a number of
cases, the IPs of the cluster peers may not be known before. This is common during
the usage of cloud providers or when the network is running through DHCP. In this
scenario, rather than stating a static configuration, it uses a present etcd cluster to
bootstrap a new one, called “discovery.” If no existing cluster is available, it uses the
public discovery service hosted at discovery etcd.io.

Fig. 3 etcd cluster service discovery

486 H. S. Nalawala et al.

2.2 Security Model of etchd

This model comes into effect when we want to allow our etcd cluster to communicate
directly to different clusters, servers or the clients. We will understand better with
four different scenarios.
First we have to create a CA certificate and a signed key pair for each member in
a cluster.
Scenario 1: Client-to-server transport security with HTTPS.
For this connection, we only need to have a CA certificate and a signed key
pair.
Scenario 2: Client-to-server authentication with HTTPS client certificates.
For this scenario, we have allowed client interaction directly with HTTPS
client certificates, clients will provide their certificates to the server, and then, the
server will validate that the certificate is signed and supplied by the appropriate
CA authority and then will decide whether to serve the request.
Scenario 3: Transport security & client certificates in a cluster.
etcd supports the same method as mentioned above for peer communication
in the same cluster, i.e. that all the communication between the members of the
cluster will be encrypted and authenticated using the client certificates.
Scenario 4: Automatic self-signed transport security.
etcd supports self-signed certificates where encryption is needed, but authen-
tication is not needed. This simplifies the deployment because there is no need
for managing keys outside of etcd.

3 Comparison of etcd with Other Similar Databases

Many such databases have been developed which serve the same purpose as etcd,
i.e. manage information across clusters in a distributed environment. Two most
commonly used databases are Apache’s ZooKeeper and HashiCorp’s Consul. These
databases are also compared with etcd in this section (see Table 1).

Table 1 Comparative study of etcd with relevant distributed databases

Properties etcd Zookeeper Consul
Latest version 3.4.0 3.6.2 1.9.1
Maximum database Several gigabytes Few megabytes to few Hundreds of megabytes
size gigabytes
HTTP/JSON API Yes No No
support
Multi-version Yes No No
concurrency control
A Comprehensive Study of “etcd—An Open-Source … 487

3.1 ZooKeeper

ZooKeeper is the oldest technology compared to others, and it is created with the aim
to enable coordination and control between Apache Hadoop Clusters. Hadoop is an
open-source framework which allows the storage and processing of large volumes
of data [11, 12]. It stores the data in a hierarchical key-value store, much like a file
system. It is used by companies like Reddit, Yelp, Yahoo!, Facebook, and Twitter.
Biggest advantages of ZooKeeper are that it is mature, robust, supports high avail-
ability, reliable, and it has a rich feature set. ZooKeeper performs much better in read-
heavy environments, where there are more read operations than write. ZooKeeper is
written in Java, which proves it is robust and reliable. ZooKeeper is quite complex
compared to its contemporaries, and as a result, it requires a deeper knowledge than
expected for the use case. The rich set of features it provides turns out to be more
of a liability than an asset. Further, ZooKeeper supports limited language binding
because it implements Jute RPC protocol, which is unique.
ZooKeeper is older than etcd, and it gave many lessons to the latter. Some of the
improvements etcd made over ZooKeeper include performing stable read and write
operation even under high load, a robust key monitoring mechanism which always
gives a notification prior to dropping any event, dynamic reconfiguration of cluster
membership, a multi-version data model for concurrency control, and a wider range
of language and framework support. It can be concluded that ZooKeeper paved the
path for services like etcd, which are considerably simpler and lightweight [13, 14].

3.2 Consul

Consul is a service discovery framework. Unlike etcd and ZooKeeper, consul imple-
ments service discovery framework within it (embedded), and so there is no need to
use a third party tool. Users only need to register their services and discover them
using HTTP or DNS. Similar to etcd, it implements a key-value database which is also
based on RAFT algorithm and APIs which can be used with HTTP/JSON. Consul
provides built-in facilities for failure detection, health check-up, and services related
to DNS. Data is stored in key-value format and can be retrieved using RESTful APIs
with HTTP. Some out-of-the-box features are service discovery, health checking,
multi-data centre, and secure service communication. etcd and Consul have different
aims and solve different problems. While etcd is a good choice for a distributed
key-value store with reliability being a chief characteristic, Consul is more aimed
towards end-to-end service discovery of a cluster. Also, etcd is preferred when the
system needs to be fault tolerant, while Consul provides an overall good infrastruc-
ture system for service discovery. In case of etcd, the fault tolerance mechanism
increases latency, while Consul is more prone to network failures [13, 14].
488 H. S. Nalawala et al.

4 Application Domains and Research Opportunities

Earlier section talked about etcd as a single source of information for the entire cluster
of nodes like Kubernetes. An analogy to understand this could be of a job scheduler,
which is currently executing some work. After finishing the work, the scheduler
would want to communicate the status of finished tasks with the entire cluster. Here,
it is important that this communication happens as quickly as possible to keep the
entire system in a stable state. etcd, being fault tolerant as well, presents itself as an
ideal choice for backend Kubernetes. Another popular use case of etcd is serving as a
backend of Chinese giant Alibaba Group. Since 2017, Alibaba Group has been using
etcd to store its key metadata information [15]. Note that they have taken advantage
of the open-source nature of etcd and optimized it to make it more efficient and serve
its need. Some of the modifications include using a dedicated machine to run etcd,
improving memory index, optimizing boltdb (It is used by etcd to store key-values;
it can be thought of as a database for databases.), and also client side best practices
like not creating key-values which change constantly, avoiding large values when
performing calculations and limiting number of lease objects created at a single
time [16]. They also created a maintenance and management platform called Alpha,
which manages etcd clusters effectively, and is a stop solution for different tasks. It is
based on an existing platform called etcd operator, but is designed to cater Alibaba’s
needs. The result of all these optimizations is that etcd could withstand even the most
extremes of network traffic, like the Double 11 shopping event of Alibaba, which
generated around 325,000 orders/second. We can see how the open-source nature of
etcd can be used to tweak its performance and serve the business needs [17].
As we know that etcd was developed to provide organizations with superior control
over their microservice architectures, but there have been some gaps lately as all
the focus of the development is shifted towards optimizing the etcd for working
with Kubernetes. Some aspects of developing the etcd for organizational work and
personal uses are lagging. We would like to point out some fields where it can be
improved, but before that we should know that etcd now is an open-source project
available freely on GitHub, so one can change it according to their needs. etcd was
initially created so that smaller companies could get data centre facilities like bigger
companies, but looking at the current time it provides open-source service all thanks
to Kubernetes.
• Optimization of RAFT for better latency.
• When there is a large amount of data pumped in directly to etcd node.

5 Conclusion

This study focused on the importance of etcd as a backbone of various distributed

systems. We started our discussion with a brief introduction to distributed systems,
and in particular, we discussed Kubernetes. In the following section, we introduced
A Comprehensive Study of “etcd—An Open-Source … 489

etcd, laid out its characteristics, and stated the reasons which make it suitable as a
backbone of distributed systems. The internal working of etcd was also presented in
this section. We also have compared etcd with other existing systems that serve the
same purpose. We also pointed out the research opportunities and areas where etcd
could improve itself. In a nutshell, the open-source nature of etcd makes it highly
flexible, and with proper configurations, it can withstand even the heaviest of the
loads. In future, we would like to undertake a detailed performance analysis of etcd
and other similar databases which serve the same purpose as etcd.

References

1. Tanenbaum AS, Steen M (2002) Distributed systems: principles and paradigms. In: Upper
Saddle River. Pearson Prentice Hall, NJ. ISBN 0-13-088893-1
2. etcd Authors (n.d.). etcd https://etcd.io/
3. AlibabaCloud. https://alibaba-cloud.medium.com/getting-started-with-kubernetes-etcd-a26
cba0b4258
4. Mizerany B (2014) About Etcd, the Distributed key-value store used for kubernetes, Google’s
cluster container manager the newstack. https://thenewstack.io/about-etcd-the-distributed-key-
value-store-used-for-kubernetes-googles-cluster-container-manager/
5. IBM Cloud Education (2019) etcd IBM. https://www.ibm.com/cloud/learn/etcd
6. Juggery L (2019) A Closer Look at Etcd: The Brain of a Kubernetes Cluster Medium. https://
medium.com/better-programming/a-closer-look-at-etcd-the-brain-of-a-kubernetes-cluster-
788c8ea759a5
7. Nangare S (2019) A Guide to Kubernetes Etcd: all you need to know to set up Etcd clusters
super user. https://superuser.openstack.org/articles/a-guide-to-kubernetes-etcd-all-you-need-
to-know-to-set-up-etcd-clusters
8. Kelverson E (2015) Getting started with etcd compose. https://www.compose.com/articles/get
ting-started-with-etcd/
9. Agrawal SS, Patel A (2019) CSG cluster: a collaborative similarity based graph clustering for
community detection in complex networks. Int J Eng Adv Technol 8(5):1682–1687. https://
www.ijeat.org/wp-content/uploads/papers/v8i5/E7475068519.pdf
10. Agrawal S, Patel A (2020) Clustering algorithm for community detection in complex network:
a comprehensive review. Rec Patents Comput Sci 13(1)
11. Desai R, Gandhi A, Agrawal S, Kathiria P, Oza P (2020) IoT-based home automation with smart
fan and AC using NodeMCU. In: Proceedings of ICRIC 2019. Springer, Cham, pp 197–207
12. Yadav S, Verma J, Agrawal S (2017) SUTRON: IoT-based industrial/home security and
automation system to compete the smarter world. Int J Appl Res Inf Technol Comput
8(2):193–198
13. Farcic V (2015) Service discovery: zookeeper vs etcd vs Consul technology conversa-
tion. https://technologyconversations.com/2015/09/08/service-discovery-zookeeper-vs-etcd-
vs-consul/
14. etcd Authors (2021) etcd versus other key-value stores etcd. https://etcd.io/docs/v3.3.12/lea
rning/why/
15. AlibabaCloud. https://www.alibabacloud.com/blog/fast-stable-and-efficient-etcd-perfor
mance-after-2019-double-11_595736
16. Masani KI, Oza P, Agrawal S (2019) Predictive maintenance and monitoring of industrial
machine using machine learning. Scalable Comput Pract Exp 20(4):663–668
17. Xingyu C (2020) Fast, stable, and efficient: etcd Performance After 2019 Double
AgriBot: Smart Autonomous Agriculture
Robot for Multipurpose Farming
Application Using IOT

Hari Mohan Rai, Manish Chauhan, Himanshu Sharma, Netik Bhardwaj,

and Lokesh Kumar

Abstract Internet of Things (IoT) is used all around the globe for connecting things
with each other. IoT is a term broadly used for devices that are connected to each
other via embedded sensors or with the use of wireless networks may be cellular or
Wi-Fi. The proposed system is basically an Agricultural Robot or “AgriBots” used
for increasing the productivity and quality of the crop and also to reduce the time and
labor cost. The system explains about the network of sensors and the applications of
different sensors in the agricultural fields. There are number of Agricultural Robots
that already exist at present but they are used at small scale only. In the existing
system, the monitoring of the parameters such as the soil moisture and temperature
are done by using the manual method. In the proposed system, the IoT is integrated
with the Arduino UNO to improve the efficiency of the agricultural fields. The details
collected by the Robot from the agricultural field will be stored on cloud and can be
monitored without any human interaction.

Keywords Agriculture robot · Seeding · Ploughing · Autonomous · Farming ·

Sensors

1 Introduction

Encouraging the modernization of the Agricultural industry has become more impor-
tant in the recent years. The agricultural industry has many challenges to overcome,
such as the increasing demands of food due to the rapidly increasing population
across the globe, to maintain and keep a check on the quantity and quality of the
food items and also the lacking interest of the young people in the Agriculture sector.
The development of robotics is spreading in every sector all around the world, as
robotics technology provides better implementation and improved quality for several
purposes [1]. In the upcoming years, there will be a boost in the number of “AgriBots”
or Agricultural Robots. Although the driverless tractors have already developed by

H. M. Rai · M. Chauhan (B) · H. Sharma · N. Bhardwaj · L. Kumar

Krishna Engineering College, Ghaziabad, Uttar Pradesh 201007, India
e-mail: manish.chauhan@krishnacollege.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 491
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_36
492 H. M. Rai et al.

the Engineers but this cannot replace the use of the human operated tractors fully as
this process is something which requires human assistance [2].
There are several application of IoT and it is being used in almost every application
such as, industrial internet of things (IIoT) for industries [3], internet of medical things
(IoMT) for medical sector [4], internet of vehicles (IoV) for vehicles [5], and internet
of everything (IoE) [6]. So, IoT application in agriculture sector also required for
its advancement. As we need more grains on the same agriculture land because the
population is increasing with very rapid rate, and it will be approximately 9.7 billion
at the end of 2050 [7].
The objective is to design an automated IoT enabled robot which can perform all
the farming operations without any manual intervention. Designing, implementing,
and employing a Robot (Agricultural Robot) in farm are considered as building a
whole Digital farming project [8]. Digitalization of farming refers to the automation
of each and every aspect of Farming. With the help of Robot, one can monitor each
and every activity going on in the field. There can be different Robots for a particular
activity or a Robot who can do multi-tasking. The implementation of Robots in
Agriculture is not at large scale in present. The robotics market is still at its early
stages of Agricultural development [9]. While the development of these AgriBots
(Agricultural Robots) the key points that are to be kept in mind are—the speed,
energy efficiency, accuracy [10], light weight, and the autonomous behavior of the
Robot.
In India, almost 60% of people are indulged in farming as their occupation [11].
The farmers tend to be the backbone of the Food production [12]. Till date, the most
common method used by farmers for farming is the Conventional method of farming
which is a time taking process and requires more manpower in the fields. The Robots
enable the farmers to monitor their fields virtually with the help of IoT. The Robot
can perform various processes like Seed sowing, spraying pesticides, ploughing,
harvesting, irrigation, and many other activities. During harvesting, the Robot can
pick up the right fruit at the right time based on the inputs provided to it and the
data is continuously passed to the farmer to monitor it [13]. Like in the proposed
system, the data is stored on the Cloud server and for its continuous monitoring. The
deployment of Robots in the farm will provide great help and ease to the farmers
and also provide right analysis in various aspects of farming like maintaining seed
planting uniformly will help in increasing the efficiency of seed sowing [14].
Agriculture is emerging as the high-tech industry today and many big companies
are involving into it [15]. The application of these Agricultural Robots is increasing
day to day because this provide effective solution with return of investment and also
this is very important in case of doing the activities that might result in risk of health
of the farmers. This will provide safety to farmers. Apart from various activities such
as Seed sowing, ploughing, irrigation, and others, the proposed system will check
the soil moisture with the use of moisture sensor and will water the plants whenever
required.
In this work, we have designed and implemented a robot which is capable of
performing several farming operations such as Seed sowing, ploughing, irrigation,
AgriBot: Smart Autonomous Agriculture Robot … 493

fertilizer spreading, and harvesting. The developed robot can be operated using Blue-
tooth, remote control, and also by feeding automatic data. The robot can be operated
through Wi-fi or using GSM module remotely and the data can be stored, analyzed,
and controlled through cloud.

2 Literature Survey

In recent past, several research has been presented by various researchers to improve
the agriculture sector by introducing smart vehicles or IoTs. Auat Cheein and Carelli
[8] presented an agricultural robot that provides unmanned service for a variety of
agricultural operations. In this paper, automated applications such as terrain leveling,
irrigation, weed detection robotics are discussed and analyzed. Ayaz et al. [16]
presented the use of IoT with wireless sensors to make agriculture smart and the
problem faced in integrating these devices with the classical farming process. In this
work, the authors also discuss the use of types of sensors to suit a variety of agri-
cultural applications such as seeding, harvesting, packing, transportation, irrigation,
etc. LeVoir et al. [17] proposed chip and high accuracy adaptive location sensing
technique for the self-governing farming robot in precision agriculture. Based on
low cost camera and global positioning system, the real time location is traced of
the rover. Bodkhe et al. [18] have presented a detail study of blockchain based irri-
gation system, its challenges, and opportunity. In this paper, authors have surveyed
the advantages, disadvantages, application of blockchain in precise irrigation and
also regarding security related to the payment transaction between stakeholders.
They have also surveyed and proposed the integration of blockchain with precision
irrigation which will secure the sensors of the connected IoT devices.
Aashish et al. [19] reviewed the overall planning, development, and the fabrication
of the Agrobot or the Agriculture Robot that works with the help of battery or any
other alternate energy supply. The Robot is able to dig the soil, level the soil, water
spray the field and fertilizer separately. The autonomous vehicle has gained the rapid
interest in the agricultural field. Almost 40% of the population chooses Agriculture
as their fir occupation. The system consists of a Relay switch and Bluetooth. As this
language is familiar to most people so it will be easy to interact with the Robot. The
Robot uses solar energy to work. The advantage of the Robot is that it is automatic
hence, hands-free, and fast input operations of the data. The Robot intends to execute
all the essential functions required to be conducted within the farms. Gonzalez-De-
Santos et al. [7] proposed a smart farming vehicle called Unmanned Ground Vehicle
(UGV). The work discusses the design and development of automated convention
vehicle robots and the design and development of mobile platforms. Two types of
robots analyzed, wheeled and wheeled, and their adaptability to crops and terrain are
discussed. Cihan Yurtseveri et al. [12] developed and presented a robot in farming
with row-like seeding operation. It has four different types of sub modules for digging,
irrigation, wheel movement, and seeding. All of the Robot’s mechanical and electrical
characteristics are determined by the architectural requirements, while the device
494 H. M. Rai et al.

control and interface are determined by stakeholder preferences. In their proposed

model, they have assembled and integrated all four subsystem into the robot. They
have verified the model using test result and practical performance based on all
proposed four types of sub modules. Oktarina et al. [20] have proposed a harvesting
robot in their pilot project for tomato plant. In their project, they used to select the
tomatoes based on their color and sizes. They have used image processing technique
for classing and proving command based on that to robot for selecting the tomatoes.
In their agriculture robot, they have also classified the tomatoes based on their sizes.
So in their pilot, project they have built and design a robot which can automatically
detect the, picked and classify the types of tomatoes based on their color and sizes.
Poonguzhali and Gomathi [21] implemented and designed a IoT based agricul-
ture robot for seeding and ploughing application. In this work, Wi-Fi technology
has been used to control robotic vehicles and farming is also done without human
intervention of manpower. Three main functions have been implemented, automatic
sowing, tillage, and sprinkling of water based on time intervals. Bhupal Naik et al.
[14] presented automated IoT based smart farming methods, such as highly rough
yield control and data collection task. In this work, various types of sensors are used
to extract information about the crop and farm and crop monitoring are also done
using these sensors. The information extracted through the sensor is stored on the
Internet and based on that, crop and field forecasts are made. Patel et al. [22] have
presented a method using 5G network for improving the supply chain of agriculture
food based on block chain technology. The proposed method is a credit based system
which helps farmers and also guarantees transparency and trust in the application
of food chain supply. The presented model provides better result as compared to
traditional schemes.
Although a lot of research has been done on agricultural robots in recent years,
there is a lot of room for improvement. Very little work has been done on automated
agricultural land selection and updation using Land Rover. Most agricultural robots
are not fully automated and at the same time they are not capable of doing all the
agricultural tasks alone without human intervention.
In this paper, we have developed an agricultural robot (AgriBot) which is capable
of performing all the agricultural task such as seed sowing, harvesting, irrigation,
soil digging, fertilizer spraying, and many more. The proposed AgriBot is fully
automated with multiple operation and can be handled without human intervention
using Bluetooth, remote, and wi-fi module.
The paper is structured as: Sect. 1 presents the motivation, objective, and general
discussion. The literature survey is conducted on the recently published work in
Sect. 2. Section 3 deals with modeling and analysis where the hardware components
used are described. The proposed methodology used for this work is presented in
Sect. 4. Section 5 describes the results and discussion on the proposed work and
conclusion and future work are summarized in Sect. 6.
AgriBot: Smart Autonomous Agriculture Robot … 495

3 Modeling and Analysis

3.1 Hardware Components

Arduino UNO, NodeMCU ESP8266, General Purpose PCB, DC Motor, Motor Pump,
Moisture Sensor, Vibrator, Relay, Capacitor, Batteries, Connecting wires, Water
tubes, Platform and wheels (for Structure).
Arduino UNO. Arduino UNO is the backbone of our prototyped model which has
been implemented for multiple agricultural operations. The power supply, connection
of different types of motors for various operation has been connected in Arduino
Uno. It is used to control all the operation of proposed robotic vehicle with software
embedding. It is an open-source and flexible electronic platform prototype which
enables easy use of both hardware and software. The microcontroller Arduino UNO
is based on ATmega328P. It consists of 14 digital Input/Output pins, 6 analog pins, a
USB connection, and a reset button. Table 1 shows the technical specifications and
components used for implementation of this work.
NodeMCU ESP8266. The Arduino UNO board has been connected on the
NodeMCU ESP8266 is a development board which is used to combine control of
Inputs/Outputs and the Wi-Fi capabilities. It is affordable to use and an open-source

Table 1 Technical
Name of components Description
specifications of Arduino Uno
Microcontroller ATmega328P
Operating voltage 5V
Input voltage (recommended) 7–12 V
Input voltage (limit) 6–20 V
Digital I10 Pins 14
PWM digital 1/0 Pins 6
Analog input pins 6
DC current per I/0 Pin 20 mA
DC current for 3.3 V Pin 50 mA
Flash memory 32 KB (ATmega328P)
SRAM 2 KB (ATmega328P)
EEPROM 1 KB (ATmega328P)
Clock speed 16 MHz
Length 68.6 mm
Width 53.4 mm
Weight 25 g
496 H. M. Rai et al.

Fig. 1 a Arduino Uno. b NodeMCU ESP8266

IoT platform. We have used Node MCU and Arduino Uno for connecting and imple-
menting all other components. Figure 1 shows the pictorial view of Arduino Uno
and Node MCU ESP8266 used in this work.
General Purpose PCB. Since our developed AgriBot is a prototype model, for its
implementation instead of designed PCB we have used general purpose PCB. Printed
Circuit Boards (PCBs) are used to support the electronic components mechanically
and connect them electrically using the conductive pathways, tracks from copper
sheets laminated onto the non-conducting substrate.
DC Motor. There are several DC motor that has been utilized for different agricultural
operations such as for digging, for harvesting, for seed sowing, for ploughing, etc. DC
Motors converts the electricity into motion by utilizing electromagnetic induction.
By using a variable supply of voltage or by changing the strength of current in its
field windings, it is possible to control the speed of the DC Motor.
Motor Pump. A Motor pump is used to transfer or move any type of liquid/gases
from one place to another with the help of mechanical action. The pump converts
the motor’s energy into mechanical flow of liquid.
Moisture Sensor. The moisture sensor is used to measure the volumetric content of
water present in the soil. The Moisture sensor works as the resistance is inversely
proportional to the moisture present in the soil. If there is more water present in the
soil better will be the conductivity, hence, results in low resistance and vice-versa.
Vibration Sensor. Vibration sensor is a coreless motor and is compact in size. It has
magnetic properties and is small sized. Based on these features, its performance is
highly consistent. Here, it is used for seed sowing, the vibrator vibrates and the seeds
fall on the desired place.
Relay. A Relay is a switch that connects or disconnects two circuits and is electrically
operated. The Relay or switch has a set of input terminals for controlling single or
multiple signals and also a set of operating contact terminals. Small electrical stimuli
are converted into larger currents with the help of Relay. A Relay may a different
AgriBot: Smart Autonomous Agriculture Robot … 497

number of contacts in many different forms, such as make contacts, break contacts,
or combinations thereof.
Batteries. Batteries are used to give power supply to the Robot. They convert chem-
ical energy to electrical energy. In this project, we have used rechargeable lithium ion
battery which is used to enable the Node MCU, Arduino Uno, and other necessary
components. Connecting wires are used for the connections between the different
components of the Robot. As the project is a prototype device hence we have used
connection wires for connecting all components with each other. Also, there are the
structural components used in designing the Robot.

3.2 Software Components

Software is one the important section of this project, some software used in this
project are; Arduino IDE, Embedded C, OrCAD design, and Cloud server.
Arduino IDE. The Arduino Integrated Development Environment (IDE) is used to
program Arduino in Java programming language. It is common to all boards and
can run both offline and online. It contains a text editor where the code is written,
a message area, console, common functions on toolbar with buttons and series of
menus. The Arduino IDE connects the Arduino and the hardware to get the programs
uploaded and communicate among them.
Embedded C. Embedded C is the name given to the programming language C
associated with hardware architecture. Embedded C is the extended version of the C
programming language with some additional header files.
OrCAD design. OrCAD software is used to create electronic semantics, to perform
mixed signal simulation and electronic prints to manufacture printed circuit boards.
It is used by electronic engineers or technicians for various projects.
Cloud Server. The data monitored is then sent to the cloud server with the help of the
Wi-Fi module embedded on the Arduino UNO. There are many cloud platforms that
are free source and are available on the internet to which the data is sent to the cloud
using the Wi-Fi module and the users can monitor the output i.e., the volumetric
water content present in the soil.

4 Proposed Methodology

The main problem in the modern era in the field of Agriculture is that it lacks the
manpower i.e., the availability of farm labor, lack of knowledge about the field and
soil, increase in the labor wages, and the wastage of resources and raw material
required for farming. The innovation of the Agricultural robot is used to overcome
498 H. M. Rai et al.

Fig. 2 Block diagram of the proposed system

these disadvantages. The use of Robotics technology is new in the field of Agriculture,
the opportunities have enhanced the Robots. The Robots perform various operations
such as Ploughing, Seed sowing, Irrigation, and Harvesting. Figure 2 shows the
block diagram of the proposed system. This has enabled the farmers to reduce the
environmental impacts, increase the accuracy and efficiency and to manage each
individual plant in an efficient manner.
This autonomous Robot can go the Agricultural land through the crop lines and
can easily perform the tasks that require more energy and are really tiring, also might
be unsafe for the farmers. The structure of the Robot is designed in such a way that
the Robot can perform different tasks such as seeding, plug system easily even on
the uneven land. The wheels of the Agribot are designed so it can travel easily even
when the soil is wet or uneven.
The measurements of field collected are provided to the robot through the farmer’s
mobile, data from mobile will be sent to the cloud server from which the data will
be passed to the Arduino with the help of NodeMCU and based on these readings or
measurements the Robot will follow the path in the field.
AgriBot: Smart Autonomous Agriculture Robot … 499

4.1 Ploughing Operation

Ploughing is done using cultivator, the farmer will provide field measurements as
Input through his/her mobile phone and that data will go to the cloud server, then
from the cloud server the data will come to the Robot via NodeMCU and cultivator is
attached to the Robot. Robot will then follow the path according to the measurements
given by farmer.

4.2 Seed Sowing Operation

Farmer will fill the seed dispenser with seeds (whichever he/she wants to grow in
the field), again the Robot follows the path according to measurement provided by
farmer, there is a narrow hole at the bottom of seed dispenser and then when the
vibrator vibrates the seeds will fall on the desired location.

4.3 Irrigation Operation

Irrigation is done based on the Output given by moisture sensor, the data from mois-
ture sensor goes to NodeMCU ESP8266. NodeMCU will then communicate to the
cloud server and will send the data to the cloud, from cloud the data will come to
the water pump controller which will then take a decision based on the previously
feed values to decide whether to Switch ON the pump or not. The type of Irrigation
method used in the proposed system is Drip irrigation system which will conserve
water. The water flows through the water tubes in the whole field. The motor will
Turn OFF after some fixed amount of time. The robot possesses a humidity sensor
and temperature sensor which will record the temperature and humidity of the field
and send the data using the NodeMCU to the cloud server and from the cloud that
data will go to the mobile phone of farmer. Based on this data the farmer can decide
which crop will be good for the field as different crops require different necessary
conditions to grow.

4.4 Harvesting Operation

Harvesting is done using Straw walker for wheat like crops and Corn head for core
like crops. A DC motor is used to rotate the Straw walker and Corn head work simply.
In this mode also the robot will follow the path based on the measurements only.
500 H. M. Rai et al.

5 Results and Discussions

The proposed system performs the following functions—Ploughing, seed sowing,

harvesting, and irrigation. All these functions are performed with the help of different
sensors. The temperature and moisture sensors are used to collect the data from the
surroundings and send it to the cloud server which then goes to the Robot and then
the further action takes place and the Robot sprinkles the water in the field. Then
comes the Ploughing mechanism, especially before the seed sown are loaded into
the module. Some factors are pre-defined according to the type of seeds that are to
be sown into the field such as the depth of the ground to be dug and the distance
to be maintained. Figure 3 shows the designed agricultural robot with ploughing
mechanism.
The DC motor is used for the movement of the Robot which is controlled by
the software modules. The Agricultural robots can easily and efficiently perform
the tasks like Ploughing, seeding, and ground leveling. Figure 4 visualizes the seed
dispenser used in AgriBot for performing the seeding operation.
The traditional method of seed sowing is much time taking process. So, agricul-
tural robots are useful for sowing the seeds in the field over the traditional method.
With the help of the AgriBot, seed can be sown throughout the field at the right place
with much more accuracy. For this, geographical mapping also plays an important
role when used for large scale purposes. Both the Robotics and geographical mapping
work well together in developing an autonomous Agriculture Robot. AgriBot with
harvesting mechanism for crop cutting operation can be seen in Fig. 5.
Harvesting is an important activity of Farming or Cultivation and can be auto-
mated with the help of Robotics technology and using a combine harvester. Although
in some cases there may be a need of using the manual harvesting process but
for collecting the apples, grapes, and some other fruits and crops, the autonomous
Agriculture Robot is a great approach.

Fig. 3 Designed AgriBot with Ploughing mechanism

AgriBot: Smart Autonomous Agriculture Robot … 501

Fig. 4 (e) AgriBot with seed dispenser for seed sowing operation

Fig. 5 AgriBot with

harvesting mechanism for
crop cutting operation

In the Irrigation process, the moisture sensor is used to observe the environmental
conditions. Based on the observations given by the moisture sensor, the Robot takes
the further action that whether to spray water to the plants or remain in steady state.
The irrigation process is somewhat same as that of spraying the fertilizers, but the
only difference is that every crop has its own requirement of fertilizers, as some
need in the initial phase when the seed is germinating and some when the seed starts
developing in the field.
502 H. M. Rai et al.

6 Conclusion and Future Scope

The proposed system provides an innovation in the field of Agriculture and Agribusi-
ness. Also, in the money related angle it is very useful for the farmers. This also
reduced certain kinds of monotonous work in the Agriculture sector and thus empha-
size that many more young individuals to opt for farming as their occupation. When
comparing the work done by the Robot and the manpower used in doing the same
work, it can be concluded that the average cost for the Robot’s maintenance is less
than the total average cost in maintaining the human being including all the fringes.
Also, the productivity and the quality of the food increased. The use of Robots has
improved the management control as they carry pre-planned procedures with great
accuracy. The Robots meet all the Occupational safety and the Health Administration
standards. In the proposed system, we constructed a multipurpose Robot or Agri-
culture vehicle that can perform different mechanism like Ploughing, seed sowing,
irrigation, harvesting, and many more other mechanisms can also be added it in. To
increase the use of Agricultural Robots, the old barriers of limiting their use should
break up now to move to a more advance level in the field of Agriculture.
In future scope, these Agriculture based Agribots can be enhanced in many other
ways like Nursery bot, Herder bot, Wine bot, Bee bot, and Hamster bots. This would
qualify the standards from the current precision to autonomous farming methodolo-
gies. Apart from the Agricultural sector, the Agricultural Robots are also used in the
Horticulture and also for Managing and monitoring of the environment.

References

1. Sowjanya KD, Sindhu R, Parijatham M, Srikanth K, Bhargav P (2017) Multipurpose

autonomous agricultural robot. In: 2017 International conference of electronics, communica-
tion and aerospace technology (ICECA). IEEE, pp 696–699 https://doi.org/10.1109/ICECA.
2017.8212756
2. Krishnan A, Swarna SS, BH (2020) Robotics, IoT, and AI in the automation of agricultural
industry: a review. In: 2020 IEEE Bangalore humanitarian technology conference (B-HTC).
IEEE, pp 1–6. https://doi.org/10.1109/B-HTC50970.2020.9297856
3. Boyes H, Hallaq B, Cunningham J, Watson T (2018) The industrial internet of things (IIoT):
an analysis framework. Comput Ind 101:1–12. https://doi.org/10.1016/j.compind.2018.04.015
4. Khan SR, Sikandar M, Almogren A, Ud Din I, Guerrieri A, Fortino G (2020) IoMT-based
computational approach for detecting brain tumor. Futur Gener Comput Syst 109:360–367.
https://doi.org/10.1016/j.future.2020.03.054
5. Priyan MK, Devi GU (2019) A survey on internet of vehicles: applications, technologies,
challenges and opportunities. Int J Adv Intell Paradig 12:98. https://doi.org/10.1504/IJAIP.
2019.096957
6. Majeed A (2017) Developing countries and Internet-of-Everything (IoE). In: 2017 IEEE 7th
annual computing and communication workshop and conference (CCWC). IEEE, pp 1–4.
https://doi.org/10.1109/CCWC.2017.7868459
7. Gonzalez-De-Santos P, Fernández R, Sepúlveda D, Navas E, Armada M (2020) Unmanned
ground vehicles for smart farms. In: Agronomy—climate change and food security. IntechOpen.
https://doi.org/10.5772/intechopen.90683
AgriBot: Smart Autonomous Agriculture Robot … 503

8. Auat Cheein FA, Carelli R (2013) Agricultural robotics: unmanned robotic service units in
agricultural tasks. IEEE Ind Electron Mag 7:48–58. https://doi.org/10.1109/MIE.2013.225
2957
9. Ponnusamy V, Natarajan S (2021) Precision agriculture using advanced technology of IoT,
unmanned aerial vehicle, augmented reality, and machine learning. Presented at the (2021).
https://doi.org/10.1007/978-3-030-52624-5_14
10. Barman A, Neogi B, Pal S (2020) Solar-powered automated IoT-based drip irrigation system.
Presented at the (2020). https://doi.org/10.1007/978-981-13-9177-4_2
11. Arjun KM (2013) Indian agriculture- status, importance. Int J Agric Food Sci Technol 4:343–
346
12. Yurtsever C, Ertas Y, Sustam O, Ulu C (2020) Development of an Agriculture Robot for
Row-type Seed Sowing Applications. J Selcuk Spec ICAT’ 20:228–242
13. Apalak MK, Karaboga D, Akay B (2014) The artificial bee colony algorithm in layer optimiza-
tion for the maximum fundamental frequency of symmetrical laminated composite plates. Eng
Optim 46:420–437
14. Bhupal Naik DS, Ramakrishna Sajja V, Jhansi Lakshmi P, Venkatesulu D (2021) Smart farming
using IoT. Presented at the (2021). https://doi.org/10.1007/978-981-15-9516-5_34
15. Baylou P (1987) Agricultural robots. IFAC Proc 20:111–119. https://doi.org/10.1016/s1474-
6670(17)55251-9
16. Ayaz M, Ammad-Uddin M, Sharif Z, Mansour A, Aggoune EHM (2019) Internet-of-Things
(IoT)-based smart agriculture: toward making the fields talk. IEEE Access 7:129551–129583.
https://doi.org/10.1109/ACCESS.2019.2932609
17. LeVoir SJ, Farley PA, Sun T, Xu C (2020) High-accuracy adaptive low-cost location sensing
subsystems for autonomous rover in precision agriculture. IEEE Open J Ind Appl 1:74–94.
https://doi.org/10.1109/ojia.2020.3015253
18. Bodkhe U, Tanwar S, Bhattacharya P, Kumar N (2020) Blockchain for precision irrigation:
opportunities and challenges. Trans Emerg Telecommun Technol. https://doi.org/10.1002/ett.
4059
19. Kumar AAA (2020) Review on multipurpose agriculture robot. Int J Res Appl Sci Eng Technol
8:1314–1318. https://doi.org/10.22214/ijraset.2020.5209
20. Oktarina Y, Dewi T, Risma P, Nawawi M (2020) Tomato harvesting arm robot manipulator; a
pilot project. In: Journal of physics: conference series. Institute of Physics Publishing, p 12003.
https://doi.org/10.1088/1742-6596/1500/1/012003
21. Poonguzhali S, Gomathi T (2021) Design and implementation of ploughing and seeding of
agriculture robot using IOT. Presented at the (2021). https://doi.org/10.1007/978-981-15-7394-
1_58
22. Patel N, Shukla A, Tanwar S, Singh D (2021) KRanTi: Blockchain-based farmer’s credit scheme
for agriculture-food supply chain. Trans Emerg Telecommun Technol. https://doi.org/10.1002/
ett.4286
Mobile Application Voting System:
A Means to Achieve a Seamless Election
Process in Developing Countries

Abidemi Emmanuel Adeniyi, Roseline Oluwaseun Ogundokun,

Sanjay Misra, Jonathan Oluranti, and Ravin Ahuja

Abstract Voting is a concept used to describe the part of the election process. It is a
means by which the citizens choose who to lead them for a designated period. There
is various type of manual and electronic voting processes currently in use. Manual
voting processes have become a tool by which government bodies in Nigeria and
other African countries at considerable take advantage of to push unworthy people
into power. The Nigeria voting system is a typical example of this misfortune, where
voters are subjected to long queues before they can perform their legal duty as a
citizen. This existing system is faced with numerous challenges such as hooliganism
where glorified thugs snatch ballot boxes and disrupt the peace and tranquillity of the
voting process. Therefore, a loyal citizen who is bound to vote is unable to perform
their legal duty, leading to the manipulation of results and other voting crises. This
research proposed a mobile voting platform to deal with the challenges as mentioned
earlier associated with a manual voting system that is ineffective and inconvenient
for citizens. The proposed system will improve how the election is being conducted
in Nigeria and other countries that are faced with similar challenges in the voting
process. The scheme aims to allow eligible voters with registered voters card (PVC)
in Nigeria and diaspora to cast their votes in their respective places of residence as
long as the mobile application is accessible on their mobile devices which will be

A. E. Adeniyi · R. O. Ogundokun
Department of Computer Science, University of Ilorin, Ilorin, Nigeria
e-mail: adeniyi.emmanuel@lmu.edu.ng
R. O. Ogundokun
e-mail: ogundokun.roseline@lmu.edu.ng
S. Misra (B)
Department of Computer Science and Communication, Ostfold University College, Halden,
Norway
e-mail: sanjay.misra@hiof.no
J. Oluranti
Centre for ICT/ICE Research, Covenant University of Technology, Otta, Nigeria
e-mail: jonathan.oluranti@covenantuniversity.edu.ng
R. Ahuja
Shri Viswakarma Skill University, Gurgaon, Hariyana, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 505
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_37
506 A. E. Adeniyi et al.

available on various versions such as Android, iOS, Windows operating system. Each
voter’s details will be secure through the use of various cryptographic techniques and
verified with the use of one-time password during the voting process. This process
will make the election process flawless, efficient, convenient, secured and timely
in the area of result compilation and final verdict. Also, the system will eliminate
violence and result in manipulation.

Keywords Voting system · Election process · Mobile application · Mobile voting

system · PVC

1 Introduction

In a democratic form of government, voting is the legal right of the citizen to choose
who they want to lead then for a specific period. The event which takes place every
four years known as the electoral process is a period preceded by active preparation
by members of a registered political party under the rule of law of the government.
In most developing countries, many people rally around prominent people who can
afford a ticket to contest for office; these individuals gain the support of a political
party which gives them the opportunity to be further showcased to the entire country.
Unlike the military government where commanders are appointed based on ranks
and professional achievement, people are ruled based on a set of principles that
may be somewhat difficult or unbearable for members of society. This gave birth to
democracy which gives the people the authority to choose their governing legislation,
it is a pure representation of which people are and how power is distributed among
them are the base to the democratic development and constitution.
In most cases, how people choose who governs them is based on some apparent
factors which span through correctness, fairness, a sense of purpose and achievement,
and more importantly, the future, and this is an essential requirement to integrity in
the electoral process. However, many individuals are biased in the opinion, and
this set of people mostly wants their candidate to emerge triumphant in an election
process. Voting technology has essentially remained stagnant for long [1]. Neverthe-
less, certain aspects stayed the same even longer than anticipated, such as a public
ballot where voters signed their signatures on a general vote under the name of one
candidate or another. This approach had the benefit of being easy to count and diffi-
cult to falsify, but the procedure was best suited for smaller elections. Then came
along the printed ballot as the government grew and parties in Nigeria became more
organized, and printed ballots became more popular. Each group will distribute multi-
page ballots with the names of the officials running for the different election offices.
Voters would take the vote to be counted from the party they chose and drop it in the
ballot box.
The cost of modern voting machines is a recurrent barrier for states trying to
boost [2]. For instance, in Nigeria, the price for acquiring a card reader is higher,
and most of these machine is not adequately designed for electoral purposes which
Mobile Application Voting System: A Means to Achieve … 507

makes it more challenging to get an accurate compilation of results at the end of an

election. It is quite evident that this method has only made the voting process less
secured, as most of these devices fail on the field when people are trying to cast
their votes. The findings of the analysis carried out on the existing voting system
in Nigeria using the Hypertext processor Web platform and Google Android mobile
application in the native language of people show impressive outcomes. In future
elections in Nigeria, it will ensure a high level of citizen participation and free,
equal, open, convenient and confidential electoral processes [3]. The need for a
multifactorial authentication solution, such as one-time passcode (OTP) for security
authentication, plays an essential role in this project, and it is our main focus. Using
mobile as voting mechanisms can resolve most of the above worries. If properly
implemented, the remote votes on handheld phones could improve voter turnout,
minimize election management costs and enable voters to engage with recognizable
technologies.
Although voters may not be able to visit voting booths physically for several
reasons, they may need to vote remotely, for example, from home or when travelling
abroad. Hence, there is a great demand for simple, straightforward, and, most impor-
tantly, safe remote voting procedures. Today, the most popular way to vote remotely
is by postal where electors cast their ballots by mail. This ignores adequate authen-
tication, however, and requires a time-consuming process. Internet voting has been
implemented to allow for more versatility [4]. The voting process should remain
secrete when voting takes place, and no individual vote or authority should be
connected to the voter who casts a vote. This is essential even if the voter is not
aware of the security concerns, this may raise. In this work, we look to improve
the way people vote seamlessly in Nigeria and hope this method can be adopted in
various sectors of the world. The current system consists of techniques such as paper
ballots, voting machine booster, punch card and voting machine visual. The biggest
problem with the new system was a time-intensive one that used to require a lot of
time to register. In the new scheme, the paper-based polling process was used, which
also yielded the results of bogus polling [5].
The elector gets a blank ballot and uses a pen or marker to indicate that he wishes
to vote for which candidate. Hand-counted polls are a time-consuming and labour-
consuming process, but paper ballots are easy to make, and the votes can be retained
for verification, this type is still the most common way of voting. Appropriate training
will need to be provided to the polling duty supervise associates [5]. During polling
day, the staff fellows concerned are requisite to be present half an hour before the
polling booth/station opening to check that all arrangements are being made correctly.
After voting, another group of officers will take care of counting the ballots [5]. The
process can prove tedious, error-prone and costly with all these phases, groups and
protocols involved. However, some technological implementation currently in the
Jamaican environment makes the process semi-manual, but this is far from what
a completely ICT-driven process might accomplish. The semi-manual method only
requires the government to retain information about electors in a database that can be
accessed on a computer at the voting day to enable quicker searches [5]. Electronic
voting (EV) means using a machine instead of using a conventional vote at voting
508 A. E. Adeniyi et al.

centres or via postal mail [6]. It includes the process of making a decision or voicing a
view about a group such as an assembly of an electorate, usually after consultations,
arguments, or election campaigns [1]. This combines several forms of polling, such as
kiosks, the Internet, telephones, punch cards and mark-sense or optical ballots. India,
as the largest democracy in the world with a 1.1 billion [7] population, developed
electronic voting machines (EVMs). Embraced and sponsored by-election electors
will address issues related to the conventional paper-based voting system. In 2003,
Estonia’s experience with the implementation of online voting in the USA and about
3.4% of electors were able to use mobile e-voting, and by 2007 mobile e-voting
elections were proven healthy amid questions about hacker attacks, identity theft
and abuse of voting counts [8].
This study therefore proposed a mobile voting platform to deal with the challenges
as mentioned earlier associated with a manual voting system, that is, ineffective and
inconvenient for citizens. The proposed system will improve how the election is
being conducted in Nigeria and other countries that are faced with similar challenges
in the voting process. The scheme aims to allow eligible voters with registered voters
card (PVC) in Nigeria and diaspora to cast their votes in their respective places
of residence as long as the mobile application is accessible on their mobile devices
which will be available on various versions such as Android, iOS, Windows operating
system. Each voter’s details will be secure through the use of various cryptographic
techniques and verified with the use of one-time password during the voting process.
The paper is structured as follows: Sect. 2 discusses the related research to the
proposed study. Section 3 discusses the material and method used in this study
while Sect. 4 discusses the experimentation results and discussion on the findings
discovered from the study. Section 5 concludes the study, and future works were
suggested in the section as well.

2 Literature Review

Sontakke et al. [9] proposed the development of an e-voting platform which is an

online voting system that allows people to cast their ballots through their gadgets
or via the use of an e-voting website. The study used the one-time-password (OTP)
technique that is used more commonly on the web to tell the difference between a
human using a web server and an automated bot. If the results of the matching algo-
rithm match three points, it verifies the voter ID of this individual and then checked
with Aadhaar ID if he is entitled to vote and then be given a voting form which will
lead to the person to the third stage of the authentication using the one-time pass-
word (OTP) method. Kohno et al. [10] explained automated voting system security
functionality, and e-voting is more comfortable than physical voting. The author also
reveals that electors will cast total ballots without insider rights without any processes
being found inside the voting terminal program. Ciprian Stănică-Ezeanu [11] anal-
ysed the system of e-voting by explaining its benefits and drawbacks. His research
focused mainly on security mechanisms including firewalls or SSL communications,
Mobile Application Voting System: A Means to Achieve … 509

which are necessary but not sufficient to ensure the specific security of e-voting spec-
ifications. The author also explains the additional layer of quality improvement in
online voting process using pseudorandom one-time password (OTP) and special-
izing information systems to tackle the unique threats posed by automated voting
and maintaining essential security standards such as the safety of voters, election
fairness and voter identification. Equally, the author proposed using biometrics and
smartcard to authenticate people. One big concern discussed by the author is the
disparity between biometric and “classical” verification, such as smart cards. The
suggested e-voting program does not interfere in any manner with the real user’s
biometric features but also authenticates the user on the smart card with the aid of
the user’s certificate of authentication.
Rossler [12] proposed using remote Internet voting to improve the accessibility of
voters, raise the confidence of voters and increase turnout. Authors proposed remote
poll-site online voting as the next move forward in the survey because it offers greater
voter comfort but does not risk protection at the same time. Avi [13] discussed
the security precautions required for the interactive electronic voting method by
reflecting on two situations where electors cast their votes over the web–Arizona
Democratic Primary 2000 and Student Council Votes at the University of Virginia.
The author states that there are four critical criteria for a secure voting system:
security, affordability, secrecy and honesty. Shin-Yan et al. [14], this paper proposes
a new, oblivious signature scheme with a proxy signature feature to fulfil security
criteria, such as identity confidentiality, safety and message authentication, and to
ensure that no one can deceive other users (including the server). They suggest an
online voting system based on the existing oblivious and proxy signing scheme
and incorporate this scheme in a mobile framework to enable users to vote safely
and conveniently. Safety analyses and efficiency analyses are presented to show the
functionality and efficacy of the method proposed. Ullah et al. [15] proposed a hybrid
cryptosystem-mobile phone voting system (MPVS) with an electronic mobile voting
process. In the event of casting votes first from the cell phone and then pooling booth,
the device avoids double voting. The suggested method is more effective and accurate,
given that data would be transmitted via secure SMS to the election commission
server. Our machine required no Internet and no unique hardware computer that
reduced the cost. The device only needed a cell telephone and SIM card.
Patil et al. [16] successfully introduced our smartphone voting program in this
project and have used Firebase as the database. This program helps the user to build
and vote for the candidates without being on the polling booth physically present.
The application is encrypted and does not require the person to vote again or register
with the same telephone number or the same Aadhaar card number also. As Microsoft
Azure with cognitive software face API’ is not a free source, the face authentication
feature was not built into this program. But for this, a separate module has been
developed. The various reviewed work proposed different techniques [17, 18] of the
electronic voting process and government systems [19–21], but there are still some
lapses in the process. The first reviewed work does not put security concern into
consideration, Ullah et al. [15] lacked timely delivery of data to the service since it
was not Internet-enabled, and this makes cause delay of the voting process or casting
510 A. E. Adeniyi et al.

of votes by the voters. The last reviewed work uses a facial authentication feature;
this feature can consume more space and memory usage of the user device. This
proposed system is user-friendly and consumes less memory and space on the user
device. It also deals with the security concern by providing a one-time password
login authentication.

3 Material and Method

The research aims to develop an accessible and usable system allowing voters to
cast a vote instantly without the time and place limit. Mobile development tools are
used to design the mobile voting application. There are native mobile development
tools, as well as cross-platform mobile development tools. The cross-platform tool
is used in this research for the application to be able to run on Windows, Android
and iOS platforms. The proposed application will be easily installed on consumer
mobile phones, with minimal images or graphics for cheap mobile devices. Adobe
XD was used to model the system while the ionic framework was used because
it allows for quick prototyping and the pleasant environment when integrated with
angular. Mobile phones are portable devices; so far, they can be moved easily from
one place to another [22–27]. The mobile phone as proposed in this research will be
the best tool to achieve a seamless voting process in this part of the world and as well
eliminate any form of chaos during and aftermath of the election process in Nigeria.
This will increase our integrity as a country in electing who to lead the country for
a given period.
There are techniques available that facilitate the registration and voting process of
this proposed system to ensure seamless voting process, to secure the votes and details
of each voters from the third party and to tackle the issue of multiple voting by a voter.
The details of each voters will be secured using various cryptographic techniques and
save of the server. Digital information will be used to ensure the identity of voters
and that each voter votes only once. Each voters will be sent an OTP code for the
authentication of their digital information before the voting processing.

3.1 Security Techniques Used

The Rivest–Shamir–Adleman (RSA) cryptographic algorithm was used to secure

the voters details because of it large factorization of large prime numbers. The RSA
cryptographic is widely used for data and messages security in network and computer
system which involves two keys, public and private keys. The public key is known to
everyone and is used to encrypt data. The private is kept secret from knowing everyone
except owner, and it is used to decrypt data. The RSA cryptographic algorithm goes
thus:
Select two large prime numbers p and q
Mobile Application Voting System: A Means to Achieve … 511

Compute n such that n = p ∗ q

Compute phi(n) such that phi(n) = ( p − 1)(q − 1)
Choose a random integer e, e < phi(n) and gcd(e, phi(n)) = 1
that e ∗ d = 1mod phi(n)
Private key is (n, e)
Public key is (n, d)
Encryption is C = M e modn
Decryption is M = C d modn
OTP is often used as an extra element for multivariate verification and authenti-
cation applications. They are only relevant for a particular order for official approval
or identity verification. To prevent password lists, it is easy to send an OTP to the
voters via short message service. The voter’s mobile number must be registered with
a provider that offers SMS OTPs for verification or authorization. OTPs are very
common as an added permission or authentication element for mobile and web-based
services.

3.2 Hardware Setup for the Proposed System

The voters will have to secure the facility of multimedia phone with Android 4.0
Operating System compatible.
The user mobile SIM must be a registered SIM from the telecommunication
company.
The process will require the user mobile phone to be connected to the Internet of
at least the second generation of network.
The voter’s phone must be able to receive message from the third party in order
to receive the OTP authentication code for voting process.
Features of the proposed system include:
Availability: the proposed application is available for use as and when required.
Accuracy: the result computation is done automatically and accumulated as the
vote is being cast appropriately. This reduces errors encountered during manual
calculations.
Flexibility: the application can populate different political parties and candidates’
identity to conduct the election efficiently.
Mobility: the application is convenient and possesses high usability for portable
mobile devices.
Security: the proposed system provides anonymity of the votes being cast and as
well ensures data security from the third party (Fig. 1).
512 A. E. Adeniyi et al.

Fig. 1 Proposed system model

4 Experimentation Results and Discussion

The proposed system is a mobile-based voting system. This application operates on

two modules; the first module is modelled on registration, accreditation of voters, and
the voting process. In contrast, the second module is modelled on result collation,
result checker, and graph analysis of the result. Each registered voter is to download
the application on their mobile phones which come in a different versions (Android,
Windows, iOS) and register their devices with the Permanent Voters Card (PVC)
pin; after this has been done, they will be granted access to use the application to
vote for their preferred candidate. The voting process begins by each user login into
the application and online time password (OTP) will be sent to the registered voter’s
number to continue the process, after which it has been authenticated, the user can
cast his/her votes and wait for the allotted time before he/she can go back to check
the result and graph analysis of the result.
This section, therefore, describes the various performance of the proposed system.
It explains how each component works together to achieve a seamless mobile voting
process. The goal of the proposed approach is to simplify the voting process in
Nigeria whereby automate the counting and result generation process of the casted
Mobile Application Voting System: A Means to Achieve … 513

vote, thereby eliminating any form of doubt which is the central issue in the existing
system.
According to Fig. 2, if the user card number is valid, a feedback OTP message
interface will be displayed for the user to proceed with the voting process.
Figure 3 displays the registration page for the first-time user of the application.
After registration, the mobile voting application will be activated on the user’s mobile
phone.
When the user tries to log in, a token is activated, which sent a one-time password
to the user to authenticate and continue the process. As shown in Fig. 4, if the user
enters the correct OTP within the stipulated time, the application opens the users’
details page where the necessary details of the user can be confirmed and then proceed
to the voting page menu else the application will take the user back to the main login
page if the OTP is not correct or it has expired.

Fig. 2 Mobile voting application login page with feedback message

Fig. 3 Registration page with feedback message

514 A. E. Adeniyi et al.

Fig. 4 OTP interface with voter’s details page

Upon login and successfully verified the user’s essential detail, the voting page
will be loaded for the user to cast his/her vote, as shown in Fig. 5. The user clicks on
the vote button, and it will load a category page where the user can select which of
the position that is active to vote. Meanwhile, the check result button will be disabled
during the voting and be enabled immediately after the voting process has ended.
Figure 6 displays message feedback to the user after the vote has been cast.
This page will be activated after the end of the voting process, as shown in Fig. 7;
the user will have to login again before he can have access to the result of the casted
vote.

Fig. 5 Voting and category selection menu

Mobile Application Voting System: A Means to Achieve … 515

Fig. 6 Message feedback

Fig. 7 Result page and graph analysis

This page will be displayed when the users try to vote when the voting process
has been ended as shown in Fig. 8, while the second error message will be displayed
when the users try to view the result when the voting process is ongoing.

5 Technical Limitation of Mobile Voting Applications

The following are the most common concerns of mobile voting application:
516 A. E. Adeniyi et al.

Fig. 8 Error feedback message

Physical Environment factors: the physical factors can be major grouped into
internal and external factors of the firm.
Internal: this is the most pressing concerns due to the lack of expertise of the mobile
voting application usage and management of the digital voter information.
External: electoral bodies may not be involved in the mechanism of enrolment and
its consistency and credibility. Officials can gain access to the system and comprise
their ability to count votes correctly or to influence the election, driven by power or
wealth.
Technical factors: several computers have currently been built with differing speeds
and degrees of performance and durability that can cause disruption and crashes.
This often occurs in biometrics authentication. Modification in any of the biometrics
features like skin wrinkles can influence the devices from functioning properly. This
proposed system uses OTP and RSA cryptographic techniques in place of biometrics
fingerprint to minimize the technical limitation of mobile voting system and enhance
the performance of the system.

6 Conclusion

The main focus of this proposed system is to provide an easy means of conducting
a free and fair election in the country (Nigeria). This proposed system will help the
eligible citizen to cast their vote for the candidate of their choice at the convenient of
their house, thereby eliminating all form of physical stress (queue and favouritism)
that usually leads to commotion during the existing voting process. The system was
built in a way that promotes one person one vote’s law and also eliminates fraud
Mobile Application Voting System: A Means to Achieve … 517

voters in the sense that each user has a unique voter’s number and instant one-
time password to activate their mobile application during the voting process. Once
a vote is being cast, it will be automatically added up to the database server for
automatic summary and result generation. This feature makes the proposed system
to be seamless, saves time and voters life during the voting process. The proposed
approach is user-friendly, which makes it easy to use and efficient in its operation.
The authors recommend that the mobile voting system should be used in various
organizations, public and private sectors, even in the education system when trying
to choose their leader.

References

1. Bellis M (2007) The history of voting machines. Retrieved 9th November 2016, from http://
inventors.about.com/library/weekly/aa111300b.html Another reference
2. Okediran OO (2019) Mobile phones: a panacea for the implementation of E-voting in Nigeria.
Asian J Res Comput Sci 1–15
3. Falade A, Adebiyi AA, Ayo CK, Adebiyi M, Okesola O (2019) E-voting system: the pathway
to the free and fair election in Nigeria. Electron Government an Int J 15(4):439–452
4. Kulyk O, Volkamer M (2016) Efficiency comparison of various approaches in e-voting proto-
cols. In: International conference on financial cryptography and data security. Springer, Berlin,
Heidelberg, pp 209–223
5. Jumb V, Martin J, Figer P, Rebello A (2015) Mobile voting using finger print authentication.
Int J Eng Adv Technol (IJEAT) 4(4):141
6. Anderson C (2006) A timeline of electronic voting in the United States. Retrieved 28th
November 2016, from http://www.indypendent.org/?p=608
7. Das A (2015) Usability of the electronic voting system in India and innovatory approach. Int
J Appl Sci Eng Res 4(5):633–642
8. Heiberg S, Laud P, Willemson J (2011) The application of voting for Estonian parliamen-
tary elections of 2011. In: International conference on E-voting and identity, Springer, Berlin
Heidelberg, pp 208–223
9. Sontakke C, Payghan S, Raut S, Deshmukh S, Chande M, Manowar DJ (2017) Online voting
system via mobile. Int J Eng Sci Comput 7(5):12176–12178
10. Kohno T, Stubblefield A, Rubin A, Wallach DS (2004) Analysis of an electronic voting system.
In: Proceedings of IEEE symposium on security and privacy 2004. pp 1–23
11. CiprianStănică-Ezeanu (2008) E-voting security. Buletinul Universităţii Petrol–Gaze din
Ploieşti LX (2):93–97
12. Rossler TG (2011) E-voting: a survey and introduction. Available at http://wiki.agoraciudadana.
org/ images/ 5/5 6/An%2 BIntroduction%2Bto%2BElectronic%2BVoting%2BSchemes.pdf
Retrieved on 15th June 2012
13. Rubin A (2001) Security considerations for electronic remote voting over the internet. AT&T
Labs–Research Florham Park, NJ. Available at http://avirubin.com/e-voting.security.html. Date
Accessed 7th July 2012
14. Shin-Yan C, Tsung-Ju W, Jiun-Ming C (2017) Design and implementation of a mobile voting
system using a novel oblivious and proxy signature. Secur Commun Netw. https://doi.org/10.
1155/2017/3075210
15. Ullah M, Umar AI, Amin N, Nizamuddin (2016) An efficient and secure mobile phone voting
system. IEEE, pp 332–336. 978–1–4799–0615–4/13/$31.00 ©2013
16. Patil H, Barot H, Gawhale K, Mhaisgawali A, Chaudhari S (2019) Mobile based voting
application. Int J Res Appl Sci Eng Technol (IJRASET) 7(5):2181–2185
518 A. E. Adeniyi et al.

17. Abayomi-Zannu TP, Odun-Ayo I, Tatama BF, Misra S (2020) Implementing a mobile voting
system utilizing blockchain technology and two-factor authentication in Nigeria. In: Proceed-
ings of first international conference on computing, communications, and cyber-security (IC4S
2019). Springer, Singapore, pp 857–872
18. Anagha H, Chetana A, Jyothi B (2019). Mobile voting system. Int J Sci Eng Technol Res
(IJSETR) 6(4). ISSN: 2278–7798
19. Jonathan O, Ayo CK, Misra S (2014) A comparative study of e-Government successful imple-
mentation between Nigeria and Republic of Korea. In: Asia-Pacific World congress on computer
science and engineering, November. IEEE, pp 1–7
20. Edikan E, Misra S, Ahuja R, Sisa FP, Oluranti J (2019) Data acquisition for effective E-
Governance: Nigeria, a case study. In: International conference on recent developments in
science, engineering and technology, November. Springer, Singapore, pp 397–411
21. Okewu E, Misra S, Fernandez-Sanz L, Maskeliunas R, Damasevicius R (2018) An e-
environment system for socio-economic sustainability and national security. Problemy Ekoroz-
woju/Problems of Sustain Developm 13(1):121–132
22. Jonathan O, Ogbunude C, Misra S, Damaševičius R, Maskeliunas R, Ahuja R (2018) Design and
implementation of a mobile-based personal digital assistant (MPDA). In: International confer-
ence on computational intelligence, communications, and business analytics, July. Springer,
Singapore, pp 15–28
23. Adeniyi EA, Awotunde JB, Ogundokun RO, Kolawole PO, Abiodun MK, Adeniyi AA
(2020) Mobile health application and Covid-19: opportunities and challenges. J Critical Rev
7(15):3481–3488. https://doi.org/10.31838/Jcr.07.15.473
24. Sadiku PO, Ogundokun RO, Habib EAA, Akande A (2019) Design and implementation of an
android based tourist guide. Int J Modern Hospital Tourism 1(1):1–33
25. Emmanuel AA, Adedoyin AE, Mukaila O, Roseline OO (2020) Application of smartphone
qrcode scanner as a means of authenticating student identity card. Int J Eng Res Technol
13(1):48–53
26. Sowunmi OY, Misra S, Omoregbe N, Damasevicius R, Maskeliūnas R (2017) A semantic web-
based framework for information retrieval in E-learning systems. In: International conference
on recent developments in science, engineering and technology, October, Springer, Singapore,
pp 96–106
27. Adewumi A, Omoregbe N, Misra S (2016) Usability evaluation of mobile access to institutional
repository. Int J Pharmacy Technol 8(4):22892–22905
Adoption of Blockchain for Data Privacy
in 6G-Envisioned Augmented Reality:
Opportunities and Challenges

Umesh Bodkhe, Ashwin Verma, Deepti Saraswat, Pronaya Bhattacharya,

and Sudeep Tanwar

Abstract The widespread adoption of real-interactivity, fuelled with developments

in wireless communication infrastructures, such as sixth generation (6G), made
augmented reality (AR) a possibility in a range of vertical applications. AR inte-
grates the digital visuals into a virtual space that emulates the real-world feedback
for the user. It supports a range of haptic and interactive communication and thus
has attracted the interest of the research community to provide useful AR frame-
works. However, as the user-private data is communicated through open channels,
it raises privacy and security concerns. In parallel, decentralized computing has
shifted resource provisioning to near device interactions that support the real-time
connectivity of AR. To preserve privacy and induce trust among AR stakeholders,
blockchain (BC)-based AR use cases are presented by researchers. However, the
existing surveys have not focused on reference architectures and implementation
challenges. Motivated by the research gap, the paper presents a systematic survey of
the integration of BC to induce data privacy in AR ecosystems. A comparative anal-
ysis is presented, and the potential application of BC-based AR is discussed. A case
study on VoRtex enterprise is presented, and a supporting architectural framework is
discussed. The survey intends to present the readers deep insights about BC adoption
in AR that would motivate industry practitioners, academicians and researchers to
propose scalable BC-integrated solutions in AR/VR-based ecosystems.

U. Bodkhe (B) · A. Verma · D. Saraswat · P. Bhattacharya · S. Tanwar

Department of Computer Science and Engineering, Institute of Technology, Nirma University,
Gujarat, India
e-mail: umesh.bodkhe@nirmauni.ac.in
A. Verma
e-mail: ashwin.verma@nirmauni.ac.in
D. Saraswat
e-mail: deepti.saraswat@nirmauni.ac.in
P. Bhattacharya
e-mail: pronoya.bhattacharya@nirmauni.ac.in
S. Tanwar
e-mail: sudeep.tanwar@nirmauni.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 519
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_38
520 U. Bodkhe et al.

Keywords Augmented reality · Blockchain · 6G · Data privacy · Security

1 Introduction

The advancements in holographic interactions in the beyond fifth-generation (5G)

and emerging sixth-generation (6G) infrastructures have driven industry projects
towards the design of scalable and interactive augmented reality (AR) designs. AR
expands the physical world objects and interactive elements into a virtual space and
can be processed through information layers, such as audio, visual graphics and
cognitive sensing elements. Developing complete artificial world by replacing real
things with virtual things is not possible in AR, while virtual reality (VR) has the
capability to do the same. AR can be visualized by incorporating voices, pictures
and graphics in the real world. By superimposing the computer-generated images,
actual real-world environment can be viewed differently through AR by changing the
reality [1]. In the same environment, AR allows users to experience the coexistence
of real and virtual entities.
Several corporate companies amount to huge investments in the AR market.
Development and launch of new hardware and software accessories, involvement
in a variety of applications and increased funding in market sectors are fuelling AR
growth. AR provides a variety of applications starting from entertainment, health
care, advertising, real-estate, virtual experiences market, art, monetization of games,
tourism, e-commerce to education and training, and many more. AR devices collect
data from various sources but experience security and privacy issues due to the scale,
scope and sensitivity of the latter. The information collected by AR devices can
be classified as Observed (location data, motion tracking and user-provided infor-
mation/generated data), Observable (like virtual person/object, avatar and real-time
world/in-application), Computed (recommendation/advertising and biometric infor-
mation) and Associated (login/contacts/payment information, virtual assets and IP
address). These data in one or another form may be attacked by adversaries. There are
various possible attacks on AR data that are private for a person or an organization.
The breaches mainly include the following attack scenarios.
• A malicious attacker can utilize the identity of another person for whom the
desired immersive experience is not created, enabling fraud and theft.
• Revealing biographical, health information, web-browsing, way of living, age/
gender/race, etc., can cause significant harm if disclosed.
• Misuse and outflow of derived sensitive information may lead to potential damage
of identity of the intended user.
• Linking of user data (IP address, username/password with other descriptive infor-
mation results in access to private/public and social accounts. This may lead to
economic and reputational harms.
• Data, location services, sound interaction, surveillance information collected from
AR device, which includes user information can be used for various purposes, viz.
Adoption of Blockchain for Data Privacy … 521

selling, marketing, overlaying, spying and third-party interaction which imposes

privacy threats.
There are several risks associated with developing immersive technologies or
content like consumer privacy, data security, infringement, copyright and malicious
attack restoration. Recently, there are developments of distrust, scepticism among
consumers with corporate firms due to misuse of data. A privacy agreement between
users and the company is required to address the lack of transparency and account-
ability. Moreover, the data demands security to avoid fraudulence and end-to-end
delivery, processing, exchange and delivery to the consumer-oriented application to
ascertain particular behaviour. These challenges can be catered through BC which
acts as a ledger and stores all the transactions in a decentralized way.

1.1 Benefits of Adopting BC in AR

Figure 1 depicts the benefits of BC adoption in various AR applications as follows.

1. Requirement of fast image/video display and processing enables AR assets to
be stored locally. Moreover, AR devices exchange data to and from the local
server because of memory and processing limitation.
2. AR-oriented applications are data and bandwidth-hungry. Overloading of the
central server may occur in the case of multiple AR devices asynchronous
communication. BC can provide data decentralization in such a situation.

Fig.1 Potential applications of BC in AR

522 U. Bodkhe et al.

3. BC can help in the commercialization of AR devices through the creation of a

user-defined marketplace for storing and uploading AR content in a decentral-
ized fashion. Moreover, the data can be stored and peered using an interplanetary
file system (IPFS) that allows authorized stakeholders to access records with
IPFS key, and hashed IPFS references are stored as external references in the
BC ledger. This improves the scalability of BC, as more transactions can be
fitted in each block due to the small transaction size.
4. BC-based tokenization enhances financial transactions where AR devices act
as hardware and perform peer-to-peer transactions on digital tokens.

1.2 Motivation

The motivations for the paper are as follows.

– The shift towards beyond 5G and emerging 6G communications has transformed
networks to support terahertz (THz) bands, sub-millimetre wave communications
and photonic communications, at extremely low latency. Moreover, 6G allows
services to be provided as network-in-a-box (NIB) requirements that simplify
network and management operations [2]. Due to this, a range of holographic
technologies is designed that would break the distance barriers and support real-
time interactions to support three-dimensional (3D) imagery, automated cars, AR,
digital-twin technology, scalable industry 4.0, virtual reality (VR) and massive
Internet of Things (IoT) communications. In the design of massive and scalable
AR frameworks, however, the simplicity and flexibility of network configurations,
and design of network virtual functions to drive 6G devices, have increased the
risk of data leakage, and thus, user data privacy, authorization and security are of
prime importance. The need is intensified in decentralized edge computations, and
thus, BC leverages trust, chronology and auditability in stored data as transactional
ledgers [3]. Recently, with rising privacy regulations and data sharing concerns,
research studies have shifted towards security in AR applications [4]. However, the
integration of BC in AR communications is not explored in detail by researchers
globally. Thus, the survey presents a systematic discussion of data privacy and
security issues in BC-envisioned and responsive AR applications.

1.3 Key Takeaways

The major key takeaways of the survey are as follows.

– The novelty of this survey is to emphasize on privacy and security concerns
achieved by blending BC with AR which help researchers to include these
technologies in their relevant domains.
Adoption of Blockchain for Data Privacy … 523

– We present a comparative analysis of existing surveys and based on identified

research gaps have proposed key discussions on emerging BC-based AR/VR
projects, and the open issues and challenges of the integration.
– A real-world case study VoRtex presents that amalgamates BC in scalable AR/VR
space to present an interactive education platform. The user-centric design and
access restrictions are presented as part of the case study.
This paper is organized as follows. Section 2 presents the related work. Section 3
illustrates the key discussions on the data privacy issues in single and networked
AR applications. Section 4 presents the integration of BC in AR verticals. Section 5
presents some real-world BC-assisted AR/VR deployments. Section 6 presents the
open issues and challenges in deploying BC in AR applications. Section 7 presents
the case study of VoRtex Enterprise Platform, a BC-based AR interactive platform.
Finally, Sect. 8 presents the concluding remarks and future scope of the work.

2 Related Work

Till date, surveys have majorly focused on secured techniques and approaches to
envision data privacy in AR applications. The authors have analysed some param-
eters such as privacy, integrity, security and privacy-preserving algorithms. Table 1
presents a comparative analysis of state-of-the-art approaches in AR/VR.
Nguyen et al. [5] discussed the importance of BC-based projects in AR domain
through the use of digital ledger. Authors also highlighted the potential Table 1:
Comparative analysis of existing state-of-the-art surveys in AR/VR with the proposed
survey.
benefits of BC which can improve a security in AR. Carames et al. [8] presented
a BC-based holistic approach in AR. Mewes et al. [13] presented projector-based
AR visualization concepts, but not considered any of the security issues. Abbas et al.
[9] performed in-depth survey on variety of BC applications including AR. They
presented BC as the most efficient security scheme to resolve the privacy issues in
intelligent mobile augmented reality (IMAR) project. Cannavo et al. [10] outlined a
comprehensive survey on the integration of VR, AR and BC. They discussed different
use cases for AR/VR with virtual experience marketplace and monetization of games.
BC is the key solution for the betterment of user’s interaction with various digital
content by using gestures, gaze and various natural interfaces. Authors in [14] that,
similar service is currently offered by the company named ‘CEEK’ [14], which offers
its user to participate in live music events with AR/VR by enhancing the entertainment
experience. According to [15], the opportunities, use cases and challenges of AR are
frequently studied and illustrated based on the review of 5G developments the data
movement and speed of the communication still exist challenges in the AR/VR in
terms of security too. Hence, researchers integrated 6G with AR/VR for a better
response as well an edge of artificial intelligence.
524 U. Bodkhe et al.

Table 1 Comparative analysis of state-of-the-art approaches in AR/VR

Authors Year Objective Pros Cons
Nguyen et al. [5] 2018 BC-based projects Use of digital ledger Experimental
were identified and analysis was not
optimized the found
security of the
project
French et al. [6] 2018 Integration of BC, Discussed several Attack scenarios
5G and AR models to increase about user
BC-assisted authorization, and
manufacturing data confidentiality is
capabilities, with 5G not discussed
network orchestration
and management
support
Werner et al. [7] 2018 Detailed study on Location-based Data tampering
AR and its system with digital
implementation in maps and images
analogue and digital
maps and images
Carames et al. [8] 2019 Focused on the use BC can benefit Interoperability
of BC in the automotive industry issues of BC
cyber-resilient and ensure data adaption to different
automotive industry security, anonymity AR and IoT verticals
and accountability are not discussed
Abbas et al. [9] 2019 An overview of BC Efficient security Privacy of the mobile
to provide security to scheme was proposed data in augmented
objects of intelligent reality was not
mobile AR and BC discussed
Cannavo et al. 10] 2020 Author has drawn Integrate AR and BC Privacy during data
attention from the within exchange was
traditional use of BC unified platforms maintained between
to AR and VR the users on different
platform
Lim et al . [11] 2021 Identified BC as a Proposed use of BC Un-secure data
key solution on the to secure the AR/VR exchange between
vulnerability of the system the users on different
VR/AR software platforms
platform
Liao et al. [12] 2021 proposed novel Quality of Service Deployment was
consensus (QoS) possible in 5G and
mechanism as yet to finalize on 6G
Proof-of-Cache- nodes
Offloading (PoCO)
(continued)
Adoption of Blockchain for Data Privacy … 525

Table 1 (continued)
Authors Year Objective Pros Cons
Proposed 2021 Integration of BC in Security and privacy -
AR/VR space, along issues of emerging
with possible AR applications,
industry along with possible
deployments, and benefits of mitigation
potential challenges of attack vectors
are discussed through BC are
discussed. A suitable
case study is
discussed

3 Data Privacy Issues in AR

In today’s new era, AR is becoming more and more popular. Though there are
various security and privacy risks present in AR systems, for example, eye tracking
systems that mainly employ biometric identification, AR can enhance the accuracy
for authentication and improves the user experience, but the same would require
intricate and explicit details of user identification to be stored in released tables, and
that might hinder the privacy of the customer. An adversary can spoof the identity
by using fake identities as well data privacy can be spoiled in AR systems.

3.1 Issues in Multiple Applications

Multiple applications in AR, this category is having more than one single application
in a single AR system.
– Data Access: There is a lot of shared data available in multiple application environ-
ments. Access for this cross-application sharing data needs to be given properly
else it can lead to the use of one application data accessed by some malicious
application in the network [16].
– Input: Usually users interact with the AR system using voice or embedded gloves,
etc. In multiple application scenarios when a user tries to interact with one appli-
cation, might be brought to other applications’ focus. And malicious applications
might steal the input of another application and by registering that input, later be
used to access that application.
– Output: In multiple applications of AR, applications have more than one output
device. Conflict among them might be created when more than one application
attempts to use that output device, it could lead to security concerns. For instance,
a malicious application tries to access content presented by another application.
526 U. Bodkhe et al.

3.2 Issues in Single Application

– Data Access: AR applications are having various data such as sensor data, GPS
data, video and audio feeds, temperature, and accelerometers reading. Whom to
provide how much access to these data is difficult in AR single application, which
might not be able to secure data privacy.
– Input: Input in single AR applications also faces similar input validations issues
as a conventional application such as integrity is not maintained.
– Output: In a single application, a malicious application might use some techniques
which can sensory overload the users. Output attacks in a single application are
more serious in the AR domain, as it is difficult to distinguish feedback of users
as virtual or real and due to that users might not be able to shut down the system
which leads to loss of data in running mode. BC-based solution for data privacy
in AR.

4 Integration of BC with AR

In this section, we presented the role of BC and its integration with AR to preserve
the privacy of user data. BC technology guarantees security and authenticates trans-
actions and records by cryptography [17–19]. Theoretically, to boost the efficiency of
AR platforms, decentralized systems such as BC can provide instant and unchanging
data transfer [20]. The network of BC is used to store data retrieved from client devices
as well as data providers for protection of integrity at which hashed data is posted
[21, 22]. Data accessibility from the provider, each request for data access should
be handled using a decentralized permission management procedure to get approval
from the data holder. A decentralized protocol for the control of permissions is a key
part of a decentralized network [23]. BC technology is an emerging technology. Due
to this, companies have started to integrate BC technology and AR [24, 25]. There is
a centralized system to store or retrieve information from a huge volume of data. But
BC technology is having distributed ledger technology which ensures data security
at a higher level.
BC-based AR applications are usually having two chains—one is for basic data
of the application and another is for data used by transactions. Some AR applica-
tions need to be more secure than other ones such as AR shopping applications.
Application built on BC technology is providing high security as it is immutable and
transparent. Therefore with the changing time, it needs to be adapt faster-growing
technologies together to provide a way for securing important data over the Internet.
BC is advantageous in the AR field in many ways. For instance, deciding whether
a digital photo is an original photograph or a copy of a digital photograph is very
difficult. But BC can handle this issue easily by allowing the developers to create
unique digital assets that cannot be copied by other developers or users [26–28].
Thus, the integration of BC technology in AR, for sure gives trust to customers for
relying on such real-time applications.
Adoption of Blockchain for Data Privacy … 527

5 Real-Life Project Deployments of BC-Assisted AR/VR

Space

This section describes BC-integrated AR/VR projects which are excellent instances
of BC technologies being used beyond the financial sector. Table 2 is the feature table
based on the software specifications and standard scenarios for the development of
BC-virtual reality platform. Here label is used to represent essential feature.
– CEEK: CEEK virtual reality is a producer of immersive VR content which
provides multiple VR channels with social and 3D experience. Virtual reality
experiences involve 360 music videos, charities and sports events. The CEEK
service platform is compatible with all devices including smartphones, desktops
and smart TVs.
– Decentraland: Decentraland (MANA) describes itself as an Ethereum BC-virtual
reality platform empowering users to develop, access, monetize content and appli-
cations. Users in this virtual world buy plots of land which they can explore, build
on and monetize later [29].
– Vibehub: VibeHub is a ground-breaking platform specializing in new AR tech-
nologies are designed to revolutionize how people communicate and monetize
content on the internet. Build a cutting-edge network that uses both AR and VR,
allowing users to attend a variety of events being conducted such as live concerts,
develop new skills, and in virtual reality connect with friends and family.
– MARKSPACE: MARK.SPACE is a platform that is open source for the develop-
ment of 3D / AR / VR-compatible spaces and entities of any type, as well as for
their rapid launch and incorporation into a specific environment. This platform is
BC-based and offers a crypto-economy.

Table 2 Comparison of BC-virtual reality projects

Criteria/features VoRtex CEEK Decentraland Vibehub Virtual universe Mark space
3D computer • • • • • •
graphics
360 degree ° • ° ° ° °
videos
VW and source • ° • • • •
or SDK tools
Open source or • • • ° ° °
SDK tools
AI and • ° ° • ° °
story-telling
Crypto-currency • • • • • •
support
• Parameter is considered, °Parameter is not considered
528 U. Bodkhe et al.

6 Open Issues of Integration of 6G and BC in AR

and Challenges During Deployments

This section gives glimpses for meaningful insights to the research community and
readers about the potential of decentralized BC in the AR domain to overcome secu-
rity and privacy-related issues. BC has also some limitations in terms of scalability,
network measurements and privacy attack vulnerability in the smart contract. In this
section, we presented open issues of BC in AR and challenges during the implemen-
tation phase in detail. Figure 2 presents the open issues and challenges of mainstream
and widespread adoption of AR/VR in industry projects.
In the figure, we have presented the challenges and have proposed the possible
key directions in blue-boxes. Inclusion of 5G and 6G services, like ultrareliable low-
latency services (uRLLC), 6G-based enhanced reliable low-latency services, 5G-
enhanced mobile broadband (eMBB) and 6G-further enhanced mobile broadband
(FeMBB). 5G-supports a peak data rate of 20 Gbps, user-experienced data rate of
0.1 Gbps and connection density of 106 devices/per sq. km. The round-trip time
latency of 5G tactile and haptic communications is < 1 ms. Thus, 5G can address
some of the key challenges of AR deployments. 6G services, on the other hand,
offers a user-experienced data rate of 1 Tbps, and a user-experienced data rate of 1
Gbps. The connection density is 10 × times of 5G and is 107 devices/per sq. km.
The round-trip latency over the air (OTA) interface in 6G varies from 10 − 100
microseconds (µ s). Thus, 6G supports a range of AR verticals and addresses the
quality-of-interaction (QoI), further than 5G. For security and privacy considerations,
users can design databases to follow K-anonymity and differential, privacy-based
models. Similarly, BC-based issues can be addressed through decentralized storage
via interplanetary file systems (IPFS), permissioned chain structures that improve
the mining latency and improve bandwidth. For smart contract verification, we can

Fig. 2 Open issues and possible research directions of integration of 6G and BC in AR/VR
deployments
Adoption of Blockchain for Data Privacy … 529

predict the possible attack vectors to contracts like gas flaws, injection dependence,
transaction re-entrancy and many more.
– Peer-to-peer bandwidth: The first is downloading speed since it has historically
been too slow to get a file from distributed peer-to-peer (P2P) storage network.
– Avatar Designs: The other challenge for P2P designs, however, is to include
updated information for each avatar (profile) on which other avatars (profiles) are
its neighbours.
– Lack of uniformity in standards: Space with a central location and transparent
boundaries where users can move and communicate with objects and control
materials, with interactive features such as role play and participation in teamwork.
– Variable application service provisioning: Enables users to interact in a network
or an environment using variable rates of audio depending on the average distance
among users.
– Scalability and high latency: Sometimes the use of BC for this network will
enhance the scalability because of its decentralized nature and it provides less
throughput with high latency.
– High requirement of Storage and Computation power: BC-based consensus algo-
rithms require the high computation power and high storage facility to process
and store the data.
– Vulnerability of smart contracts: Smart contracts are susceptible to various attacks
such as re-entrancy, timestamp dependence, short address and delegate call.

7 Case Study: VoRtex Enterprise: BC-Assisted

Decentralized VR Platform

VoRtex is an educational platform that is developed using BC technology integrated

with VR. This platform enterprise artificial industrial environment which is similar
to reality to employ training from various fields like health care, automotive and
space. To provide trust and security about the privacy of data in the virtual world
(VW). Jovanovic et al. [20] proposed a solution platform with combination of BC
technology. This platform is having a user-centric design and allows the processing
of data using a BC distributed environment, by which data ownership and integrity
functionality can be provided. The decentralized BC-based model works as a back-
end, where data is represented in the form of blocks and is synchronized among
connected users. Figure 3 represents the high-level software architecture of VoRtex
platform. The proposed solution architecture is listed as follows.

8 Concluding Remarks and Future Scope

Technical advancements in wireless communication infrastructures and the emer-

gence of connected things made the realization of real-time visual communications
a reality. With the emergence of beyond 5G and 6G-driven applications, AR/VR
530 U. Bodkhe et al.

Fig. 3 Component-based software architecture of VoRtex

has transitioned data-driven applications with virtual and haptic connectivity, with
massive device-connectivity. AR/VR space has become a complex ecosystem, with
high ingested data, which has made security and privacy issues more paramount.
Thus, the adoption of BC in AR/VR-based ecosystems handles the inherent issues of
trust, security and privacy among AR stakeholders. The survey systematically high-
lights the potential BC-envisioned AR use cases and presents the key drivers and
benefits. As part of the future scope, the authors intend to investigate the highlighted
open issues, and possible frameworks that would help practitioners in the design
of scalable decentralized AR solutions, with a view of addressing the trade-offs of
required security and privacy counterparts.

References

1. Forescout (2020) What is augmented reality (AR) and how does it work. https://
thinkmobiles.com/blog/what-is-augmented-reality/ (2021), [Online; Accessed 03 –December
2020]
2. Hsu CH, Manogaran G, Srivastava G, Chilamkurti N (2021) 6g-enabled network in box (nib)
for industrial applications and services. IEEE Trans Indust Inform 1–1. https://doi.org/10.1109/
TII.2021.306770
3. Bodkhe U, Tanwar S, Bhattacharya P, Verma A (2021) Blockchain adoption for trusted medical
records in healthcare 4.0 applications: a survey. In: Singh PK, Wierzchoń ST, Tanwar S,
Ganzha M, Rodrigues JJPC (eds) Proceedings of second international conference on computing,
communications, and cybersecurity. Springer Singapore, Singapore, pp 759–774
4. Yamakami T (2020) A privacy threat model in xr applications. In: Barolli L, Okada Y, Amato
F (eds) Advances in internet, data and web technologies. Springer International Publishing,
Cham, pp 384–394
Adoption of Blockchain for Data Privacy … 531

5. Nguyen QK, Dang QV (2018) Blockchain technology-opportunities for emerging economies.

In: 2018 4th international conference on green technology and sustainable development
(GTSD). IEEE, pp 478–482
6. French A, Risius M, Shim JP, Bordnick P, Parsons J, Holtman D (2018) Virtual reality and
blockchain technology in the era of 5g: Disrupting business and society
7. Werner PA (2019) Review of implementation of augmented reality into the georeferenced
analogue and digital maps and images. Information 10(1):12
8. Fraga-Lamas P, Fernández-Caramés TM (2019) A review on blockchain technologies for an
advanced and cyber-resilient automotive industry. IEEE Access 7:17578–17598
9. Abbas QE, Sung-Bong J (2019) A survey of blockchain and its applications. In: 2019 Interna-
tional conference on artificial intelligence in information and communication (ICAIIC). IEEE,
pp 001–003
10. Cannavo A, Lamberti F (2020) How blockchain, virtual reality and augmented reality are
converging, and why. IEEE Consumer Electron Magazine
11. How can blockchain accelerate the virtual reality revolution? https://www.entrepreneur.com/
article/365482, note = Accessed: 2021–02–03
12. Liao S, Wu J, Li J, Konstantin K (2020) Information-centric massive IOT based ubiquitous
connected vr/ar in 6g: A proposed caching consensus approach. IEEE Internet of Things J
13. Mewes A, Heinrich F, Hensen B, Wacker F, Lawonn K, Hansen C (2018) Concepts
for augmented reality visualisation to support needle guidance inside the mri. Healthcare
Technology Letters 5(5):172–176. https://doi.org/10.1049/htl.2018.5076
14. Ceek AV (2020) Ceek:.http://www.ceek.com/ (May 2020)
15. Tang F, Kawamoto Y, Kato N, Liu J (2019) Future intelligent and secure vehicular network
toward 6g: Machine-learning approaches. Proc IEEE 108(2):292–307
16. Ladha A, Bhattacharya P, Chaubey N, Bodkhe U (2020) Iigpts: Iot-based framework for intel-
ligent green public transportation system. In: Singh PK, Pawłowski W, Kumar N, Tanwar
S, Rodrigues JJPC, Obaidat MSO (eds) Proceedings of first international conference on
computing, communications, and cyber-security (IC4S 2019), vol 121. Springer International
Publishing, pp 183–195
17. Bodkhe U, Tanwar S (2020) A taxonomy of secure data dissemination techniques for IOT
environment. IET Software 1–12. https://doi.org/10.1049/ietsen.2020.0006
18. Forescout (2021) ”privacy and blockchain. https://en.wikipedia.org/wiki/ Privacyand-
blockchain. [Online; Accessed 13 December 2020]
19. Tanwar S, Bhatia Q, Patel P, Kumari A, Singh PK, Hong WC (2020) Machine learning adoption
in blockchain-based smart applications: the challenges, and a way forward. IEEE Access 8:474–
488. https://doi.org/10.1109/ACCESS.2019.2961372
20. Jovanović A, Milosavljević A (2019) Decentralized virtual reality blockchain-based platform.
Facta Universitatis Series: Autom Control Robot 18(1):057–077. https://doi.org/10.22190/FUA
CR1901057J
21. Bhattacharya P, Tanwar S, Bodke U, Tyagi S, Kumar N (2019) Bindaas: blockchain-based deep-
learning as-a-service in healthcare 4.0 applications. IEEE Trans Netw Sci Eng 1–1. https://doi.
org/10.1109/TNSE.2019.2961932
22. Bodkhe U, Bhattacharya P, Tanwar S, Tyagi S, Kumar N, Obaidat MS (2019) Blohost:
blockchain enabled smart tourism and hospitality management. In: 2019 international confer-
ence on computer, information and telecommunication systems (CITS), Shanghai, China.
IEEE, pp 1–5
23. Bodkhe U, Tanwar S (2021) Secure data dissemination techniques for IOT applications:
Research challenges and opportunities. Software: Practice and Exper n/a(n/a) 1–23. https://
doi.org/10.1002/spe.2811, https://onlinelibrary.wiley.com/doi/abs/https://doi.org/10.1002/spe.
2811
24. Bodkhe U, Tanwar S, Bhattacharya P, Kumar N (2020) Blockchain for precision irrigation:
opportunities and challenges. Trans Emerg Telecommun Technol e4059. https://doi.org/10.
1002/ett.4059, https: //onlinelibrary.wiley.com/doi/abs/https://doi.org/10.1002/ett.4059
532 U. Bodkhe et al.

25. Dhalor M (2021) Blockchain in augmented reality. https://www.ommzi.com/blockchain-aug

mented-reality/ [Online; Accessed 13- December 2020
26. Bodkhe U, Tanwar S, Parekh K, Khanpara P, Tyagi S, Kumar N, Alazab M (2020) Blockchain
for industry 4.0: a comprehensive review. IEEE Access 8:79764–79800
27. Mehta D, Tanwar S, Bodkhe U, Shukla A, Kumar N (2021) Blockchainbased royalty contract
transactions scheme for industry 4.0 supply-chain management. Inform Process Managem
58:1–20. https://doi.org/10.1016/j.ipm.2021.102586
28. Price (2021) Next reality ”the future of augmented reality and blockchain tech-
nology. https://medium.com/@NextRealityNews/the-future-of-augmented-reality-blockchai
ntechnology-9a24b8a0a1ec. Online; Accessed 13 Dec 2020
29. Coinmarketcap (2021) Decentraland (mana. https://coinmarketcap.com/currencies/decentral
and/. [Online; Accessed 13 December 2020]
Integration Blockchain for Data Sharing
and Collaboration in Mobile Healthcare
Applications

Inaam Zia, Pawan Singh, Anil Kr. Tiwari, and Amit Pandey

Abstract Enabled by way of cellular and wearable era, private fitness record
provides huge and growing importance of healthcare, which is beneficial for both
care vendors and medical studies. The secure method and handy sharing of personal
health statistics is vital to the improvement of the interplay and collaborating with the
healthcare enterprise. Confronted with the capability privacy troubles and vulnera-
bilities present in current private health statistics garage and system of sharing, which
also include the concept of self-sovereign facts ownership, we advocate an modern
person-centric fitness data sharing solution via making use of a decentralized and
permissioned blockchain to guard privateness the usage of channel formation scheme
and beautify the identification management the use of the club service supported with
the aid of the blockchain. A cellular utility is deployed to accumulate fitness data
from non-public wearable devices, guide input, and scientific gadgets, and synchro-
nize information to the cloud for records sharing with healthcare insurance agencies
and healthcare companies. To maintain the integrity of fitness statistics, inside each
document, an evidence of validation and integrity is permanently retrieved from
database stored in cloud location and is very important to the blockchain network.
Furthermore, for performance and scalable and considerations, we undertake a tree-
based totally processing statically and batching method to handle big facts units of
personal fitness statistics captured and uploaded by means of a mobile platform.

Keywords Health care · e-health · Privateness · Permissioned blockchain · Get

right of entry to manage · Scalability · Integrity · Wearable devices · Cell platform

I. Zia (B) · P. Singh · A. Kr. Tiwari

Amity University Uttar Pradesh, Lucknow Campus, India
A. Pandey
College of Informatics, Bule Hora University, Bule Hora, Ethiopia

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 533
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_39
534 I. Zia et al.

1 Introduction

The healthcare system is one of the most demanding and evolving sectors in the world,
any countries growth is directly proportional to the how good the health system is,
and it constitutes to certain amount of GDP of a country [1]. The industry may
include companies which mainly consist of generalization and commercialization
of the products sometimes without taking into consideration of the factor of healing
and improving of the sector. It is considered as one of the most complex system in
the world which contain counterfeiting of drugs without being able to discriminate
the bad ones from the good one, not being able to break the heavy barriers such
as distributed patient record and try to fetch each and every record of the patient
from different hospitals is a very difficult task, achieving a transparency during data
sharing is a very difficult task too and if there is not proper sharing of data among
different parties it may lead to disastrous impact on the health of a patient. There
has been cases where it has been seen that the record has been sold for money by
the middle man so we have to produce something which may lead to a system where
there is no middle man and there is security and transparency of data which may lead
to trustworthy system.
There are many research papers which has discussed about the use blockchain in
health care [2–6].

2 Blockchain

Bitcoin is one of the famous cryptocurrencies, and the technology on which it is

based on is known as blockchain [7] because of the release and use of bitcoin the
blockchain became very famous, people started talking about it and taking interest in
it. Bitcoin works same as other currency like Dollar and Pounds, but it uses different
concept to maintain and handle these cryptocurrencies. Due to its volatile nature and
anonymous founder of various cryptocurrencies, the government of various country
is banning or regulating the laws to monitors these cryptocurrencies, one of the
features of these cryptocurrencies is that it is very hard to track transactions of these
cryptocurrencies, and there are hardly any experts present with the government of
the countries to do so which makes it the most difficult for government to legalize it.
But there are various positives sides of blockchain in which are interested which are
decentralization of data which is maintaining a copy of record at every node making
the data secure from losing it and there cannot be changes done in the previous node
as there is concept of proof-of-work. So, it can be concluded that with the help of
blockchain we can develop tools which can be helpful in various aspects of life like
in the medical field or in the shipment industries. As we know any cryptocurrencies
can be mined only it matches the hash function [8] of it.
Integration Blockchain for Data Sharing and Collaboration … 535

2.1 Consensus

The consensus refers to a network of blockchain on which mechanism should be used

to build a particular network, there are various factors which are taken into consid-
eration before going for the model which determines which transaction should be
accepted or rejected based on the model selected for the network. Merely a regula-
tion of this protocol deals with the regulation of transaction whether it is valid or
not, adding the newly created node in the network or rejection of node based on the
mechanism. Basically, there are categorized into following:
1. In the case of global consensus model, there is a main block which is basically
called as genesis block, and every node which is similar to the genesis block
will be stored after validation. Most popular example is Bitcoin and Ethereum.
2. In the early of consensus, it only deals with the concerned nodes, and there is
ignorance of other nodes. For example, blockchains are nano [9] and trust chain
[10].
One of the most common and popular mechanisms is known as Nakamoto
consensus. The Nakamoto consensus deals with the concepts of proof-of-work
(POW) for accepting or rejection of new node in the block. If a miner is mining
complete a task and find a hash of the new node it will be added to the network as it
is validated by the miner this concept is known as proof-of-work. the stake [11].

2.2 Smart Contracts

A smart contract is a kind of contract which does not require any manual adjustment,
everything is done automatically, it is verified automatically, it is executed automati-
cally, and there cannot be any changes done in the code which is programmed for the
logic which later on runs on the blockchain. A custom logic or rules or guidelines
can be written on the blockchain; on the satisfaction of these rules, some actions are
triggered Mostly all the cryptocurrency are adapting these smart contracts.

2.3 Blockchain Types

Blockchain can be classified into three categories which are as follows:

• Public blockchain as the name suggests it is open to all and anybody can take
part into it. It is basically based on the trust of decentralization one of the most
common example it.
• Private blockchain is personal to an organization or group; it is made for public
use and only used by the persons in the organization, and a personal blockchain
is a permissioned blockchain centralized to 1 governing organization.
536 I. Zia et al.

2.4 Challenges and Future

In spite of having so much capacity, there are various obstacle in the blockchain.
The concept of blockchain is that it maintains a copy of transaction at every
node which makes the makes the transaction very sluggish; it is fine for very a
small transaction but not in the case of very large transaction. Due to the previous
point, it is hard to enhance the scalability of blockchain in the software. In case of
bitcoin and Ethereum also, there it is very hard to enhance the scalability though it
provides decentralization and consistency. In the future, we should be working on
improving the feature of scalability. As the scalability of the software is increased,
it opens a door for various other features which may increase the efficiency of the
software.

3 Device Layout

3.1 System Evaluation

Non-public health information sharing. 6 entities which are included, particu-

larly consumer, wearable gadgets, healthcare company, insurance employer, the
blockchain network and cloud database (Fig. 1).

Fig. 1 Consumer-centric personal fitness facts sharing

Integration Blockchain for Data Sharing and Collaboration … 537

3.2 Machine Entities

• User. System customers gather record from wearable gadgets which screen users’
health facts which include strolling distance, napping situations, and heartbeat.
The ones records are uploaded to the database on cloud hosted on trustworthy
platform through a mobile application. Consumer is the sole owner of private
fitness statistics and is liable for allowing, revoking, and denying facts get right
of entry to from another parties, which are healthcare vendors and insurance
corporations. If the person is in the search of scientific treatment, the user would
percentage the health facts with the desired doctors. If the treatment is completed,
the information gets right of entry to is revoked to disclaim similarly access
from the doctors. Identical state of affairs applies to user-insurance organization
relations. Except, user also can report normal activities in line with a specific
scientific remedy along with medication utilization to proportion with the remedy
issuer for adjustment and higher improvement.
• Wearable devices. Wearable devices serve to rework authentic health data into
user-friendly layout, and then, the information is sync with the aid of the user
to their active account. Each account is related to a fixed of wearable gadgets
and viable clinical gadgets. While a chunk of fitness facts generated, it is going
to be uploaded to the blockchain community for report preserving and integrity
protection.
• Healthcare company. In healthcare companies, doctors are appointed by a posi-
tive person to perform medical check, provide a few recommendations, or provide
clinical treatment. In the meantime, the medical remedy records may be uploaded
to the blockchain network for sharing statistics with different healthcare carriers
below the user’s permission. And the contemporary healthcare provider may
request access to previous fitness data and clinical remedy from the person.
Every information request and the records corresponding get right of entry to
are recorded at the blockchain.
• Medical insurance enterprise. User might also request a quotation of health
insurance from medical insurance businesses or marketers to choose a right health
coverage plan. To provide higher insurance regulations, insurance companies
request facts get admission to from customers consisting of user health facts
from wearable gadgets and scientific remedy history. Customers with preceding
clinical remedy(s) may also want to pay a better rate, and the history cannot be
denied by way of customers to prevent insurance fraud. Users can pick out not to
percentage workout facts because of privacy problems, but commonly they would
preference to percentage because everyday workout can bring down the insurance
pay price. But customers cannot conceal or adjust medical treatment history facts
considering the fact that the once data is permanently recorded on the blockchain
community and the integrity and trustworthiness is ensured. Furthermore, the
coverage claims also can be recorded at the blockchain.
• Blockchain community. The blockchain network is used for three functions. For
fitness records accumulated from each wearable gadgets and healthcare carriers,
538 I. Zia et al.

each of the hashed information entry is uploaded to the blockchain community

for integrity safety. For private fitness information, get admission from healthcare
company and medical insurance business enterprise, and every of the information
that gets right of entry to request must be processed to get a permission from the
owner with a decentralized permission control protocol. The received entry to
manipulate policies must be saved in a dispensed manner at the blockchain which
ensures balance. Besides, each of the get entry to request and access pastime ought
to be recorded on the blockchain for in addition auditing or investigation.
• Cloud database. The cloud database stores consumer health associated statistics,
records requests from the healthcare company and coverage businesses, records
get entry to document and records access manipulate policy. Statistics get admis-
sion to is responsible and traceable. Once information leakage is detected, the
malicious entity may be diagnosed.

3.3 Device Implementation

The device captures a personal health data using a gadget which is wearable which
can be a smart watch or any tracker. It can be used in the tracking of medical utiliza-
tion. The private information can also be stored in cloud where the user can buy a
subscription of the cloud services with sufficient storage which can store abundant
amount of data (Fig. 2).
Private health data requires integrity and validation of the data. All the public data
which are not that private information is moved from the user computer or tools to
the cloud server for the validation of the data. The wearable gadget which is in the
form of smart watches or tracker is used to collect the daily motion data of the user
which are maintained as a statistic. Merkle tree model is used for the scalability of
the product. Merkle tree is a tree structure which is in a form of binary; it contained
an information in the form hash. Based on the time it is generated, the information
is ordered. In Merkle, a pair is maintained of every two nodes, and the time hash

Fig. 2 Personal health data collection

Integration Blockchain for Data Sharing and Collaboration … 539

Fig. 3 Personal fitness information integrity safety

becomes the leaf of the node. The following step is repeated until an unused hash is
found out which then becomes the tree of the node (Fig. 3).
It is a process to improve the record validation and authenticity of it which later
on help in the scalability of the product. Merkle algorithm becomes the root of
the transaction, and verification of the data is 1/3rd dependent on the other parties.
Storing all the hash brings a lot of advantages. The hash is very effective to find any
changes or modification it can be detected very easily (Fig. 4).
Information sharing is done using a Hyperledger. The data is stored in the cloud
server, firstly the data is given by the user, and some information is also shared by the
HealthCare provider, Insurance Company. If there is any fabrication by the client, it
is checked by the peers, it is seen from the ordering services, and there are various
channels maintained to carry out transaction. There is a Hyperledger fabric network
maintained which contains a CA, peers, ordering service and channels.

3.4 System Assessment

Our system adopts a consumer-centric version for processing non-public health data
the usage of blockchain network, ensuring the statistics possession of people, as well
as records integrity. The operations at the records facts are fantastically inter-operable
and like-minded with cutting-edge structures. Via implementing access manipulate
policies, customers can cope with their non-public records without demanding about
the privateness problems. With all of the security targets proposed in segment i
accomplished, it is far crucial to assess the machine performance, concerning to
the scalability and efficiency of the information integrity proof generation and facts
540 I. Zia et al.

Fig. 4 Information sharing and collaboration the usage of Hyperledger fiber and channel for cellular
customers

validation method. We take a look at distinctive numbers of concurrent statistics

with a variety from 1 to 10,000. Parents 5 and 6 suggest the common time value,
respectively (Figs. 5 and 6).
From these two figures, we are able to conclude that the device can deal with a
huge dataset at low latency, which indicates the scalability and efficiency of the facts

Fig. 5 Average time for

integrity evidence era
Integration Blockchain for Data Sharing and Collaboration … 541

Fig. 6 Average time for

integrity evidence validation

process. Through adopting Merkle tree method to batch statistics, we implement a

set of rules with the computation complexity of o(log2n). That is a crucial gain when
the information facts are accumulated at a high frequency.

4 Conclusions

This research paper deals with fact sharing of medical records of patient and record
of the doctor. The paper deals with the removing the extra cost for the patient. It uses
the blockchain to implement the concept. It does not deal with all the various other
aspects of the blockchain. Medrec is a medical management system which deals
with the smart contract which increases the privacy of the system. In this paper, we
implement a mobile healthcare system for collection and monitoring of the system
and providing an efficient result according to it. The system deals with the privacy,
integrity, and validation of the data.

References

1. OECD.Health.spending (2018) https://doi.org/10.1787/8643de7e-enhttps://www.oecdilibrary.

org/content/data/8643de7e-en
2. Kuo T-T, Kim H-E, Ohno-Machado L (2017) Blockchain distributed ledger technologies
for biomedical and health care applications. J Amer Med Info Assoc 24(6):1211–1220.
1093/jamia/ocx068. URL https://doi.org/10.1093/jamia/ocx068
3. Dhillon V, Metcalf D, Hooper M (2017) Blockchain in health care. Apress, Berkeley, CA,
pp 125–138. ISBN 978–1–4842–3081–7. https://doi.org/10.1007/978-1-4842-3081-7_9. URL
https://doi.org/10.1007/978-1-4842-30817_9
4. Engelhardt MA, Espinosa D (2017) Hitching healthcare to the chain : an introduction to
blockchain technology in the healthcare sector. An Introduct Blockchain Technol Healthcare
Sect 7(10):22–35
542 I. Zia et al.

5. Randall D, Goel P, Abujamra R (2017) Blockchain applications and use cases in health infor-
mation technology. J Health and Med Info 8(3):8–11. https://doi.org/10.4172/2157-7420.100
0276
6. Sharma R (2018) Blockchain in Healthcare. Accessed 18 Oct 2018. http://www.fccco.org/upl
oads/publications/Blockchaininhealthcare_FCCCO_RS.pdf
7. Nakamoto S (2008) Bitcoin: a peer-to-peer electronic cash system
8. Rogaway P, Shrimpton T (2004) Cryptographic hash-function basics: Definitions, implications,
and separations for preimage resistance, second-preimage resistance, and collision resistance.
In: International workshop on fast software encryption. Springer, pp 371–388
9. LeMahieu C (2018) Nano: a feeless distributed cryptocurrency network. 2018 Accessed 10
November 2018. https: //nano.org/en/whitepaper
10. Otte P, de Vos M, Pouwelse J (2017) Trustchain: a sybil-resistant scalable blockchain. Future
Gener Comput Syst
11. Saleh F (2018) Blockchain without waste: Pproof-of-stake
12. Bell EA, Ohno-Machado L, Adela Grando M (2014) Sharing my health data: a survey of
data sharing preferences of healthy individuals. In: AMIA annual symposium proceedings.
American Medical Informatics Association, pp 1699
13. Behner FWP, Hecht M-L (2018) Fighting counterfeit pharmaceuticals. Accessed 15 Nov 2018.
https://www.strategyand.pwc.com/reports/counterfeit-pharmaceuticals
14. Script—Bitcoin Wiki (2018) Accessed 15 Nov 2018. https://en.bitcoin.it/wiki/Script
15. Wood G (2014) Ethereum: a secure decentralised generalised transaction ledger. Ethereum
project yellow paper 151:1–32
16. Peterson J, Krug J, Zoltu M, Williams A, Alexander S (2018) Augur: a decentralized oracle and
prediction market platform. Accessed 18 October 2018. https://www.augur.net/whitepaper.pdf
17. 2018 (accessed October, 2018). URL https://0xproject.com/pdfs/0x_white_paper.pdf. 0x: An
open protocol for decentralized exchange on the Ethereum blockchain
18. LeMahieu C (2018) NEM–distributed ledger technology. Accessed 10 Nov 2018. https://
nem.io/
19. Androulaki E, Barger A, Bortnikov V, Cachin C, Christidis K, De Caro A, Enyeart D, Ferris
C, Laventman G, Manevich Y et al (2018) Hyperledger fabric: a distributed operating system
for permissioned blockchains. In: Proceedings of the thirteenth eurosys conference. ACM, pp
30
20. Iris (2018) Develop the world’s new financial system. Accessed Nov15 2018. https://www.ste
llar.org/
21. Neo-project. NEO. https://neo.org/
22. Zheng Z, Xie S, Dai H, Chen X, Wang H (2017) An overview of blockchain technology: archi-
tecture, consensus, and future trends. In: Big data (BigData Congress), 2017 IEEE international
congress on IEEE, pp 557–564
23. Kish LJ, Topol EJ (2015) Unpatients-why patients should own their medical data. Nat
Biotechnol 33(9):921–924
24. Clippinger JH (2017) Why self-sovereignty matters. https: //idcubed.org/chapter-2-self-
sovereignty-matters/ Online; Accessed 7 March 2017
25. Liang X, Shetty S, Tosh D, Kamhoua C, Kwiat K, Njilla L (2017) Provchain: a blockchain-based
data provenance architecture in cloud environment with enhanced privacy and availability. In:
International symposium on cluster, cloud and grid computing. IEEE/ACM
26. Tosh DK, Shetty S, Liang X, Kamhoua CA, Kwiat KA, Njilla L (2017) Security implications
of blockchain cloud with analysis of block withholding attack. In: Proceedings of the 17th
IEEE/ACM international symposium on cluster, cloud and grid computing, ser. CCGrid ’17.
Piscataway, NJ, USA: IEEE Press, pp 458–467
27. T. O. of the National Coordinator for Health IT (ONC) (2016) the National Institute for
Standards, and T. (NIST). Use of blockchain in healthcare and research workshop
28. Cachin C (2016) Architecture of the hyperledger blockchain fabric. In: Workshop on distributed
cryptocurrencies and consensus ledgers
Integration Blockchain for Data Sharing and Collaboration … 543

29. Author F, Author S (2016) Title of a proceedings paper. In: Editor F, Editor S (eds)
CONFERENCE 2016, LNCS, vol 9999. Springer, Heidelberg, pp 1–13
30. Author F, Author S, Author T (1999) Book title, 2nd edn. Publisher, Location
31. Author F (2010) Contribution title. In: 9th international proceedings on proceedings. Publisher,
Location, pp 1–2
32. LNCS Homepage (2016). http://www.springer.com/lncs. Last Accessed 21 Nov 2016
Communication Technologies, Security
and Privacy
Reusability Estimation
of Object-Oriented (O–O) Systems Using
SOM Technique

Bharti Bisht and Parul Gandhi

Abstract Due to recent changes and advancements in software sector, there was
an alarming need to cut down time as well as efforts required during complete soft-
ware development process. While constructing and designing product and services,
assuring the quality of software product is very important aspect in order to boost
market value of that product. To attain both quality as well as productivity objectives,
there is an urge for software reusability. Reusability is an essential measure that can
be used to improve overall software quality with lesser cost and efforts. Data Mining
is a process to extract useful data from a larger set containing raw data. It also means
analyzing patterns from large collection of data. In Object-Oriented (O–O) Soft-
ware Systems, class reusability can be easily obtained by inspecting metrics values.
Inspecting metric values would save time to develop software components from
initial phase. In this paper various software metrics related to software reusability of
Object-oriented (O–O) software systems has been outlined with the help of MATLAB
environment. SOM Technique was applied to create clusters of metrics values that
were derived from two Java based system that were a part of COMET dataset that
was easily available online. The goal of this study was to identify relationship among
metrics values and different level of reusability of particular class. This study shows
how SOM technique can be applied on metrics with SOM grids of different sizes.
The study depicts elimination of two of the software metrics i.e. NOC and DIT due
to poor distribution depicted by both of them.

Keywords Object-oriented software system · Software metrics · Reusability ·

SOM

1 Introduction

Reuse based engineering [7] is an approach in which the software development

process is organized in order to reuse existing software components. This approach

B. Bisht · P. Gandhi (B)

FCA, Manav Rachna International Institute of Research and Studies Faridabad, Faridabad, India
e-mail: parul.fca@mriu.edu.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 547
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_40
548 B. Bisht and P. Gandhi

comes with many advantages such as reduced cost as well as time and even contributes
in improving software quality. Metrics represents standard of measure of particular
property that a software system possess. These also act as a indicator of quality of
software system that will help managers of particular project to control software
development process and will also help software developers to analyze software
quality [15]. Metrics plays an important role in measuring reusability of OO (Object-
Oriented) systems. OO (Object-Oriented) approach is even contributing towards
construction of software in component form [7] and also plays an important role in
reuse process. In current scenario, there is much demand for OO (Object-Oriented)
techniques that would directly contribute in improving reusability level of software
[10]. Software Reusability is one of the remarkable approach [8] to reveal all those
artifacts [10] from extant components for formulating new software systems.
According to [11], software reusability is distinctive approach for a software
component that portrays software’s proficiency of reuse [3]. Reusability also points
out the reuse proficiency of particular software component [3]. It means that [5] if
reusability of particular software component is low, then that component would be
less reusable. From [12], reusability means the employment of hitherto software [20]
written in the form of design together with its code. This mechanism has been already
inspected generally during development evolution of different software projects. The
most imperative advantage showcased by software reusability is that it contributes
in reducing the total bugs present in software [10].
The main objective behind this study was as follows:
(1) This study attempts to use metrics as an indicator for measuring the capability
of the reusable components.
(2) This study would help to predict those components that can be reused early
during software development process.
(3) SOM technique has been used in this study would help to categorize reusable
components using OO (Object-Oriented) metrics.

2 Reusability Concepts

Researcher through its study [11] was able to explore that a efficient as well as effec-
tive software is required which can be produced in short span of time. This study also
helped to find the profitable contribution [11] of reusability in improving produc-
tivity as well as quality of software. Authors in [9] has mentioned about the main
purpose of Component based approach that this creates only reusable modules not
the complete system that can be easily reused. So through their study they highlighted
about the richness of O–O (Object oriented).In past there have been lot of studies
which have contributed towards introduction of various approaches [15] to find out
various software characteristics using different metrics. Authors through their study
[19] measured the reusability level of software using two metrics i.e. fan-in metric
[19] and fan-out metric [19]. Kayarvizhy and Kanmani [21] developed their own
metrics to compute software modules quality characteristics such as complexity as
Reusability Estimation of Object-Oriented (O–O) Systems … 549

Fig. 1 Levels of reusability. (Source Self)

well as reusability. Further in this study they used two methods. In first method using
methods present in a software component, CR (Component Reusability) [21] was
calculated whereas in second method, level of reusability was computed using metric
known as CRL (Component Reuse level) [21].
Reusability is a process of amending software [9] through extant reusable units
or modules rather than formulating a very brand new software system. These
reusable modules have been precisely tested and also being verified [22], which
make them suitable enough for contributing in the improvement of software quality
[22]. Reusability helps to increase productivity and improve quality of software. It
is not only dependent on code but also envelopes software development cycle valu-
ables like software components [8], test suites [8] documentations [8] and designs
[8]. Reusability can also be attained by adopting software metrics approach. Metrics
play an important role during software development process and as well as manage-
ment of software. Organization uses different metrics to apply for their software
development phase which includes various reuse metrics, Object-oriented metrics as
well as quality metrics. These metrics also helps in shaping reuse practices applied
during entire software development life cycle [11]. Software components which are
developed within any organization are not only confined to be used in that particular
environment but also can be dispersed as object code and can be easily reused by
any other environments also.
Figure 1 represents different Reusability levels in Software Development life-
cycle:

2.1 Code Reusability [8]

This is the most common level of software reuse approach. This is applied during
implementation stage of software development life-cycle.

2.2 Design Reusability [8]

This one is considered as the highest reusability level. In this, software design
miniature is reused.
550 B. Bisht and P. Gandhi

2.3 Specification Reusability [8]

In this, issues at the specification phase which occurs due to inefficient reusable code
[8] is removed.

2.4 Application System Reusability [8]

This is the special case of reusability as this helps to reuse complete software system.
They are combined and even their accuracy is also assessed.
The different benefits provided by Software Reusability are [18]:
• Dependability is increased
• Improves Productivity
• Accelerated Development [8]
• Effectiveness Increased [8]
• Reduces Operational costs [15]
In this study SOM technique has been used through which it is very easy to
categorize the components that are easily reusable.

3 SOM Technique

This is one of the familiar pattern recognition as well as classification algorithm. It

is one of the Artificial Neural Network model that represents competitive learning
[7] and is also unsupervised technique. SOM uses arranged collection of neurons
in the form of two-dimensional grid. This reduction of high-dimensional grid into
two-dimensional grid is very useful in data visualization technique which in turn
helps to determine similar patterns [6]. It helps to transfer incoming input signals
into 2-D map [8] and even explore underlying structure of input.
SOM contains 2 neuron layers i.e. input and output layers. Every input layer
is connected to every output layer. The degree of minutiae’s of findings helps to
determine SOM grid size [9], more generalized findings requires less sized grid.
Similar kinds of clusters are represented as neighbors in grid but different kind of
clusters are at very greater distance to each other.
Rotaru and Dobre [29] used SOM technique to details of various components, for
characterizing these software modules present in repositories that are extracted from
manual of software and the results obtained during the study were propitious. In [31],
authors proposed new technique i.e. SSOM technique in order to classify software
components. In [25], authors have applied SOM technique in analyzing software
components with the help of different metrics values for detecting quality dimen-
sions. This approach helped designers to investigate and have better understanding
Reusability Estimation of Object-Oriented (O–O) Systems … 551

Fig. 2 SOM architecture M0 M1 M2

[10]

Y Y Y
0 1 …… n

of software clusters. To describe the self-organizing mechanism, the collection of

input variables can be defined as real vector [7] y = {a1 , a2 …an }T ź Rn . SOM array
contains I neurons, each neuron is associated with real vector ni = {bi1, bi2… .bin }T
ź Rn . Each input variable has an image on SOM array, which can be described as an
element of an array nc that meets with y. This matching is calculated with the help of
distance function i.e. d(y, ni ).Applying this function to all input variables is defined
as a mapping of n-dimensional input space to 2-D SOM array [7]. During learning
phase, when these real vectors are amended, the nodes that are near in the array to
the present matching node will stimulate other to grasp values of attributes obtained
from input vector y.
Figure 2 represents architecture of SOM containing 3 clusters and n input vectors:
M0 .
Figure 3 represents different steps involved in SOM.
In this study SOM technique has been used as it is very easy to interpret as well as
understood the data. It is also an easy method to find out similarities among various
software modules.

4 Methodology Used

The main aim of conducting this study is to apprehend, analyze and project effects
of values of software modules metric on level of reusability. In this study, SOM
technique has been used to define various reusability levels. This study contains
following stages shown below by Fig 4.
552 B. Bisht and P. Gandhi

Fig. 3 Steps involved in

SOM. (Source Self) Input
Vector

Data is
preprocessed

Map is
initialized

Map is
trained

Map is
visualized

Results are
analyzed

Information is
extracted

Fig. 4 Stages of study

4.1 Collection and Preprocessing of Data

For this study, online available COMET dataset [10] available online is taken as input
dataset. Semi-automated approach [7] was applied on datasets in order to measure
values of software metrics which are used in this study. Complete information of
Reusability Estimation of Object-Oriented (O–O) Systems … 553

classes with their attributes and the methods used by the particular class were used
as input vector.
Class level metrics are used for experimenting with SOM technique. The various
metrics used for study are:
• Weighted Methods per Class (WMC)
• Response for a Class (RFC)
• Lack of Cohesion in Methods(LCOM) [4]
• Depth of Inheritance Tree of a Class (DIT) [13]
• Number of Children (NOC)
• Coupling between Objects(CBO) [15]
The values of above used metrics were collected [6] from 2 selected systems i.e.
Eclipse JDT [5] and Hibernate [5].
Table 1 provides the minimum, maximum, standard deviation, mean, median
values for different metrics used in study (NOC, DIT, LCOM, WMC, CBO and RFC
[19] for Eclipse JDT system [5] which is used during study [10].
Table 2 provides the minimum, maximum, standard deviation, mean, median
values for different metrics used in study (NOC, DIT, LCOM, WMC, CBO and RFC
[19] for Hibernate system [5] which is used during study [10].
The above analytical descriptions of both Python programs revealed that NOC
and DIT metric have low median value as well as mean value. This results shows
that inheritance is not a part of these programs which were used as input dataset.
Table 3 represents the threshold value for class level metrics used in this study
[19].

Table 1 Analytical description (Eclipse JDT system)

WMC NOC DIT RFC CBO LCOM
Minimum value 1 0 1 0 0 0
Maximum value 2076 24 9 664 227 80,987
Standard deviation value 158.26 1.96 1.77 65.87 21.38 2765.67
Mean value 64.61 0.52 2.68 42.55 17.88 234.37
Median value 24 0 2 25 7 6

Table 2 Analytical description (Hibernate system)

WMC NOC DIT RFC CBO LCOM
Minimum value 0 0 1 0 0 0
Maximum value 227 106 7 337 128 1374
Standard deviation value 26.72 3.85 1.51 35.63 16.34 82.64
Mean value 24.59 0.54 2.27 37.19 16.10 18.73
Median value 14 0 2 27 15 1
554 B. Bisht and P. Gandhi

Table 3 Threshold value of class level metrics

WMC NOC DIT RFC CBO LCOM
Threshold value 21 3 3 45 14 21
Reference used [18] [18] [21] [21] [18] [16]

Threshold value for selected metrics used as quantitative approach to specify

qualitative for software quality [20] and this in turn useful to find out classes with
high risk [20].The threshold values helps developers in complete review process of
classes of particular software. Exceeded threshold upper bound value gives a sign of
problem and helps to find out error in the design of class of particular software [34].
During data preprocessing stage, there should be preparation of data that can easily
available to be used by any of the clustering techniques [18].The dataset which has
been used contains different Object-Oriented classes with metrics value -1 which
shows that particular class don’t exist in a particular time.
In this study, the used dataset is normalized by the Min-Max normalization
technique [12]

vi − min
vi = (new_ max A −new_ min A )
max A − min A
+ new_ min A (1)

4.2 Classification by SOM Technique

This study uses Silhouette score technique to validate clustering process. This method
can be used to analyze the distance separating different resulting clusters [11]. Silhou-
ette graph shows how close different points in a cluster to the different points of
neighboring cluster to previous one. If the coefficient value is very near to + 1 then
this means that the selected cluster is very far from the neighboring one. Value of
0 tells that selected cluster is very near to or on boundary of neighboring one and
negative one tells that the selected clusters are not correct one to be selected for
clustering process.
The steps which are followed to find out Silhouette score of data set S containing
m objects are as follows:
• Find the average distance between object j ź S and other objects part of that cluster
to which j belongs. Let this value be p(j)
• Now calculate minimum average distance [4] from j to all clusters of which j is
not a part. Let this value be q(j)
• Silhouette score s(j) is calculated using below Eq. 2 [20]:
Reusability Estimation of Object-Oriented (O–O) Systems … 555

Fig. 5 Simple learning

network [11]

q( j) − p( j)
s( j) = (2)
max{q( j), p( j)

Figure 5 represents Simple learning network that has been taken into the
consideration for this study:

5 Experimental Analysis

In this section experiments were analyzed separately for each Java based systems [6]
which was used as the dataset for the study. Size of SOM Grid and the epoch’s number
was updated during experiments. Values of learning rate used during experiment are:
Learning rate value of Ordering stage = 0.7
Steps of Ordering stage = 1005
Learning rate value of Tuning Stage = 0.05 Neighborhood distance of Tuning
Stage = 2.
It was observed that as epoch’s number was increasing during experiment, the
average values of Silhouette coefficient [24] was also increasing. The epoch’s number
[5] was then modified to 1000 because Silhouette coefficient became constant [8] for
all the input dataset. Highest value of Silhouette coefficient was selected and then
the value of metrics used in study was analyzed for every cluster in order to find the
connection with different reusability level of classes.
After evaluating every cluster during experiment it was observed that the process
of clustering was dominated by NOC metric as well as DIT metric. The values
of different metrics were also examined in every cluster and were not able to find
common connection between classes that are a part of same cluster [5]. The results
found during experiment showed that if NOC metric as well as DIT metric are
eliminated, then the results can be enhanced. After eliminating both metrics i.e.
NOC and DIT metric [5] again the experiments were applied.
All the experiments done on Eclipse JDT [5] are shown below by Table 4.
It can be found from above table that Silhouette Coefficient Average is changed
whenever there is a change in Cluster Size.
Figure 6 represents Silhouette Coefficient plot of Experiment 7 of Eclipse JDT
which was performed during this study:
556 B. Bisht and P. Gandhi

Table 4 Experiments done

Number of experiment Cluster size Silhouette coefficient
on Eclipse JDT system [5]
average
1 [9] 0.4938
2 [8, 9] 0.4601
3 [7] 0.4588
4 [6, 7] 0.5412
5 [6] 0.5456
6 [4, 5] 0.6345
7 [2] 0.7080
8 [2, 3] 0.6181

Fig. 6 Silhouette coefficient

plot of eclipse JDT [12]

Negative value of Silhouette Coefficient is not preferred because it indicates that

the object is much related to other cluster more than the cluster to which it belongs.
The desired values are the values which approach to 1.
The classes which are a part of different clusters can be categorized into 3 divi-
sions. These divisions summarize the relationship between class level metrics and
reusability factor [8]. The divisions are as follows:
• Division 1: Highly reusable cluster [9] consists those classes which are not
exceeding threshold values and have low values of LCOM, WMC, CBO, and
RFC metrics [11]
• Division 2: Medium reusable cluster [9] consists of those classes that are near
threshold values and even values of LCOM, WMC, CBO, and RFC metrics is
medium
Reusability Estimation of Object-Oriented (O–O) Systems … 557

• Division 3: Low Reusable cluster [9] consists those classes that are exceeding
threshold values and values of LCOM, WMC, CBO, and RFC metrics is highest
Based on above divisions, the complete analysis of each and every cluster
vectors [28] used in experiment 7 of Eclipse JDT is represented in Table 5.
In the above table, if the Avg. of any one of the metric part of particular cluster is
high then it can be concluded that the classes that are included [3] in this cluster have
metrics with higher values, so they have less reusable ability. When there is higher
percentage value of Ex. %, it can be concluded that there are lot of classes in this
particular cluster that are outrunning the threshold value, so this also becomes less
reusable [11].
From above table, it can be concluded that experiment 3 and 4 [11] show low
reusability level because almost every classes in both experiments exceeds threshold
values [6] and the average of Silhouette coefficient is very high. Experiment 2 [11]
consists of every classes that are reusable and metrics have minimum value, so this
one is Highly Reusable. Experiment 1 [11] has medium reusability level because
almost every classes in this experiment exceeds threshold value [6] and average of
Silhouette coefficient is lesser than Experiment 3 and 4 but more than experiment 2
[11].
Hibernate as well as Eclipse JDT systems were analyzed many times [3] to find
out the highest average value of Silhouette coefficient. In Eclipse JDT, value of
NOC (Number of Children) metric is 0 in 81.2% of the classes and 80.5% classes
have values 4, 5 and 6 for DIT (Depth of Inheritance Tree of a Class) metric [4]. In
Hibernate, value of NOC (Number of Children) metric is 0 in 82.3% of the classes
[6] and 76.5% classes have values of 4 and 5 for DIT (Depth of Inheritance Tree of
a Class) metric [5]. Therefore it can be concluded that DIT as well as NOC metric
had poor distribution of classes, so these 2 metrics were removed from input vector
used in experiments done in Eclipse JDT [7].
All the experiments done on Hibernate [5] are shown below by Table 6.
Figure 7 represents Silhouette Coefficient plot of Experiment 7 of Hibernate which
was performed during this study:
The complete analysis of each and every cluster vectors used in experiment 7 of
Hibernate is represented in Table 7.

Table 5 Analysis of experiment 7 done on eclipse JDT system [7]

Cluster WMC RFC CBO LCOM No of classes
number
Avg. Ex. % Avg. Ex. % Avg. Ex. % Avg. Ex. %
1 77.5 95.2 63.5 82.4 23.6 95.2 102.1 45 223
2 14.6 27.5 14.4 2.4 5.8 7.8 14.4 27.8 703
3 1090.5 100 5 100 135.2 100 9054 100 15
4 254.3 100 152. 3 100 52.5 97.3 1211.2 76.7 74
Threshold 22 46 15 22
558 B. Bisht and P. Gandhi

Table 6 .
Number of experiment Cluster size Silhouette coefficient
average
1 [10] 0.4555
2 [9, 10] 0.4458
3 [8] 0.4257
4 [7, 8] 0.5299
5 [6] 0.5301
6 [5, 6] 0.6137
7 [2] 0.8011
8 [2, 3] 0.7097

Fig. 7 Silhouette coefficient

plot of hibernate [12]

Table 7 Analysis of experiment 7 done on hibernate [10]

Cluster number WMC RFC CBO LCOM No of classes
Avg. Ex. % Avg. Ex. % Avg. Ex. % Avg. Ex. %
1 43.7 82.5 72.6 79.4 28.4 99.5 253.8 57 185
2 7.8 5.6 12.7 0.9 7.7 12.7 10.4 11 969
3 463 100 588 100 128.9 100 12958 100 2
4 221.7 100 275.6 100 85.7 100 2825.9 89.6 19
Threshold 22 46 15 22

From the above table it can be concluded that Cluster number 3 has high values
with two vectors for all class level metrics used in this study [11]. Cluster Number 3
as well as Cluster Number 4 have low reusability value. Cluster Number 2 has high
reusability value because all metrics have minimum values [10]. Cluster number
1 has medium reusability level as the classes in this cluster exceeds the threshold
Reusability Estimation of Object-Oriented (O–O) Systems … 559

value and average of silhouette coefficient is lesser than cluster number 3 and cluster
number 4 but greater than cluster number 2 [13].

6 Conclusion and Future Work

In this paper, SOM Technique is used to cluster class level metrics. This clustering
was build on threshold values of class level metrics. Through this paper we have tried
to show that for clustering metrics, SOM technique is applied and even this technique
helps to visualize relationship between metrics and reusability level of software.
SOM technique is used due to its capability of clustering data and even its special
property of spatial autocorrelation [14]. This technique also helped in finding out
patterns in class level metrics and the relation with different reusability level of soft-
ware. Validity of clustering was done on basis of highest average value of silhouette
coefficient after applying different sizes of grid and different epoch’s number. At first
stage, we applied SOM technique on all class level metrics. We found that DIT and
NOC metric [27] influenced results of clustering process due to their poor distribu-
tion [11]. The results of experiments revealed that clustering process becomes more
meaningful [13].
The future study would investigate SOM technique on more metrics to get much
better results than the current results also. We would use same dataset of both systems
with elimination of three metrics i.e. NOC,DIT and LCOM due to poor distribution
[12].

Acknowledgements I would like to acknowledge Dr Parul Gandhi, Professor and Ph.D. Coor-
dinator in Faculty of Computer Applications, MRIIRS, Faridabad for her continuous support and
helpful guidance during the preparation time of this article.

References

1. Abreu B, Carapuca R (1994) Candidate metrics for object-oriented software within a taxonomy
framework. J Syst Softw 26:87–96
2. Aggarwal K, Singh Y, Kaur A, Malhotra R (2006) Empirical study of object-oriented metrics.
J Object Technol 5(8):110–115
3. Arifa B, Mohamed N, Archana K (2013) Study of software reusability in software components.
IJET 5:2455–2460
4. Amin F, Mahmood A, Oxley A (2011) Reusability assessment of open source components for
software product lines. Int J New Comput Archit Appl (IJNCAA) 1(3):519–533
5. Alexandre A, Almeida E, Meira S (2006) A software component quality model: a preliminary
evaluation. IEEE 4:1–8
6. Barnard J (1998) A new reusability metric for object-oriented software. Software Qual J
7(1):20–25
7. Basili J, B. V.R, Melo W (1996) How reuse influences productivity in object-oriented systems.
Commun ACM 30(10):104–114
560 B. Bisht and P. Gandhi

8. Bhatia P, Mann R (2009) An approach to measure software reusability of OO design. COIT-

2009, pp 26–30
9. Budhija N, Singh B, Ahuja S (2013) Detection of reusable components in object oriented
programming using quality metrics. Int J Adv Res Comput Sci Softw Eng 3(1)
10. Caldiera G, Basili V (1991) Identtifying and quantiffying reusable software components. IEEE
pp 61–70
11. Chahal K, Singh H (2008) A metrics based approach to evaluate design of software components.
IEEE 269–272
12. Chidamber R, Kemerer F (1994) A metrics suite for object-oriented design. IEEE Trans Softw
Eng 20(6):476–492
13. Deeba K, Amutha B (2016) Classification algorithms of data mining. Indian J Sci Technol
9:2–5
14. Gandhi P, Bhatia K (2010) Reusability metrics for object-oriented system: an alternative
approach. IJSE 1(4):63–72
15. Gandhi P, Bhatia PK (2011) Estimation of generic reusability for object-oriented software: an
empirical approach. ACM 36(3):1–4
16. Gill SN (2006) Importance of software component characterization for better software
reusability. In ACM SIGSOFT Softw Eng Notes 31:1–3
17. Goel B, Bhatia P (2013) Analysis of reusability of object-oriented systems using object-oriented
metrics. ACM 38:1–5
18. Hudaib A, Huneiti A (2016) Software reusability classification and prediction using self-
organizing map (SOM). Commun Netw 8:179–192
19. Bhagwan J, Oberoi A (2011) Software modules clustering: an effective approach for reusability.
J Inform Eng Appl 1(4)
20. Kakkar P, Sharma M, Sandu P (2012) Modeling of reusability of procedure based software
components using Naïve Bayes classifier approach. Int J Comput Appl 55:12–17
21. Kayarvizhy N, Kanmani S (2011) Analysis of quality of object-oriented systems using object-
oriented metrics. ICECT 2:25–32
22. Maheswari G, Chitra K (2019) Enhancing reusability and measuring performance merits of
software component. Int J Innov Technol Explor Eng 8:1577–1583
23. Manhas S, Vashisht R, Sandhu P, Neeru N (2010) Reusability evaluation model for procedure
based software systems. Int J Comput Electri Eng 2:1107–1110
24. Mateen A, Kausar S, Sattar A (2017) A software reuse approach and its effect on software
quality, an empirical study for the software industry. Int J Managem IT Eng 7(2):266–279
25. Morisio EM, Tully C (2002) Success and failure factors in software reuse. IEEE Trans Softw
Eng 28:340–357
26. Nyasente S, Mwangi P, Kimani S (2014) A metrics-based framework for measuring the
reusability of object-oriented software components. J Inform Eng Appl 4:71–84
27. Olague H, Etzkorn HL, Gholston S, Quattlebaum S (2007) Empirical validation of three soft-
ware metrics suites to predict fault-proneness of object-oriented classes developed using highly
iterative or agile software development processes. IEEE Trans Software Eng 33:402–419
28. Padhy N, Singh R, Satapathy (2018) Software reusability metrics estimation: algorithms,
models and optimization techniques. Elsevier 69:653–66
29. Rotaru P, Dobre M (2014) Reusability metrics for software components. IEEE 21:24–29
30. Sandhu P, Singh H (2006) A reusability evaluation model for OO-based software components.
Int J Comput Sci 1(4):259–264
31. Sharma A, Grover PS, Kumar R (2009) Reusability assessment for software components. ACM
SIGSOFT Softw Eng Notes 34:1–6
32. Shatnawi A, Abdelhak DS (2013) Mining reusable software components from object-oriented
source code of a set of similar software, information reuse and integration. In: 2013 IEEE 14th
international conference on IEEE, pp 2–18
33. Singh G (2013) Metrics for measuring the quality of object-oriented software. ACMSIGSOFT
38:1–5
Reusability Estimation of Object-Oriented (O–O) Systems … 561

34. Subramanyam R, Krishnan S (2003) Empirical analysis of CK metrics for object-oriented

design complexity: implications for software defects software engineering. IEEE Trans 29:297–
310
35. Tully C (2018) Success and failure factors in software reuse. IEEE Trans Software Eng 28:340–
357
The First Corpus for Detecting Fake
News in Hausa Language

Sukairaj Hafiz Imam, Abubakar Ahmad Musa, and Ankur Choudhary

Abstract The rapid spread of deceptive news especially in Africa has become a
global issue in last decade. This triggers the attention of the research community
to develop efficient and reliable classification approaches for fake news detection
so as to prevent its spread in the community. It has been explored that fake news
in regional languages spread with a faster pace as compare to English language in
local regions. Hausa is a very common language in Nigeria and some West African
countries. So, it opens the challenge to detect the fake news in Hausa language.
This paper presents the first corpus for the detection of fake news in Hausa. A
dataset has been formed by collecting the labeled real and fake news consists of
2600 articles. In order to classify the fake news in Hausa, six different classifiers have
been utilized. The performance of these approaches is then evaluated over different
metrics and compared to determine the best model on the Hausa language dataset.
The experimental results indicate that support vector machine (SVM) outperformed
the other classifiers used by achieving 85% accuracy while AdaBoost happens to
emerge as the fair model with 70% accuracy.

Keywords Hausa · Corpus · Machine learning (ML) · Natural language

processing (NLP) · LSTM · Bi-LSTM · CNN · RNN

1 Introduction

It is not wrong to say that deception is the act of creating and spreading information
across the globe with the sole aim of misleading people or achieving a negative goal
[1–3]. Nowadays, the emergence of internet and online social media platforms has
speed up the pace of news spreading, which has its own pros and cons [4]. The positive
side of this spreading is the important news reached to the population very easily but
the negative side is fake news. Fake news is a good example of this deception which
has been popular these days, as several fake news contents have been initiated and

S. H. Imam (B) · A. A. Musa · A. Choudhary

Department of Computer Science and Engineering, Sharda University, Greater Noida, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 563
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_41
564 S. H. Imam et al.

propagated, thereby creating a lot chaos and abnormalities within the environment
[5, 6]. This is a global problem that requires an immediate attention.
Hausa is adherent of the Afro-asiatic and is the most broadly spoken language
within the Chadic branch of that family. Ethnologies estimated that it was spoken
as a first language by some 47 million people and as a second language by
another 25 million, bringing the total number of Hausa speakers to an estimated 72
million. According to more recent estimations, Hausa would be spoken by 100–150
million people.1
In line of this huge population, the spread of unauthentic news is a great threat to
the people speaking Hausa language because it is ease to dismay the society.
Taking Nigeria as a case study, where most of the northern parts of its populace
are Hausa speakers, fake news contents have been dispersed through social media
platforms like Facebook, Twitter. This leads to the occurrence of violence which
destroys the lives and properties of thousands of people. For example: A fake news
broadcasted by Nigerian Television Authority (NTA) in November 1989, related to
the death of the first Governor General and President of Nigeria, Dr. Nnamdi Azikwe.
After listening this news, many newspapers had published it on their front page.
It takes much effort and time before Azikwe proved his health status and convinced
the world that president was still alive and the wrong news was managed. In 2019
(Thirty years later), rumors spread that President Buhari had died during one of
his prolonged nonappearances from Nigeria in 2017 due to medical issues. The
imaginary death of President Buhari spread like wildfire on top social media portals
like Twitter, Facebook and WhatsApp. Most recently was the fake news on vital
issues like: religious matters and ethics, security issues, coronavirus, damaging the
reputation of respected personnel and government officials, as such, the negative
impact, danger and harmful challenges of fake news written in Hausa language
cannot be overemphasized.
Literature revealed that lots of techniques have been employed to identify fake
news so as to limit its danger. This ranges from traditional based approaches where
individuals try to distinguish between real and fake news, considering some basic
features of the news content, and modern artificial intelligence-based approaches. It
also comprises of natural language progressing techniques, machine learning-based
classifiers, deep learning-based models, recurrent learning-based approaches, and
data mining-based techniques [1, 3, 4, 7–20].
These technology-driven solutions proposed in the literature have proven their
performance mainly on English language-based content [10–13, 15–21] but the
approaches utilized for other language news content are relatively very a few as
compare to English language [1, 4, 7–9]. However, currently no effort has been
made public on the automatic detection of fake news written in Hausa language
despite the large size of its speakers and the danger of misleading its huge populace.
This paper aims to fill this gap by presenting a corpus for detecting fake
news written in Hausa language. The contribution proposed by this paper can be
summarized as follows:

1 https://en.wikipedia.org/wiki/Hausa_language.
The First Corpus for Detecting Fake News in Hausa Language 565

• The paper is solely on Hausa Language.

• The first labeled dataset has been created for detection of fake news in Hausa
language. The dataset consists of 2600 news articles comprising of real and fake
news selected from key topics like: Business, health, entertainment, sports, politics
as well as religion.
• Furthermore, we studied and employed some of the most used machine learning
classifiers to examine and compare their performance on this dataset. Hoping that
this work will serve as a baseline approach for Hausa-based fake news detection
systems.
• The different parts of this paper comprise of Sect. 2—related literature review,
Sect. 3—the Hausa corpus and the methodology used, Sect. 4—presentation and
discussion of results, and Sect. 5 which concludes the whole process.

2 Related Work

In the past years, many researches have been done for detecting fake news. Super-
vised and unsupervised machine learning is the most popular algorithms used by the
researchers. Furthermore, few researchers used deep learning and semi-supervised
algorithms.
Most of the work done for detecting fake news were English-based [1, 10–13, 15–
25]. Few researches were carried out using non-English medium. Al-yahya et al. [26]
used NN and transformer-based language models for Arabic fake news detection and
compared their performance to each other where transformer-based outperform NN.
Pratiwi et al. [27] built an automatic method for detecting fake news in Indonesian
language where PHP-ML library was utilized to classify between real and fake news
content. Mansur and Onan et al. [28] presented a dataset for satirical text classifica-
tion in Turkish language in which nine different surveys carried out on non-identical
text classification. Naïve Bayes, LR, C4.5, and SVM were employed and SVM got
the best accuracy of 89.70%.Vogel and Jiang et al. [22] presented the first dataset in
German language for fake news detection. SVM and CNN resulted an outstanding
performance of 72 and 89%, respectively. Alves et al. [29] carried out a survey on
the 2018 Brazilian election record shared over the social media platform. LSTM
(Naïve), bidirectional LSTM (Bi-LSTM), and deep Bi-LSTM achieved the highest
accuracy. Duran et al. [30] trained different algorithms on lexical features such as
bag of words (BOW), part of speech (POS tags), n-grams, and n-grams combination
for confirming the authenticity of news content in Spanish language. The outcome
revealed that SVM outperformed the rest of the algorithms used. Alzanin and Azmi
[31] utilized supervised Gaussian Naïve Bayes (GNB) and semi-supervised expec-
tation–maximization (E–M) to detect fake news in Arabic tweets. The supervised
approach excelled the semi-supervised resulting up to 78.6% accuracy. Vicario et al.
[32] proposed a scheme that identifies a polarizing news content on social media to
forecast the authenticity of news in Italian Facebook dataset with the aid of different
classifiers like LR, SVM, KNN, NN. Kim et al. [33] used unified key sentence
566 S. H. Imam et al.

information in Korean article dataset for distinguishing the genuineness of the news
content by using bilateral multi-perspective matching model using five layers. Liu
et al. [34] presented a French dataset for detecting sarcasm news using machine
learning algorithms in which logistic regression outpaces the rest with 92.17%. Silva
et al. [1] presented an extensive survey on how to automatically classify different
news contents as fake or real in Portuguese. SVM and logistic regression performed
better in the different experiments conducted. Hossain et al. [4] created the first
dataset in Bangla and compare between human and machine performance in clas-
sifying fake and real news content. Different models like SVM, LR, RF, LSTM,
CNN, and Bidirectional Encoder Representations from Transformers (BERT) were
employed. Amjad [7] provided the first dataset in Urdu language for confirming news
authenticity. SVM, LR, RF, multinomial Naïve Bayes, Bernoulli Naïve Bayes, DT,
and AdaBoost were used. However, AdaBoost was resulted to be the most accurate
compared to others.

3 The First Corpus for Detecting Fake News in Hausa

Language

This section provides an overview of the data collection process as well as the compi-
lation statistics. We collected thousands of real news articles from different reliable
sources, a special thanks to Abubakar Rimi Television, Kano, for providing us the
news scripts contains more than one thousand authentic news. This dataset compiled
news content from six different domain: (i) Business domain, (ii) health domain, (iii)
entertainment domain, (iv) sports domain, (v) politics as well as (vi) religious. The
next subsections discuss dataset details such as: news sources, news domains, etc.

3.1 Authentic and Fake News Collection

Different authentic and reputed news channels, radio stations and newspaper sites
have been chosen for collecting the data and creating the news dataset. The major
points keep in to consideration in handling and collecting news data is as follows:
• The dataset was prepared by collection and annotation of data manually.
The news was considered as authentic if any of the condition is satisfied:
• If the news was published in trustworthy newspaper and news agency.
• If the same news was published by different newspapers with evidence (Fig. 1).

Table 1 shows the different chosen sources for authentic news.

Table 2 described the categories of news.
The First Corpus for Detecting Fake News in Hausa Language 567

Fig. 1 Frequency of news

by category in Hausa corpus

Table 1 Sources of authentic news

S. no Source Origin Url
1 Abubakar Rimi television, Kano Nigeria www.artvkano.org
2 Freedom Radio, Kano Nigeria freedomradionig.com
3 Aminiya Nigeria aminiya.dailytrust.ng.com
4 Guarantee Radio Nigeria guaranteeradio.com
5 Daily Nigerian Nigeria dailynigerian.com
6 Legit Hausa Nigeria hausa.legit.ng
8 Sahara Hausa Nigeria sahara.hausang.com
9 Premium Times Hausa Nigeria hausa.premiumtimesng.com
10 Rariya Nigeria rariyajarida.com
11 Leadership Hausa Nigeria hausa.leadership.ng
12 Radio Kano Nigeria radiokano.com
13 Rahama radio Nigeria rahamradio.com
14 DW Hausa Germany www.dwhausa.com
15 RFI Hausa France www.rfi.fr
16 CRI Hausa China hausa.cri.cn
17 BBC Hausa England www.bbchausa.com
18 VOA Hausa America www.voahausa.com
19 Mikiya Nigeria mikiyahausa.com
568 S. H. Imam et al.

Table 2 News categories

S. no Categories Real news count Fake news count
1 Business 125 100
2 Entertainment 151 200
3 Health 175 260
4 Politics 366 500
5 Religious 100 94
6 Sport 189 340
7 Totals 1106 1494

3.2 Pipeline: Text Data

An overview of the pipeline composed of five phases: News collection, data prepro-
cessing, feature extraction, models selected and verification of effectiveness of algo-
rithms Selected. This process described how news is collected until the detection of
final result (Fig. 2).

3.2.1 Data Preprocessing

In preprocessing, different approaches were used to remove undesired characteristics

from the collected dataset.

Fig. 2 Text pipeline

The First Corpus for Detecting Fake News in Hausa Language 569

3.2.2 Tokenization

It is the initial stage of NLP. The main advantage of this phase is to translate a word
into phrases, symbols, or other meaningful meaning known as tokens [35].

3.2.3 Extraction of Stop Words

They are the English words which are inessential and ignoring it has no any impact
to a sentence meaning? NOTE: only 22 languages have stop words in Python2 and
Hausa is not among them. Stop words were used only because few words in Hausa
resembles that of English language.

3.2.4 Stemming

It is the method of eliminating suffixes to get to the origin form of the word. It refers
to heuristics that split suffixes [36].

3.2.5 Tf-Idf

It is an advancement from IDF which was proposed by Spärck Jones [37]. The term
frequency (TF) was used in calculating the number of times a word appears in the
dataset divided by the total number of words in the dataset. The demonstration of
the equation represented by

ni, j
t f i, j =  (1)
kni, j

TF-IDF weight of term i in document j in a corpus of N documents is calculated

as:
N
Weight i, j = t f i, j ∗ log( ) (2)
dfi

(where (tfij) is a number of times term i appear in document j and (dfi) is a number
of documents containing term i). The main idea behind the TF-IDF is from the theory
of language modeling where the terms in given documents are categorized into words
with relevancy and those without relevancy for a given document. Furthermore, TF-
IDF can be used to measure the effectiveness of a term in in the document collection
[38].

2 https://pypi.org/project/stop-words/#available-languages.
570 S. H. Imam et al.

3.2.6 Doc2vec

It is an extended version of word2vec introduced by Le and Mikolov. It is applied

to a document as whole instead of individual words. Doc2Vec works on the logic
that the denotation of a word is solely depends on the document that it is present in.
The vectors computed by Doc2Vec can be utilized for finding resemblances between
documents [39–41]. In some sources, Doc2Vec referred to paragraph2vec and its
modified version of word2vec algorithm [42].

3.2.7 Performance Evaluation Metrics

Evaluation metrics are constantly used in selecting relevant algorithms, and it gives
us chance to test the effectiveness of the algorithms. It helps in comparing/evaluating
the performances of different types of algorithms for detecting fake news. The most
common metrics are true positive (TP), false positive (FP), true negative (TN), false
negative (FN), precision (Pr), accuracy (Acc) and recall (Re). The performance was
evaluated using the following criteria:

TP +TN
Acc = (3)
T P + T N + FP + FN
TP
Pr = (4)
T P + FP
TP
Re = (5)
T P + FN
2∗ Re∗ Pr
F_score = (6)
Re + Pr

The selection of this metrics is in line with the work in [3, 15].

3.3 Algorithms

This contains the detail explanation of the proposed algorithm used for detecting
fake news in Hausa language. Initially, we started by preprocessing our dataset by
removing redundancy and characters such as numbers. The application of feature
extraction helps in reducing the dimension of feature state. The last but not the
least was choosing the models that fits our dataset and provide good performance
or high accuracy. Six different algorithms were selected due to their high predictive
performance and broad use in related research papers. These include: NB [7, 9, 11,
12, 20], SVM [1, 4, 7, 9, 12, 20], AdaBoost [7, 9, 12], RF [1, 4, 7, 9, 20] and LR [7,
9, 12], PAC [43].
The First Corpus for Detecting Fake News in Hausa Language 571

3.3.1 Logistic Regression (LR)

It is used in categorizing text base on an extensive feature set, in which the paired
output is (fake (0) /real (1)). It is an algorithm which is used in classifying data base
on binary or numerous sections. Hyperparameters tuning is applied for generating
an excellent and an accurate outcome on the datasets. Mathematically, the LR can
be defined as follow:
1
h θ (X) = (7)
1+ e−(βo +β1 X )

3.3.2 Support Vector Machine (SVM)

This also used for paired categorization task and is accessible via several kernels
tasks. The objective causes the use of SVM algorithm which is to evaluate a marginal
boarder on the origin of attribute set to classified data points. The length of the deci-
sion boundary differs depending on the quantity of attributes. Different chances of
the decision boundary occurring in N-dimensional planetary. The charge is detecting
that the planetary separating the two types of data points has a strong margin. A
mathematical illustration of the SVM can be defined as:

1 2
n
J(θ) = θ (8)
2 j=1 j

Such that

θ t x (i) ≥ 1, y (i) = 1 (9)

θ t x (i) ≤ −1, y (i) = 0 (10)

3.3.3 Random Forest (RF)

The model was introduced as improvement technique of decision trees. It comprises

huge amount of decision trees, dedicated for predicting the outcome of the category,
and concluding estimate is grounded on the category with the most votes. The rate
of mistakes/errors in RF is quiet less due to little correlation among trees. There are
different algorithms that can determine the division in the decision tree based on
regression or classification problems. For categorization problems, this paper invites
the Gini index as a cost function to guess the division of the dataset. The mathematical
equation is:
572 S. H. Imam et al.


c
 2
Gind = 1 − pi (11)
i+1

3.3.4 K-Nearest Neighbors (KNN)

It is a learning mode in which an experimental variable does not need to guess

the result of the precise data. The KNN algorithm evaluates the distance between
a new data point and its closest neighbor; the advantage of K is to calculate the
preponderance of votes from its neighbors, if the value of K is 1, and amount of
data point is assigned to the closest class. The mathematical equation to evaluate the
interval between 2 dots can be considered as:

 k

Euclidean distance =  (xi− yi )2 (12)
i=1

3.3.5 Passive-Aggressive Classifier (PAC)

This classifier is used for extensive learning. This classifier became aggressive any
erroneousness or unsuitable outcome but remain reflexive if exact result is achieved
while categorizing.
The general steps for passive-aggressive classifier are as follows:
Step 1: Initialization of weights wt = [0, 0, 0, . . .].
Step 2: Select the document for classification dc = (dc1, dc2 . . . , dcn ).
Step 3: Perform TF-IDF on document and then normalize dc.
Step 4: Apply conditions if dcT wt > = 0, + 1 and -1. Based on that fix the output
as + 1 or –1.
Step 5: Compute Loss.

4 Comparison and Discussion of Result

4.1 Machine Learning Approach

The dataset was splitted into 80% for training set and 20% for testing on the selected
models using scikit-learn [44]. The implementation was carried out using DOC2VEC
and TF-IDF in which two different word embedding techniques utilized to map the
news content and compare the performance of the selected models on each technique.
The concluded result is summarize below (Figs. 3 and 4):
The First Corpus for Detecting Fake News in Hausa Language 573

Fig. 3 Model performance using DOC2VEC

Fig. 4 Model performance using TF-IDF

It is clearly shown from the above classification of results that models trained
and tested with TF-IDF outperform the models trained and tested using Doc2Vec.
The results unveiled that some classifiers did a magnificent work in differentiating
authentic news content. However, some classifiers showed fair performance. It was
observed that support vector machine (SVM) indicated an impressive performance in
detecting fake news in Hausa language with 85% using TF-IDF and AdaBoost 69%
using Doc2Vec. Classifiers such as AdaBoost to be the fair classifier using TF-IDF
and multinomial Naïve Bayes using Doc2Vec.

5 Conclusion

The first labeled dataset is presented in Hausa language; the real news contents
were generated from the internet and other legitimate sources, while the fake news
contents were drafted by the expert journalist of ARTV Kano. The dataset was
annotated manually cleaned and processed and used for this work. Six different
classifiers were employed for this experiment; however, it was observed that SVM
outperformed all the classifiers used to achieve 85% accuracy. Due to several factors,
we could only annotate 2600 news contents; we look forward to the improvement of
this dataset in the future, also the viability of incorporating other languages like the
574 S. H. Imam et al.

main three Nigerian languages (Hausa, Yoruba & Igbo) and other African languages
like Swahili.

References

1. Silva RM, Santos RLS, Almeida TA, Pardo TAS (2020) Towards automatically filtering
fake news in Portuguese. Expert Syst Appl 146:113199. https://doi.org/10.1016/j.eswa.2020.
113199
2. De keersmaecker J, Roets A (2017) ‘Fake news’: Incorrect, but hard to correct. The role
of cognitive ability on the impact of false information on social impressions. Intelligence.
65:107–110. https://doi.org/10.1016/j.intell.2017.10.005.
3. Zhang X, Ghorbani AA (2020) An overview of online fake news: characterization, detection,
and discussion. Inf Process Manag 57:102025. https://doi.org/10.1016/j.ipm.2019.03.004
4. Hossain MZ, Rahman MA, Islam MS, Kar S (2020) BanFakeNews: a dataset for detecting fake
news in Bangla
5. Kim J, Tabibian B, Oh A, Schölkopf B, Gomez-Rodriguez M (2018) Leveraging the crowd
to detect and reduce the spread of fake news and misinformation. WSDM 2018—proceedings
11th ACM international confernce web search data Mining 2018 Feburary, pp 324–332. https://
doi.org/10.1145/3159652.3159734.
6. Sadiku M, Eze T, Musa S (2018) Fake news and misinformation. Int J Adv Sci Res Eng
4:187–190. https://doi.org/10.31695/ijasre.2018.32728.
7. Amjad M, Sidorov G, Zhila A, Gómez-Adorno H, Voronkov I, Gelbukh A (2020) Bend the
truth: benchmark dataset for fake news detection in urdu language and its evaluation. J Intell
Fuzzy Syst 1–13. https://doi.org/10.3233/jifs-179905
8. Abonizio HQ, de Morais JI, Tavares GM, Junior SB (2020) Language-independent fake news
detection: english, Portuguese, and Spanish mutual features. Futur Internet 12:1–18. https://
doi.org/10.3390/FI12050087
9. Faustini PHA, Covões TF (2020) Fake news detection in multiple platforms and languages.
Expert Syst Appl 158:113503.https://doi.org/10.1016/j.eswa.2020.113503
10. Ozbay FA, Alatas B (2020) Fake news detection within online social media using supervised
artificial intelligence algorithms. Phys A Stat Mech its Appl 540:123174.https://doi.org/10.
1016/j.physa.2019.123174
11. Mahabub A (2020) A robust technique of fake news detection using ensemble voting classifier
and comparison with other classifiers. SN Appl Sci 2:1–9. https://doi.org/10.1007/s42452-020-
2326-y
12. Singh VK, Ghosh I, Sonagara D (2020) Detecting fake news stories via multimodal analysis.
J Assoc Inf Sci Technol 1–15. https://doi.org/10.1002/asi.24359
13. Kumar S, Asthana R, Upadhyay S, Upreti N, Akbar M (2020) Fake news detection using deep
learning models: a novel approach. Trans Emerg Telecommun Technol 31:1–23. https://doi.
org/10.1002/ett.3767
14. Kong SH, Tan LM, Gan KH, Samsudin NH (2020) Fake news detection using deep learning.
In: ISCAIE 2020—IEEE 10th symposium computing application industry electronics, pp 102–
107. https://doi.org/10.1109/ISCAIE47305.2020.9108841
15. Kaliyar RK, Goswami A, Narang P, Sinha S (2020) FNDNet—a deep convolutional neural
network for fake news detection. Cogn Syst Res 61:32–44. https://doi.org/10.1016/j.cogsys.
2019.12.005
16. Kaliyar RK, Goswami A, Narang P (2020) DeepFakE: improving fake news detection using
tensor decomposition-based deep neural network. J Supercomput.https://doi.org/10.1007/s11
227-020-03294-y
17. Huang YF, Chen PH (2020) Fake news detection using an ensemble learning model based
on self-adaptive harmony search algorithms. Expert Syst Appl 159:113584.https://doi.org/10.
1016/j.eswa.2020.113584
The First Corpus for Detecting Fake News in Hausa Language 575

18. Antoun W, Baly F, Achour R, Hussein A, Hajj H (2020) State of the art models for fake news
detection tasks. In: 2020 IEEE international conference informatics, IoT, enabling technology
ICIoT 2020. pp 519–524. https://doi.org/10.1109/ICIoT48696.2020.9089487
19. Wang Y, Yang W, Ma F, Xu J, Zhong B, Deng Q, Gao J (2020) Weak supervision for fake
news detection via reinforcement learning. Proc AAAI Conf Artif Intell 34:516–523. https://
doi.org/10.1609/aaai.v34i01.5389
20. Meneses Silva CV, Silva Fontes R, Colaço Júnior M (2020) Intelligent fake news detection:
a systematic mapping. J Appl Secur Res 0:1–22.https://doi.org/10.1080/19361610.2020.176
1224
21. Zhang J, Dong B, Yu PS (2020) FakeDetector: effective fake news detection with deep diffu-
sive neural network. In: Proceedings–international conference data engineering 2020-April, pp
1826–1829. https://doi.org/10.1109/ICDE48307.2020.00180
22. Vogel I, Jiang P (2019) Fake news detection with the new German dataset “GermanFakeNC”.
https://doi.org/10.1007/978-3-030-30760-8
23. Song C, Ning N, Zhang Y, Wu B (2021) A multimodal fake news detection model based on
crossmodal attention residual and multichannel convolutional neural networks. Inf Process
Manag 58:102437. https://doi.org/10.1016/j.ipm.2020.102437
24. Scott W, Bansal R, Kaushik A, Sengupta S (2021) Cross-SEAN : a cross-stitch semi-supervised
neural attention model for COVID-19 fake news detection. Appl Soft Comput J 107:107393.
https://doi.org/10.1016/j.asoc.2021.107393
25. Jiang TAO, Li JP (2021) A novel stacking approach for accurate detection of fake news. 9.
https://doi.org/10.1109/ACCESS.2021.3056079
26. Al-yahya M, Al-khalifa H, Al-baity H, Alsaeed D, Essam A (2021) Arabic fake news detection
: comparative study of neural networks and transformer-based approaches
27. Pratiwi IYR, Asmara RA, Rahutomo F (2018) Study of hoax news detection using naïve bayes
classifier in Indonesian language. In: Proceedings 11th international conference information
communications technology systems ICTS 2017, 2018-January, pp 73–78. https://doi.org/10.
1109/ICTS.2017.8265649
28. Mansur TA, Onan A (2019) Satire detection in turkish news articles: a machine learning
approach. Mansur Alp Toçoğlu1 and Aytuğ Onan2(&) 1 Faculty of Technology, Depart-
ment of Software Engineering, Manisa Celal Bayar University, 45400 Manisa, Turkey
mansur.tocoglu@cbu.edu.tr 2 Faculty of Engineering and Architecture, Department of
Computer Engineering
29. Jairo Alves L, Leila W, Paulo Q, Carlos Cardoso E, Luan C (2019) Brazilian presi-
dential elections in the era of misinformation: a machine learning approach to analyse
fake news. Fluminense Federal University, Rio de Janeiro, Brazil jairo.luciano@gmail.com,
{leila_weitzel,carloseac, luanpereiracunha}@id.uff.br 2 Universidade de Évora, Évora 17,
Portugal pq@uevora.pt. https://doi.org/10.1007/978-3-030-33904-3
30. Posadas-Durán JP, Gomez-Adorno H, Sidorov G, Escobar JJM (2019) Detection of fake news
in a new corpus for the Spanish language. J Intell Fuzzy Syst 36:4868–4876. https://doi.org/
10.3233/JIFS-179034
31. Alzanin SM, Azmi AM (2019) Rumor detection in Arabic tweets using semi-supervised and
unsupervised expectation–maximization. Knowledge-Based Syst 185:104945. https://doi.org/
10.1016/j.knosys.2019.104945
32. Vicario MD, Quattrociocchi W, Scala A, Zollo F (2019) Polarization and fake news. ACM
Trans Web 13:1–22. https://doi.org/10.1145/3316809
33. Kim N, Seo D, Jeong CS (2019) FAMOUS: fake news detection model based on unified key
sentence information. In: Proceedings IEEE International conference software engineering
services sciences ICSESS. 2018 Nov, pp 617–620. https://doi.org/10.1109/ICSESS.2018.866
3864
34. Liu Z, Shabani S, Balet NG, Sokhn M (2019) Detection of satiric news on social media:
analysis of the phenomenon with a French dataset. In: Proceedings—international confer-
ence computing communications networks, ICCCN. 2019-July, 1–6. https://doi.org/10.1109/
ICCCN.2019.8847041
576 S. H. Imam et al.

35. Verma T, Renu R, Gaur D (2014) Tokenization and filtering process in RapidMiner. Int J Appl
Inf Syst 7:16–18. https://doi.org/10.5120/ijais14-451139
36. Balakrishnan V, Ethel L-Y (2014) Stemming and lemmatization: a comparison of retrieval
performances. Lect Notes Softw Eng 2:262–267. https://doi.org/10.7763/lnse.2014.v2.134
37. Spärck Jones K (2004) A statistical interpretation of term specificity and its application in
retrieval. J Doc 60:493–502. https://doi.org/10.1108/00220410410560573
38. Zhang W, Yoshida T, Tang X (2011) A comparative study of TF*IDF, LSI and multi-words
for text classification. Expert Syst Appl 38:2758–2765. https://doi.org/10.1016/j.eswa.2010.
08.066
39. Bilgin M, Şentürk İF (2017) Sentiment analysis on Twitter data with semi-supervised Doc2Vec.
In: 2nd international conference computing science engineering UBMK 2017. pp 661–666.
https://doi.org/10.1109/UBMK.2017.8093492
40. Shuai Q, Huang Y, Jin L, Pang L (2018) Sentiment analysis on chinese hotel reviews with
Doc2Vec and classifiers. In: Proceedings 2018 IEEE 3rd advanced information technology
electronics automation control conference IAEAC 2018. pp 1171–1174. https://doi.org/10.
1109/IAEAC.2018.8577581
41. Chen Q, Sokolova M (2018) Word2Vec and Doc2Vec in unsupervised sentiment analysis of
clinical discharge summaries
42. Huang G, Bhayani (2009) Twitter sentiment analysis (final project results). J Inform
43. Saloni G, Priyanka M (2019) Fake news detection using passive-aggressive classifier
44. Buitinck L, Louppe G, Blondel M, Pedregosa F, Mueller A, Grisel O, Niculae V, Prettenhofer
P, Gramfort A, Grobler J, Layton R, Vanderplas J, Joly A, Holt B, Varoquaux G (2013) API
design for machine learning software: experiences from the scikit-learn project. pp 1–15
A Tour Towards the Security Issues
of Mobile Cloud Computing: A Survey

Vaibhav Kalra, Shubhankar Rahi, Neerav pawar, Poonam Tanwar,

and Ms Shweta Sharma

Abstract The rapid growth of mobile devices has seriously challenged mobile cloud
computing. It has become one of the biggest issues in the IT world as growth of these
smart phone devices and tablets has increased web-based malicious activities around
the world. Such as there is a lot of data and security issues like risk of data theft
and violation of privacy rights. Securing data is more hard and critical in mobile
cloud computing. The purpose of this study is to present major security and privacy
challenges in this field which has been of great interest to academics and the research
community.

Keywords Mobile cloud computing · Mobile computing · Cloud computing ·

Research directions

1 Introduction

Lastly, organized processing and wanted applications have encouraged unstable

growth of use system, such as distributed computing, programming as an assistance,
interpersonal organizations, online stores, and so on. As an important application
model in the Internet age, cloud computing has been an important research topic
for scientific and industrial communities since 2007. Generally, cloud computing is
defined as a wide range of malware provided by an online system program. These
bundled programs may have a set of economical servers or PCs, which organize
various computing resources according to a specific management plan and provide
customers with secure, consistent, quick, transparent and simple services such as
data warehouse, computer access and use. According to the topmost technologies
provided by Gartner [1–3], ubiquity and mobility are two important characteristics
of the next-generation network that offers various personalized network services
through multiple methods of communication and network access. The primary tech-
nologies of cloud computing include computing, resources and specific applications

V. Kalra · S. Rahi · N. pawar · P. Tanwar (B) · M. S. Sharma

Manav Rachna International Institute of Research and Studies Faridabad, Faridabad, Haryana,
India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 577
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_42
578 V. Kalra et al.

such as things that can be sold, such as water, gas or electricity to users. There-
fore, the combination of the ubiquitous network and cloud computing creates a
new computing model, namely mobile cloud computing. Although mobile cloud
computing contributes a lot to our daily life, it will also bring many problems. In
short, at the heart of these challenges and difficulties is the question of how to combine
the two technologies to make them smaller: on the one hand, to ensure that mobile
devices take full advantage of the benefits of cloud computing to improve and expand
their operations [4].

2 Literature View

In this paper, we have given extensive data in regards to the cloud security issues. The
cloud board secuity issues and cloud access technique security issues are likewise
featured. We additionally proposed the portable distributed computing application
and the difficulties which have existing arrangements and ways to deal with the beat
these difficulties. This paper gives us the key open exploration issues related with
the portable utilization of the distributed computing which assists us with for the
most part perusing and have an outline of the MCC, their issues, existing arrange-
ments, and the distinctive approaches. The creator inform the point that branch of
software engineering proposes and carried out that at whatever point another client
verification component of versatile distributed computing utilizing the unique finger
impression acknowledgment framework in which a global diary of computer science
and designing innovation present us the idea of MCC and told the new wording
Mobile Cloud and other different MCC applications, significant concern and secu-
rity worries with a portion of the anticipations. We discussed the engineering of the
MCC with these various administrations required by their customer and the worker in
MCC. We additionally discussed the major distinctive security issues which emerges
about how much safe the portable distributed computing climate is left [3, 5–21].

3 Security Issue

Security and protection are significant issues in versatile distributed computing appli-
cations and still face some tremendous difficulties. Client protection and the honesty
of information or applications are one of the fundamental issues in accomplishing
versatile distributed computing. It is a mix of distributed computing and portable
organizations. For this, security issues are isolated in to two classes: cloud secu-
rity and portable organization client security. A mobile network user’s security and
protection are significant issues in portable distributed computing applications and
still face some colossal difficulties. Client protection and the honesty of informa-
tion or applications are one of the fundamental issues in accomplishing versatile
A Tour Toward the Security Issues of Mobile … 579

distributed computing. It is a mix of distributed computing and portable organi-

zations. For this, security issues are isolated into two classes: cloud security and
portable organization clients security [1–25].

3.1 Security for Mobile Applications

The requesting way to deal with identifying security issues is low to present and
executing security programming and antivirus programs on portable devices. All
versatile contraptions are fixed with upkeep and force limits. These dangers can
be additionally undermined with PCs that are viable with ensuring the contraption.
A few techniques are intended to look for security apparatuses in the cloud. Prior
to utilizing a particular portable application on versatile, it might encounter some
degree of hazard evaluation. To start with, on the off chance that it is checked and
not noxious, the record will be shipped off the customer’s portable contraptions.
Versatile contraptions are a gentle exercise, for instance, running on a cloud security
worker rather than plague programming or hazard ID [3, 5–11].

4 Security

It creates situations for protection issues by determining the current area and giving
private data, including significant data from the client the utilization of area-based
administrations (LBS) given by global positioning system (GPS) gadgets. Different
dangers to revealing individual information can be limited by choosing and exam-
ining the requirements of the endeavor, and just the predetermined administrations
ought to be acquired and carried out locally through enemy of infection programming
or danger location programs. Programs should move to the cloud networks.

4.1 Getting Information on the Cloud

Individuals and associations can impact the cloud to store a lot of information or appli-
cations. In any case, care should be taken to guarantee the respectability, confirmation,
and computerized privileges of the information or application during handling.

5 Integrity

The whole portable cloud client should guarantee the respectability of their data put
away on the cloud organization. All entrance should be validated and confirmed.
580 V. Kalra et al.

It is proposed to give various ways to deal with keep up the respectability of the
information put away in the cloud. For instance, all the data put away by every
individual or association in the cloud network is labeled or empowered with them,
in which they just move, refresh, or erase information.

6 Authentication

Unlike normal electronic applications, distributed computing applications offer

numerous extraordinary types of assistance to clients. Subsequently, the validation
cycle in distributed computing is totally different contrasted with an ordinary web
application. Getting to various cloud-based applications utilizing a username and
secret word is a solitary symbol. For instance, all the data put away by every indi-
vidual or association in the cloud network is labeled or empowered with them, in
which they just move, refresh, or erase information.

7 Digital Rights Management

Digital rights the board (DRM) is a deliberate way to deal with copyright insurance for
advanced media. DRM is typically carried out by inserting code that forestalls dupli-
cating, indicating the time span for which substance can be gotten to, or restricting
the quantity of gadgets that can be introduced on the media.
Mobile Cloud Computer: -
Mobile computing has become a popular word and is increasingly popular in today’s
computer field. Advances in mobile computers, network connectivity, and security
technologies such as mobile phones, smart phones, PDAs, GPS navigation, and
laptops have shown impressive growth. Alternatively, with wireless technologies
such as WiMAX, connectivity networks, and WiFi, users can display the Internet
successfully but are not limited to wires as before. Therefore, those mobile devices
are accepted as their first choice for work and have fun in their daily lives.

8 Characteristic

Characteristics are as follows: -

A Tour Toward the Security Issues of Mobile … 581

8.1 Travel

Mobile nodes on mobile computer networks that communicate with others, including
random nodes in the cable network via mobile help (MS) channels while in operation.

8.2 Different Types of Network Conditions

The network functions commonly used for mobile nodes are not specified; such
networks may be high-bandwidth wired network networks or low-bandwidth
broadband (WWAN) networks, or they may be phased out.

8.3 Frequent Termination

Mobile nodes do not always maintain connections, such as wireless communication,

charging network status, but wireless network remains idle or disconnected due to
low battery power.

8.4 Disk-Symmetric Network Connection

Servers and access points and other MSS provide powerful access, but such capa-
bilities weaken call centers. Therefore, there are connection bandwidths with high
variability between downlink and uplink.

8.5 Low Dependability

Network system terminals that signify mobile phones, networks, database platforms,
and security application development improvements should be considered as signs
of disruption and recurrence may occur.

9 Characteristics

Cloud square measurement options are as follows:

582 V. Kalra et al.

9.1 Performance

‘Cloud’ is often thought of as a resource pool wherever all underlying platforms

are used. Complete users access the desired resources through the browser and get
information from cloud computing providers while not maintaining their informa-
tion centers. In addition, some virtual machines (VMs) square measure are usually
installed on the server too much to increase the utility capacity; and VMs support
upload migration in the event of server overload.

9.2 Reliability and Extension

Cloud computing provides a secure mode for storing user information and users are
not concerned about issues such as software system changes, leak repairs, virus
attacks, and loss of information. If a failure occurs on a server or VM, cloud
computing systems transfer and store that information to different machines and
then remove those nodes that fail on the system systems to create the whole system
operating normally. Meanwhile, the cloud is usually expanded from horizontal and
overhead to a very large network, making a variety of applications from thousands
of nodes and strangers.

9.3 Independence

The associate in nursing cloud system is self-contained, repairs itself and assigns
hardware, software and storage resources to the most demanding consumers, so
administrators are clear to eliminate users.
(3) Provocation:
First, cloud computing requires the associate in nursing mechanism to be devel-
oped to produce a secure and high-powered service because of the variety of
complaints used and the infrastructure used in computer use. In addition, due to
the resource-intensive resource information centers, a resource management plan
and measurement strategies are required to avoid wasting energy. Lastly, a small,
straightforward, and easy-to-use square integration is very important for service
providers in the cloud computing, so the exact same standard as normal | equal
level} required dry.
A Tour Toward the Security Issues of Mobile … 583

10 Architecture of Mobile Cloud Computing

Versatile distributed computing, in its easiest way, alludes to the foundation that
happens outside the cell phone, both information stockpiling and information
preparing. Portable cloud applications remove processing force and information
stockpiling from cell phones and the cloud, bringing applications and MCs not exclu-
sively to cell phone clients yet to a more extensive scope of versatile supporters. Cell
phones assume a significant part in our day-by-day lives as it has gotten very accom-
modating in a compelling specialized instrument paying little heed to time and place.
The size of cell phones is restricted to a base size; there are limits on handling abilities,
battery life, information stockpiling, and show limit. Asset concentrated applications
are huge information that require maximum usage of central processing unit (CPU),
random access memory (RAM) to run these applications. These require a more drawn
out battery life, which is absent in current cell phones, and incredible applications
with high information use, for example, games and discourse acknowledgment are
genuine instances of this. To beat these impediments, portable distributed computing
has been presented with versatile administrations, with the advancement of (mobile
cloud computing (MCC) and the expanding number of cell phone applications in
cloud moving. This lessens the specialized equipment and processing force of cell
phones. A very much organized construction proposes an absence of coordination
among IT and business objectives. Many MCC structures have been created from
various stages, and at present, four kinds of designs are being utilized for the current
MCC. This is the overall construction of the MCC [25] (Fig. 1).

Fig. 1 Mobile cloud computing (MCC) architecture of [25]

584 V. Kalra et al.

This structure consists of mobile devices connected to the mobile network via
basic channels (base transceiver channels, access points, or satellites) that connect and
control the active interaction between the network and the mobile device. Medium
user data requests and information are transmitted via a central processor from a
server connected to mobile network services. Mobile users receive authentication,
authentication, and accounting (AAA) services from mobile network providers based
on user information on the home agent and data server. User requests are sent to the
cloud via the Internet. These cloud applications are managed by cloud administrators
and provide services at their request. These services are available through utility
computing, virtualization and service-oriented architecture. The current structures
below apply to all cloud-based development models.
Cloud services are usually categorized based on layer configuration. Within the
higher levels of this paradigm, infrastructure as a service (IaaS), platform as a service
(PaaS), and software as a service (SaaS) are set [1–3, 5, 12–17].
Data center layer
This layer provides hardware comfort and cloud infrastructure. In the data center
layer, many servers connected to networks are very fast to provide services to
customers. Typically, data centers with high power supply and low risk of disaster
are built in densely populated areas. Data center layer (DCL) plays an important role
in the data center because it connects all data center resources.
Infrastructure as a service (IaaS).
IaaS as a service is a distributed computing administration where organizations
lease or rent workers for registering and capacity in the cloud. Clients can run any
working framework or application on a leased worker with no support and upkeep
expenses of those workers. Infrastructure as a service is a provisioning model in
which an association rethinks gear used to help tasks, including capacity, equipment,
workers, and systems administration parts. The IaaS server farm is situated at the
highest point of the layer. It gives stockpiling, workers, organizing parts and equip-
ment on a ‘installment premise’ that you use. IaaS rapidly increases and down with
request, paying you just for your utilization. This will assist you with keeping away
from the expense and intricacy of buying and keeping up your own actual worker and
other datacenter foundation. Every asset is given as a different help unit and you just
need to employ one as long as you need it. Distributed computing specialist co-ops,
for example, Azure deals with the framework when buying, introducing, designing
and keeping up your own product working framework, middleware and applications.
Instances of Amazon EC2 and S3 IaaS. Infrastructure as a service gives essential
stockpiling and figuring abilities as standard administrations over an organization
[4, 18–25] (Fig. 2).
Software as a service (SaaS).
Software as a service (or SaaS): How programs are distributed on the Internet—
as a service. SaaS applications are sometimes called web-based software, wanted
software, or hosted software. No matter the name, SAS applications run on a SAS
provider server.
A Tour Toward the Security Issues of Mobile … 585

SaaS Gmail,salesforce CRM,EventPro,Office365

• For End user

PaaS Heroku,AWS Elastic Beanstalk.

• For software developer

IaaS Microsoft Azure,Google Compute Engine

• For IT administrators

Fig. 2 Cloud service model [25]

The most well-known example of Saleforce is Salesforce.com, but many other

examples have hit the market, including providing Google apps for basic business
services, including email and word processing (Google Microsystems, 2009). Sale-
force is a leader in promoting this service model. Microsoft Mesh Live lets you share
files and folders on multiple devices at the same time.
Platform as a Service (PaaS)
A platform as a service (PaaS) cloud is full of development and distribution
environments, with resources to help you deliver everything from simple cloud-
based applications to advanced, cloud-enabled business applications. Like IaaS, PaaS
includes infrastructure—servers, storage and communications—but at the same time
middleware, development tools, enterprise intelligence services (BI), data manage-
ment systems and more. PaaS is designed to support the entire web application life
process: build, test, deployment, storage and development. PaaS allows you to avoid
the costs and difficulties of purchasing and maintaining software licenses, built-in
app infrastructure and middleware etc.
The platform integrates software (PaaS) layer and provides it as a service that can
be used to build high-quality services. The client interacts with the platform through
the API and does what it takes to maintain and measure the platform to provide a
certain level of service (Sun Microsystems, 2009). Examples of PA are Google App
Engine, Microsoft Azure and Amazon Map Reduction / Simple Service.
In this way, on a portable cloud computing, data storage and integration is
embedded in the cloud, and the user gets seamless, desired service without worrying
about battery life or the ability to process mobile devices.

11 Idea and System

Similar to cloud computing, there are local unit loads but no consistent descrip-
tions of what mobile cloud computing is. With this paper, we tend to think of a
586 V. Kalra et al.

completely different computer mode consisting of MC and CC that provides cloud-

based client services using web and mobile devices. Whereas other side MCC may
be the advancement of a portable computer.
MCC is simply divided into cloud computing and mobile computing. Those
mobile devices are usually laptops, PDAs, smart phones and so on. It connects to
a hotspot or base station via 3G, WiFi, or GPRS. Because the computer and major
processors are migrated to the ‘cloud,’ the potential demand for mobile devices is
limited, and some low-cost mobile phones or perhaps non-smartphones can even
deliver mobile goods through a central platform. Al- even if the consumer uses a
cloud that is converted from PCs or installed devices to mobile devices, the main idea
remains the computer of the cloud. Mobile users send cloud service requests via an
Internet browser or desktop application, then cloud management assigns resources
to the application to ensure integration, and mobile computing and visualization
activities can be forced to ensure QoS integration is complete [1–24].

12 Challenges with Solutions

The main purpose of using cloud computing is to create a simple and fast way for
users to access and retrieve information from the cloud, which is a simpler and
faster way to access cloud computing services effectively through the exploitation of
contextual devices. The biggest challenge of mobile cloud computing comes from
the features of mobile devices and wireless networks, in addition to its limitations and
limitations, and such challenges make application, editing and installing on mobile
devices and additional distribution more difficult than cloud-mounted devices. In
the case of mobile cloud computing, mobile device issues, wireless communication
quality, application forms, and support from a computer cloud to mobile square
measure all important factors that contribute to testing from a computer. The second
table provides a summary of the planned challenges and a few solutions related to
computer cloud computing [1–24].

12.1 Ratings for Mobile Devices

Talking about cloud-based devices is a key factor in resource pressure. Although

smart phones are clearly developed in various fields such as hardware and memory,
storage, screen size, wireless communication, audio technology, and applications,
there are still major limitations such as restrictions on computer capabilities and
power resources, using complex systems. In this regard, PCs and laptops are in a
state of overcrowding, and these smartphones such as the iPhone 4S, humanoid
serials, Windows Mobile serials are three times less processed, eight times memory,
five to ten times in the final capacity and ten times the network data rate. Normally,
a smartphone should be charged daily such as making calls, triggering messages,
A Tour Toward the Security Issues of Mobile … 587

aquatic web, public access, and various network applications. In line with previous
development trends, computer duplication capabilities and rapid development of
screen technology can create more and more applications installed on smartphones.
If battery technology cannot be improved in a short period of time, then how to save
battery power on a smartphone can be a major problem we often encounter these
days.
The process efficiency, storage, battery time, and communication of these smart-
phones are systematically improved in the event of a portable computer. However,
such large differences could continue at a concert of major challenges in mobile
cloud computing.

12.2 Communication Quality

Separated from the cable operator using physical integration to ensure data integrity,
the data transmission in mobile cloud sites remains dynamic and therefore the orga-
nization continues due to the existing network overlay permissions. In addition, the
information center for a large business and web service provider service is often a
way to complete users, especially for mobile device users. In a wireless network,
the network delay is two hundred ms in the ‘last mile’ but only five ms in the old
wireless network.

13 Open Analysis Issues

Although some come with a cloud-based computer already shipped around the world,
there is still a long way to go for business use, and a few aspects of analysis should
be considered in any project.

13.1 Delivery of Information

Due to the nature of the resources, mobile devices have potential challenges in cloud
access, static access, data transfer, etc. These type of issues may be resolved by
the application (service) and a central location that can provide a policy for all
cloud-based computer programs [25].
588 V. Kalra et al.

13.2 Classification of Tasks

Investigators differentiate tasks/applications from mobile devices to multiple sub-

tasks and bring their number to work in the cloud, which can be a reasonable decision
for mobile devices that have been denied resources. Though the accurate strategy or
rules on how to differentiate these functions are not defined and to be processed by
the cloud and those by device.

13.3 Higher Service

The actual purpose of MC is to offer PC-friendly platform for mobile terminals. On

the other hand, because of completely different options available between mobile
devices and PCs and user are often unable to directly integrate services from PC
platforms to mobile devices. Therefore, any analysis should attempt to develop a
strategy in such a way as to provide appropriate and friendly communication services
for mobile devices.

References

1. Hewitt C (2008) Orgs for scalable, robust, privacy-friendly client cloud computing. IEEE
Internet Comput 12(5):96–99
2. Buyya R, Yeo C, Venugopal S (2008) Market-oriented cloud computing: vision , hype, and
reality for delivering it services as computing utilities. In: High performance computing and
communications 2008, HPCC’08. 10th IEEE international conference on IEEE, pp 5–13
3. Youseff L, Butrico M, Da Silva D (2008) Toward a unified ontology of cloud computing. In:
Grid computing environments workshop, GCE’08. IEEE, pp 1–10
4. Mobile cloud computing subscribers to total nearly one billion by 2014, 2009, Sept). [Online].
Available http://www.abiresearch.com/press/1484
5. Shankar S (2009) Amazon elastic compute cloud
6. Zahariev A (2009) Google app engine. Helsinki University of Technology
7. Microsoft azure homepage (2011) [Online] Available http://www.windowsazure.com/en-us/
8. McCarthy J (1961) Speech givento celebrate mits centennial [Online] Available http://en.wik
ipedia.org/wiki/JohnMcCarthy (computerscientist)
9. The customer relationship management (crm) (2009) [Online]. Avail- able: http://en.wikipe
dia.org/wiki/Customerrelationshipmanagement
10. Rochwerger B et al (2009) There servoir model and architecture for open federated cloud
computing. IBM J Res Develop 53(4):1–11
11. Boss G et al (2007) Cloud computing. IBM white paper, Version vol 1.
12. Mei L, Chan W, Tse T (2008) A tale of clouds: paradigm comparisons and some thoughts on
research issues. In: Asia-pacific services com- puting conference, 2008. APSCC’08, IEEE, pp
464–469
13. Cohen R (2010) The cloud computing opportunity by the numbers. 2010 [Online] Available
http://www.elasticvapor.com/2010/05/cloud-computing-opportunity-by-numbers.html
14. Marrapese B (2010) Google ceo: a few years later, the mobile phone becomes a super computer
Dec [Online] Available http://www.itnews-blog.com/it/21320.html
A Tour Toward the Security Issues of Mobile … 589

15. Chetanet S et al (2010) Cloud computing for mobile world. Available at chetan. ueuo.com
16. Chunet et al (2011) Elastic execution between mobile device and cloud. Published in
Proceedings of the sixth conference on computer systems. ACM, pp 301–314
17. Zhang X et al (2011) Towards an elastic application model for augmenting the computing
capabilities of mobile devices with cloud computing. Mobile Netw Appl 16(3):270–284
18. Li L, Shen (2011) Virtualized screen: a third element for cloud-mobile convergence. Multimedia
IEEE 18(2):4–11
19. Giurgiu et al (2009) Calling the cloud: enabling mobile phones as interfaces to cloud applica-
tions. In: Proceedings of the ACM/IFIP/USENIX 10th international conference on middleware.
Springer, pp 83–102
20. Rellermeyer A, Roscoe (2011) R-osgi: distributed applications through software modulariza-
tion. In: IFIP lecture notes in computer science(LNCS) vol 4834(4834). pp 1–20
21. Jeong S et al (2010) Towards an elastic application model for augmenting computing capabili-
ties of mobile platforms. In: Mobile wireless middleware, operating systems and applications,
pp 161–174
22. Satyanarayanan M, Bahl P, Caceres R, Davies N (2009) The case for vm-based cloudlets in
mobile computing. Pervasive Comput IEEE 8(4):14–23
23. Marinelli E (2009) Hyrax: cloud computing on mobile devices using mapre-duce. DTIC
Document, Tech. Rep
24. Cooney M (2011) Gartner: The top 10 strategic technology trends for 2012. [Online]. Available
http://www.networkworld.com/news/2011/101811-gartner-technology-trends-252100.html
25. Hoang et al (2011) A survey of mobile cloud computing: architecture, applications, and
approaches. Wireless Mobile Commun. https://doi.org/10.1002/wcm.1203
Subpixel Image Registration Using
FFT-Based Technique

Swati Jindal, Bhaskar Dubey, and Jitali Patel

Abstract This paper presents an efficient image matching technique with trans-
lation, subpixel translation. Fast Fourier transform technique is the most powerful
area-based technique that involves translation, rotation and other operation in fre-
quency domain. FFT and its inverse are used to convert a two-dimensional image
from the spatial domain to the frequency domain. The benefit of expressing an image
in frequency domain is that executing certain operations on the frequencies is far more
efficient than doing so in spatial domain. In this paper, we discussed different meth-
ods of subpixel estimation. Also, we validate the performance of our method under
several different noise conditions. Our main goal to develop efficient technique to
estimate translation and subpixel both by using FFT techniques. Our implementation
is based on Python programming.

Keywords Image registration · Phase correlation · FFT

1 Introduction

Image registration is a crucial step in most of the image processing task such as remote
sensing, weather forecasting and satellite image. Image registration is used to overlay
two or more images with the same scene taken from different sensors having different
orientation, the same sensor at different time and from different locations. The main
goal of image registration is to find coordinate relationship between images.

S. Jindal (B) · J. Patel

Institute of Technology, Nirma University, Ahmedabad, India
e-mail: 19mcei09@nirmauni.ac.in
J. Patel
e-mail: jitali.patel@nirmauni.ac.in
B. Dubey
Image Analysis and Quality Evaluation Division, Signal and Image Processing Group, Space
Applications Centre, Ahmedabad, India
e-mail: bhaskard@sac.isro.gov.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 591
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_43
592 S. Jindal et al.

Image registration is divided into two categories: feature-based techniques and

area-based techniques.
1. Feature-based techniques: This technique works on the object or feature point of
an image. It can be further divided into algorithm such as an algorithm that uses high-
level and low-level features, e.g. graph-theoretic method and feature-based method,
respectively [1–3]. The low-level feature means edges and corners, while high-level
feature means detect object or relation between features.
2. Area-based techniques: This technique works on intensity of images. It can
be further divided into algorithm which uses frequency domain, e.g. FFT-based &
algorithm which uses direct pixel value, e.g. cross-correlation [1–3].
Mathematical model: Let f and g be two images such that f is translated, rotated
and scaled replica of image ‘g’. That is,

1
f (x, y) = g( (x cos θ + y sin θ − x0 , −x cos θ + y cos θ − y0 )) (1)
|a|

where x0 and y0 are the translation in the x and y directions, respectively. θ and a
are the rotation angle and scale factor, respectively. Thus, the aim of any automatic
image registration model is to estimate the parameters (x0 , y0 ,θ , a).
The fast Fourier transform(FFT) is the most powerful method of area-based tech-
nique. The registration methods use the Fourier domain method to match the images.
Geometric defects such as translation, rotation and scaling between images are
corrected using FFT-based registration algorithms. This is accomplished by applying
techniques from theory of Fourier transform to image data sets.
The FFT ratio is calculated using phase correlation, which is based on the trans-
lation property of the Fourier transform. The strongest correlation is represented by
the sharp peak of the inversion of this ratio. Before phase correlation, the reference
and sensed images are converted to log-polar images from Cartesian coordinates, i.e.
the rotation in Cartesian coordinate is equivalent to translation in polar coordinates.
After that, phase correlation computes similar ratio, scaling & rotation error which
are represented as shift.

2 Literature Review

The image registration sector is highly capital-intensive sector and has attracted atten-
tion of researchers. A researcher from the world used different techniques of image
registration to match images. Study examining efficiency of different algorithms
includes work of Reddy et al.(1996), Samritjiarapon et al.(2008), Xie et al.(2003),
Tong et al.(2019), Foroosh et al.(2002) and Ezzeldeen et al.(2010).
Ezzeldeen et al. [4] analysed different techniques, i.e. FFT-based techniques,
contour-based, wavelet-based, Harris-PCNN and Harris moment-based technique.
These techniques have been evaluated based on root mean square error(RMSE) and
running time. Landsat Thematic Mapper and SPOT remote sensed image are used.
Subpixel Image Registration Using FFT-Based Technique 593

The sensed images are rotated with 1–20 angle in counterclockwise direction and,
furthermore, compared referenced image and rotated images. In the end, it is con-
cluded that most stable and best process is FFT, but FFT has higher RMSE value, and
the RMSE value of wavelet technique is least, and smaller running time technique is
contour based and FFT technique.
Foroosh et al. [5] discuss the subpixel image registration method with interpola-
tion such as phase correlation, intensity interpolation and correlation interpolation.
The accuracy of methods depends on the quality of the interpolation algorithm,
also discusses subpixel image registration without interpolation such as differential
properties of an image sequence and defines subpixel registration as an optimization
problem. So the main motive is to find subpixel image registration using phase corre-
lation. To estimate non-integer translation between images, author describes various
methods such as 2D Dirichlet kernel function. The algorithm is tested on different
set of images.
Samritjiarapon et al. [1] explore the Fourier phase correlation which is used to
measure the translation, and best first search is used to analyse the final translation
of angle or scaling parameters. Satellite image is used to implement data. The best
technique is FFT-based technique with phase correlation and best first search as
its error is less as compared to other techniques. Moreover, the computation time
of FFT-based technique with the phase correlation technique is less. So FFT-based
technique with phase correlation and the best first search techniques are observed as
more stable.
Reddy et al. [6] evaluated the efficiency of FFT-based techniques. Phase correla-
tion is used to get translation movement. To find scaling and rotation movement, use
Fourier scaling and rotation properties. The proposed algorithm is tested on set of
different images and different amount of noise added to the image. The advantage of
the algorithm is that it computes result in fixed amount of time and low computational
cost.
Xie et al. [7] conducted a study on FFT-based technique and implemented using
Interactive Data Language (IDL) and added ENvironment for Visualizing Images
(ENVI). As the image size increases, overall accuracy increases, and the accuracy
of the algorithm is good compared to manual method. The ENVI user functions are
very useful because function is inbuilt and it also reduces time.
Patel et al. [8] explore the challenge of estimating the rotation, and scaling is
reduced to one estimating a 2D translation by resampling the Fourier magnitudes on
the log-polar grid. Rather than standard normal correlation, the authors use phase
correlation to convert from Cartesian to log-polar using bilinear interpolation algo-
rithms. As a result, when it comes to registering aerial photographs, this method is
extremely effective.
According to Juan et al. [9], SIFT is the slowest and least effective at lighting
changes. SURF is quick, but it is not resistant to rotations or variations in lighting.
PCA- SIFT’s blur quality should increase. The approaches are chosen in accordance
with the application.
594 S. Jindal et al.

From depth of published paper, we learned several methods for image registra-
tion such as contour based, wavelet based, FFT and many more, but the FFT-based
technique is very efficient and accurate compared to other methods.

3 Methodology

3.1 FFT-Based Image Registration Approach

FFT-based image registration problems use the properties of Fourier transforms of

the two images (often called master and slave, or original and replica) to estimate the
translation in X and Y directions. The technique is highly efficient and robust and
gives accurate results under the presence of high noise as well. The Fourier transform
method basically is used to convert the image from spatial domain into frequency
domain, it decomposes result into sine and cosine waves [10], and each point in the
Fourier domain image is a particular frequency. Phase correlation method can be
explained as follows: let f 1 be sensed image and f 2 be translated replica of f 1 , so
they differ by (x0 , y0 ) displacement [6]:

f 2 (x, y) = f 1 (x − x0 , y − y0 ) (2)

their corresponding Fourier transform is [6]:

F2 (x, y) = exp− j2π(ux0 +vy0) ∗F1 (u, v) (3)

where F1 & F2 are Fourier transform of f 1 and f 2 . Now phase correlation of two
image f 1 and f 2 with their Fourier transform F1 and F2 is [6, 11]:

F1 (u, v)F2∗ (u, v)

= exp j2π(ux0 +vy0) (4)
|F1 (u, v)F2∗ (u, v)|

As shown in Eq. 4, F2∗ is complex conjugate of F2 . Now inverse Fourier transform

of Eq. 4 results in an impulse at the location (x0 ,y0 ). Thus by locating the position of
impulse, we get the required translation in image pairs.

3.2 Subpixel Image Registration

Our main aim is to find a method for subpixel translation with the benefits of the phase
correlation method. In subpixel image, registration images are upsampled by scaling
factor using bicubic interpolation. There are various interpolation techniques such
as nearest neighbour, bilinear and bicubic interpolation. Bicubic interpolation is the
Subpixel Image Registration Using FFT-Based Technique 595

best technique as compared to nearest neighbour and bilinear interpolation. Bicubic

interpolation technique is generally used in professional image editing software like
adobe photoshop and corel photoshop [2]. In this, we consider closest 4 × 4 neighbour
of points for a total of 16 points or pixels. These 16 points are determined using the
below equation [2] :
 3  3
p(x, y) = ai j x i y j (5)
i=0 j=0

Hence, images are upsampled by bicubic interpolation. The scaling factor calculates
the computation time of the interpolation technique.
Now upsampled images are shifted by an integer value, and upsampled shifted
images are downsampled as the same scaling factor. After that, the result of phase
correlation is observed as single peak value, while in the case of subpixel image
shift, there is coherent peak in which one is largely adjacent to each other. So, further
computing is required for subpixel shift.
Generally, subpixel image registration based on Fourier transform is divided into two
categories:
(i) Spatial domain
(ii) Frequency domain
Spatial domain means by using correlation peak, while frequency domain means
by using phase difference. To estimate subpixel shift, three techniques are used
(i) Centroid shift estimation
(ii) Parabolic fit
(iii) Dirichlet function
Centroid Shift Estimation In this, we simply calculate or estimate weighted aver-
age of local neighbourhood around phase correlation value (i.e. peak value)[12].
Weighted average is calculated using below equation:

x0 pc[0] + x1 pc[1] + · · · + xn pc[n]

x= n (6)
i=0 pc[i]

y0 pc[0] + y1 pc[1] + · · · + yn pc[n]

y= n (7)
i=0 pc[i]

where pc[0] pc[1] and …, pc[n] are phase correlation value at point and x0 , x1 , . . . ,
xn , y0 , y1 , . . . andyn are nearest neighbour of peak point. Here x and y are estimated
values of subpixel displacement image registration.
Parabolic Fit In this, we used parabolic function [3, 11, 13], i.e.

PC(x, y) = a0 x 2 + a1 y 2 + a2 x y + a3 x + a4 y + a5 (8)

First estimate peak value and then estimate subpixel value using six coefficients
(a0 , a1 , a2 , a3 , a4 , a5 ) and nine neighbourhood values.
596 S. Jindal et al.

AX = B (9)

where A is a rectangular matrix of size 9 × 6, X is coefficient matrix which contains

unknown values and B is phase correlation matrix which contains phase correlation
value as shown in below equation [11, 13]:.
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
x02 y02 x0 y0 x0 y0 1 a0 PC0
⎢x 2 y12 x1 y1 x1 y1 1⎥ ⎢a 1 ⎥ ⎢ PC1 ⎥
⎢ 1 ⎥ ⎢ ⎥ ⎢ ⎥
⎢. . . . . .⎥⎥ ⎢ ⎥ ⎢ PC2 ⎥
⎢ ⎢a2 ⎥ ⎢ ⎥
⎢. . . . . .⎥ ⎢a 3 ⎥ ⎢ PC3 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
A=⎢
⎢. . . . . .⎥ ⎢ ⎥ ⎢ ⎥
⎥ , X = ⎢a4 ⎥ and B = ⎢ PC4 ⎥ (10)
⎢. . . . . .⎥⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎢a 5 ⎥ ⎢ PC5 ⎥
⎢. . . . . .⎥⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎢a 6 ⎥ ⎢ PC6 ⎥
⎣. . . . . . ⎦ ⎣ a7 ⎦ ⎣ PC7 ⎦
x82 y82 x8 y8 x8 y8 1 a8 PC8

Here PC0 , PC1 , …and PC8 are phase correlation values, and we have to determine
coefficient matrix, i.e. X.
X = A−1 B (11)

Now, peak value at x and y can be determined using below equation [3, 11–13]:

a2 a4 − 2a1 a3 a2 a3 − 2a0 a4
Δx = , Δy = (12)
4a0 a1 − a22
4a0 a1 − a22

Hence, estimated values of x and y are determined as given in Table 4.

3.3 Dirichlet Function

Once we get peak value at point (x0 , y0 ) using phase correlation. In subpixel image,
displacement coherent peak occured which is one adjacent to each other, i.e. (x h , y0 )
& (x0 , yh ), where x h = x0 ±1, yh = y0 ±1.
Explain Dirichlet function by taking example: Consider peak which occurs at
location (x0 , y0 ) and two side peaks at (x0 +1, y0 ) and (x0 , y0 +1). By applying Eq.
(4), we will find phase correlation pc(0,0), pc(1,0) and pc(0,1). After applying below
equation [5], we obtained value of (Δx, Δy).

pc(1, 0)
Δx = (13)
pc(1, 0) ± pc(0, 0)

pc(0, 1)
Δy = (14)
pc(0, 1) ± pc(0, 0)
Subpixel Image Registration Using FFT-Based Technique 597

Here, we will obtain two solutions. So correct solution is to identify that the result
is in an interval of (0,1) and the same sign as x h - x0 and the same for y-axis.

4 Implementation and Results

4.1 Algorithm

The main goal is to develop an efficient technique using Python that can estimate
translation and rotation both. We first do our simulation experiment on Lena image;
subsequently, the technique can be applied to any image in general. The implemen-
tation steps are shown in Fig. 1.
Repeat the steps under different noise conditions and carry out a sensitivity anal-
ysis with respect to noise.
Experiments on Bidirectional Translated Lena Image The matching of proposed
algorithm is tested on a different set of images. Here Lena image is taken as test case.
All the images are 8-bit greyscale image of size 512 × 512. In this, we give integer

Original image Reference image

f(x,y) g(x,y)

FFT Module FFT Module

F(u,v) G(u,v)

Phase Correlaon
Module

Esmate Translaon

Fig. 1 Block diagram of algorithm

598 S. Jindal et al.

(a) (b) (c) (d) (e)

Fig. 2 aOriginal Lena image, b shifted Lena image by (−20, −20), c shifted Lena image by (−20,
10), d shifted Lena image by (−10, −10), e shifted Lena image by (20, 20)

Table 1 Simulated and translated Lena images

Fig X-Shift Y-Shift
a 0 0
b −20 −20
c −20 10
d −10 20
e 20 20

(a) (b) (c) (d)

Fig. 3 a Surface plot of phase correlation, b surface plot of phase correlation, c surface plot of
phase correlation, d surface plot of phase correlation

shift in both directions using Eq. 2. Now, different shifts are demonstrated as shown
in Fig. 2 (Table 1).
Translation Estimation using Phase Correlation In this, we estimate the translation
of the image using phase correlation. Here phase correlation means basically compute
their FFT ratio. And phase correlation is based on Fourier shift property. Due to
circular wrapping, −x is equivalent to 512 − x (Fig. 3; Table 2).

4.2 Generation of Subpixel Translated Lena Image

In this section, we are explaining the process of subpixel translation. Lena image was
used to demonstrate the result. For example, Lena image is amplified by scaling factor
Subpixel Image Registration Using FFT-Based Technique 599

Table 2 Estimated translation using phase correlation

Fig Estimated X-Shift Estimated Y-Shift
(a) 492(−20) 492(−20)
(b) 492(−20) 10
(c) 502(−10) 20
(d) 20 20

(a) (b) (c) (d)

Fig. 4 a Zoomed image by factor 2, b integer shifted Lena image by (201, 201), c downsampled
shifted Lena image by (100.5, 100.5), d surface plot of phase correlation of Lena image & (c)

2 using bicubic interpolation, and then image is translated by integer value(201,201).

Some portion of amplified Lena image and the translated image are shown in Fig. 4a,
b, respectively. Then translated zoomed Lena image is subsequently downsampled
by the same factor, as shown in Fig. 4c.
Subpixel Image Registration Results The algorithm has been tested on 48 test cases
of Lena image. The algorithm was applied to a different shift, which has been shown
in Fig. 5. The surface plot of phase correlation of Lena image, which is translated
in both axes, is shown in Fig. 5b, e, h, k. In subpixel shift, peak is not located at
single point but located at various points which are adjacent to each other. So, further
processing of subpixel image is required.
The simulation results of different algorithm are shown in Tables 3, 4 and 5 by
examining error between the estimate and original shift. By comparing the method,
it is found that the Dirichlet function is more efficient because the error is less as
compared to other methods, and error is in an interval of (±0.1, ±0.25).
Subpixel Translation Under Noise Condition We tested our algorithm under dif-
ferent noise conditions. To obtain noisy images, Gaussian noise has been added to the
normal image. Here 20db noise is added to the images. Due to the subpixel nature,
there is the peak which is the one adjacent to each other. To estimate subpixel shift,
Dirichlet function is used, and result shows the exact shift is recovered even under
noise conditions as shown in Fig. 6.
600 S. Jindal et al.

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

(j) (k) (l)

Fig. 5 a Shifted Lena image by (100.3, 100.3), b surface plot of phase correlation, c zoomed image
of phase correlation, d shifted Lena image by (−100.5, 80.5), e surface plot of phase correlation,
f zoomed image of phase correlation, g shifted Lena image by (−50.3, −40.3), (h) surface plot of
phase correlation, i zoomed image of phase correlation, j shifted Lena image by (100.25, 100.25),
k surface plot of phase correlation, l zoomed image of phase correlation
Subpixel Image Registration Using FFT-Based Technique 601

Table 3 Estimated subpixel displacement value using centroid function

S. No. Original (x, y) shift Estimate (x, y) shift Error
1 (0,0) (0,00 (0,0)
2 (100.5,100.5) (100.23, 100.51) (0.27, −0.01)
3 (100.33,100.33) (100.18, 100.06) (0.15,0.27)
4 (411.5,80.5) (411.48, 80.59) (0.02, −0.09)
5 (461.67,471.67) (461.81,471.84) (−0.14, −0.17)
6 (100.25,100.25) (100.10,100.02) (0.15,0.23)

Table 4 Estimated subpixel displacement value using parabolic fit

S.No. Original (x, y) shift Estimate (x, y) shift Error
1 (0,0) (0,00 (0,0)
2 (100.5,100.5) (100.39, 99.96) (0.11,0.54)
3 (100.33,100.33) (100.13,100.01) (0.2,0.23)
4 (411.5,80.5) (411.98, 80.32) (−0.48,0.18)
5 (461.67,471.67) (461.91,471.93) (−0.24, −0.26)
6 (100.25,100.25) (100.81,99.36) (−0.56,0.89)

Table 5 Estimated subpixel displacement value using Dirichlet function

S. No. Original (x, y) shift Estimate (x, y) shift Error
1 (0,0) (0,0) (0,0)
2 (100.5,100.5) (100.25,100.56) (.25,−0.06)
3 (100.33,100.33) (100.31,100.29) (−0.01,0.01)
4 (411.5,80.5) (411.40,80.68) (0.1, −0.18)
5 (461.67,471.67) (461.69,471.69) (−0.02, −0.02)
6 (100.25,100.25) (100.19,100.18) (0.06, 0.07)

(a) (b) (c) (d)

Fig. 6 a Noisy image shifted by (50.5, 40.5), b surface plot of phase correlation of shifted noisy
and original Lena image, c noisy image shifted by (−100.33, −80.33), (d)surface plot of phase
correlation of shifted noisy and original Lena image
602 S. Jindal et al.

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

Fig. 7 a Landsat satellite image, b shifted image by (151,151), c surface plot of phase correlation
of (a) & (b), d Atlanta Rareplane image, e shifted image by (−100, −80), f surface plot of phase
correlation of (d) & (e), g London Rareplane image, h shifted image by (150, −50), i surface plot
of phase correlation of (g) & (h)

4.3 Result of Satellite Image

The technique was also used to perform a separate set of satellite images from
different bands of Landsat pictures. We have taken a satellite image from Earth
Explorer, the latitude is 23.02, and the longitude is 72.57 of Ahmedabad space. The
Landsat 8 data have 11 spectral bands, and image is of size 7731 × 7881. In this
experiment, band 3 Landsat 8 image was taken as original image, and shift image
by (151,151) in both axes was taken as sensed image. The FFT technique works
only on image of size 2n . The original landsat satellite image is of size 512 × 512
after that shift image by 151, 151 in both direction. Even under case of satellite
image, algorithm provides accurate result. The result of phase correlation is impulse
function, i.e. everywhere zero except at displacement point. It was discovered that
comparing two bands of the same satellite image yielded precise results.
Subpixel Image Registration Using FFT-Based Technique 603

5 Future Work

Following is the work that needs to be done in future:

– The technique can be implemented on rotated and scaled images as well by trans-
forming to FFT images in the polar coordinates.
– We plan to test the performance of rotation estimation by giving different degree
of angle and under similar noise condition.
– Subsequently, the technique can be tested for any combination of translation,
rotation and both.
– We plan to simulate studies with highly noised image pairs which are rotated,
translated and scaled replica of each other.

6 Conclusion

An efficient FFT-based technique is implemented for automatic image registration.

The accuracy of the FFT-based image registration method is very high, even under
noise condition. We carefully designed simulation experiment that shifted the Lena
image from -2 to 3 in both directions to generate 25 test cases. Moreover, add 10db
PSNR noise to all the images and perform phase correlation. It is shown that phase
correlation estimates the translation in images very accurately in more than 80 per
cent case. So, this method is capable of estimating translation under noise condition.
The advantage is that even if image size increases, its computation time is less and
gives accurate results. The bicubic interpolation technique produced smoother result
as compared to nearest neighbour and bilinear interpolation. In comparison with the
centroid and parabolic functions, the Dirichlet function gives more reliable results
when estimating subpixel translation. The Dirichlet function produces reliable results
with an error range of (±0.1,±0.25). In the case of noise, the method produces an
accurate result, and in simulation experiments, the error is less than one pixel. To
determine the translation between satellite images taken over a three-month period
provided a reliable result.

References

1. Samritjiarapon O, Chitsobhuk O (2008) An fft-based technique and best-first search for image
registration. In: International symposium on communications and information technologies.
IEEE, pp 364–367
2. Gonzalez RC, Woods RE (2008) Digital image processing. Prentice Hall
3. Jurgen L (2019) Sub-pixel image translation estimation on a nanosatellite platform, Ph.D.
thesis. Stellenbosch University, Stellenbosch
4. Ezzeldeen R, Ramadan H, Nazmy T, Yehia MA, Abdel-Wahab M (2010) Comparative study
for image registration techniques of remote sensing images. Egyptian J Remote Sensing Space
Sci 13(1):31–36
604 S. Jindal et al.

5. Foroosh H, Zerubia JB, Berthod M (2002) Extension of phase correlation to subpixel registra-
tion. IEEE Trans Image Process 11(3):188–200
6. Reddy BS, Chatterji BN (1996) An fft-based technique for translation, rotation, and scale-
invariant image registration. IEEE Trans Image Process 5(8):1266–1271
7. Xie H, Hicks N, Keller GR, Huang H, Kreinovich V (2003) An idl/envi implementation of the
fft-based algorithm for automatic image registration. Comput Geosci 29(8):1045–1055
8. Patel D, Gandhi V, Patel V (2014) Image registration using log polar transform and fft based
scale invariant. Int J Eng Res Appl 4(7):79–84
9. Juan L, Gwun O (2009) A comparison of sift, pca-sift and surf. Int J Image Process (IJIP)
3(4):143–152
10. Sarvaiya JN, Patnaik S, Kothari K (2012) Image registration using log polar transform and
phase correlation to recover higher scale. J Pattern Recogn Res 7(1):90–105
11. Jingying J, Xiaodong H, Kexin X, Qilian Y (2002) Phase correlation-based matching method
with sub-pixel accuracy for translated and rotated images. In: 6th international conference on
signal processing, vol 1. IEEE, pp 752–755
12. Tong X, Ye Z, Xu Y, Gao S, Xie H, Du Q, Liu S, Xu X, Liu S, Luan K et al (2019) Image
registration with fourier-based image correlation: a comprehensive review of developments and
applications. IEEE J Selected Topics Appl Earth Observations Remote Sensing 12(10):4062–
4081
13. Gleason SS, Hunt MA, Jatko WB (1991) Subpixel measurement of image features based on
paraboloid surface fit. In: Machine vision systems integration in industry, vol 1386. International
Society for Optics and Photonics, pp 135–144
Impact of Green Communication
and Technology System

Sumit Dhariwal and Avinash Raipuria

Abstract Green communication implies inspecting out for the examination of

viability as for biological conditions, imperativeness capability, and the purposes
behind correspondence on mobile phones generally. The commitment to a green
exchange to enhance business environmental responsibility and encourage an ecolog-
ical period of framework types of materials and buildings. This paper attempts to
introduce the most recent examination on green connections and frameworks organi-
zation for state of the art wired and prompt and underhanded consequences for nature.
Ongoing exploration on convenient advancement incorporates the improvement of
the quantity of pinion wheels manhandled yearly that has set off the requirement
for upgrade in the field of essentialness successful correspondence. The paper is a
composing study concerning the shows to improve essentialness adequacy in green
correspondence frameworks. It clarifies the various pieces of examination, plan, scat-
tering, and advancement of shows, just as constructions of green communications
and frameworks organization.

Keywords Green communication · ICT · D2D · MIMO · GIOT and HTr network

1 Introduction

I have contributed tremendously to the communication of time, and considering that,

I have given attention to the environment by using the energy from the work of energy
through this paper. The use of the communication equipment increases the likelihood

The original version of this chapter was revised: A affiliation to an earlier paper was omitted. The
correction to this chapter was available at
https://doi.org/10.1007/978-981-19-0284-0_58
S. Dhariwal (B)
Manipal University Jaipur, Rajasthan, India
e-mail: sumit.dhariwal@jaipur.manipal.edu
A. Raipuria
Malaysia University of Science and Technology, Kota Damansara, Selangor, Malaysia
e-mail: avinash.raipuria@phd.must.edu.my

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022, 605
corrected publication 2022
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_44
606 S. Dhariwal and A. Raipuria

of the spread of radiation, which is fatal for the environment. The importance of green
communication emphasized by this paper offers a refined solution to all the problems
the researcher has attempted to solve.
It has been noticed that presently, the total 4% energy is spent by the ICT frame-
work that reasons across three percentage (3%) of the commonly CO2 expenditures,
once the contrasted with the generally CO2 releases through planes is a total CO2
quarter of the productions by different automobiles. Pursuant to ongoing study, versa-
tile endorsers comprise the majority of the global populace. The supposition made
by the media transmission market is that to reach that cutting edge, an expansion
in endorsers will be needed, as well as per supporter’s information rate, and the
rollout of extra base stations if there is an occurrence of portable systems. The move
of versatile systems is primarily centered on diminishing the vitality utilization of
terminals, whose battery power forces necessities in a similar respect.
As a reality, late thoughts of versatile innovation incorporate the development
various hardware abused each year that has introduced the importance of progressing
in the field of green communications. Get necessities the number of subscribers
and base stations have both been increased, which have inspired an expansion in
communication movement.
It talks about the problem of energy proficiency and utilization in interchanges
systems. Ericsson’s ongoing test, report says half of versatile administrator effective
costs is equal to the expenses of the energy. In this manner, media communications
applications can have an immediate, economical effect on bringing down ozone
depleting substance discharges, control utilization, and energy proficient remote
transmission strategies (Fig. 1).

Fig. 1 Energy utilization Indicators of Energy Utlization

indicators Transportation- 12.1%

Enviormental Changes
12.7%
Production
&Construction - 11.9%
Fuel Combustion- 8.5%

Industry- 4.2%

FugitiveEmission- 4.2%

E-Wast&
Industrilization- 3.1%
Electriccity & Power
Genration 29%
Agriculture - 13.9%
Impact of Green Communication and Technology System 607

2 What is Green Communication?

Green Communication is a communication system in which we communicate using

work using radiation, thereby eliminating the risk of diseases mainly including
nervous systems as well as the prevention of nature.
Green correspondence is the act of choosing energy proficient interchanges and
systems administration advancements and items, limiting asset utilized at whatever
point conceivable in all parts of correspondence. The number of portable member-
ship in the data and correspondence innovation the number of subscribers and the
number of base stations has been expanded. According to current analyses, world-
wide portable memberships have increased dramatically from 500_million in_2000
to 5_billion in the year of 2012. Also, tend to achieve worldwide infiltration of
hundred percentage (100%) after the year of 2020.

3 Techniques

There’s a compelling impulse for green communication in the improvement of 5G.

Part of the progress is less scrutinized because of the green communication used in
5G. It is stated below.

3.1 [D2D] Device-To-Device Communication

D2D indicates to communication between devices, which might be phones or vehi-

cles. This innovation opens up new gadget driven interchanges that regularly require
no immediate correspondence with network framework, consequently to address part
of the organization limit issue as 5G permits more gadgets to be associated in quicker,
more solid organizations [11]. Guarantees are normal shown in Fig. 2.

Fig. 2 Device-to-Device
Communication
608 S. Dhariwal and A. Raipuria

Fig. 3 Multiple input

multiple outputs Malik et al.
(2017)

3.2 Multiuser Multiple Input Multiple Output

Multiuser Multiple Input Multiple Output (MIMO) frameworks are on the ascent. In
such frameworks, a base station having different receiving wires all the while serves
various clients having single radio wire gear [12]. All clients in such situations can
share multiplexing advantages, as delineated in Fig. 3.

3.3 Hiterogeneous Networks

One includes the utilization of a heterogeneous organization of macro cells and some
more modest cells (e.g., microcells, pico cells, and femtocells). A backhaul network
is finished by adding the base station to the core network through wired [13], wireless,
or mixed architecture as shown in Fig. 4.
Force utilization in HetNet can be constrained by placing little cells into rest
mode when there is a low burden or no heap [12]. In [13], advancement and logical
methods are utilized to make the general framework proficient and to control power
utilization in backhaul networks.
Impact of Green Communication and Technology System 609

Fig. 4 A 5G heterogeneous network system design Malik et al. (2017)

3.4 Green Internet of Things

Green Internet of Things 5G is another possible element of green correspondence as

shown in Fig. 6. Energy Efficiency in IoT has been pointed toward giving joining of
various fields, that is concentrated in [14] (Fig. 5).

4 Types of Environment Impacts

The environmental impact of the terms is used and regularly used when examining
green progress in communication. The carbon discharges as of now get the most
consideration because of the issue of a worldwide temperature alteration and the
adjustment in the atmosphere. When thinking about the ecologically friendly arrange-
ment, air contamination, water contamination, and earth characteristic, security of the
ozone layer, common property use, squander decrease, and so on must be considered
as the principle sway.
Media communications hardware typically contains a lot of scarce resources and
heavy metals, and the biggest test is the primary test on earth to isolate these materials
through mining and waste treatment. With this technique, the amount of waste created
with and without the material can be determined. We see that the reuse is 93 kg,
610 S. Dhariwal and A. Raipuria

Fig. 5 The Internet of Things is going green Malik et al. (2017)

Diffrent Countries
Percentage (%) of CO2 Emission

35
30
25
20
15
10
5 2014
0 2015

Fig. 6 CO2 production by the most important providers Sejja et al. (2017)

whereas, without the reuse of 12 kg of auxiliary material for PC, there would be
about 500 kg of mining waste. So, to have a total perspective on the natural impact
of an object, each of the five phases must be considered, for the most part, material
extraction, manufacture, use, transportation, and end of life.
Impact of Green Communication and Technology System 611

Energy use and materials used in data communication innovation, which is iden-
tified as a carbon discharge, must also be considered. When examining such an
equivalent, we need to consider the discharge of a substance that damages ozone,
which incorporates the Kyoto convention, for example, CH4 (methane), N2 O (nitrous
oxide), CO2 (carbon dioxide), PFCs, HFCs, and SF6 (sulfur hexafluoride). Green-
house gases have an earth-wide temperature raise potential that is assumed for
a hundred-year horizon. These temperatures are most important for the EWT of
carbon dioxide. CO2 emissions are mainly associated with off-lattice locations that
include remote areas. Diesel-control generators fuel most such destinations. The
ozone harming substance outflows for atomic power are extremely low and atomic
power impacts affect nature, for example, the treatment of atomic waste.

5 Express and Secondary Influences

When breaking down the immediate and backhanded effects ecological inviting
arrangements are to be considered. For instance, actualizing an answer, which dimin-
ishes the energy utilization of an administration, brings about an immediate effect.
Aberrant effects of arrangements are identified with the more extensive ideas and
selection of the arrangement. Composition of letter supplying attendance, issue of
vehicle, use of paper, and so forth. Backhanded effect decrease ordinarily constrains
the ecological issues. Since the varieties in the political, money related, educa-
tional elements are subject to these, those decreases are normally harder to antici-
pate. Expanding proficiency is a fundamental methodology in restricting ecological
effects. On the off chance that we think about the email address to which we are
by replacing each letter sent by email to a great extent lessening the effect of those
letters. This case exhibits that instating the circuitous natural effects of arrangements
is a troublesome errand, which must be done with extraordinary consideration.

5.1 Mobile Devices

Different portable specialized gadgets, for example, cell phones, smart watches,
personal wearable specialty gadgets, human service gadgets, and AI glasses have
led us to a sensible society. With the rapid progress of remote correspondence, the
absence of co-activity between portable hubs affects the nature, as well as results in
an imbalance in the use of the environment, creating unnecessary energy use of cell
phones.
The classes coming into the system can be broadly divided into three types, for
the most part, standard cell phones, PCs, and cell phones. Ongoing testing shows
performance is estimated at 1.5 kwh per year, depending on each 50th hour of charge.
That equates to a backup state of 30% of the battery’s limit each day and 40% of
the remaining time, as well as versatile and green communication between systems
612 S. Dhariwal and A. Raipuria

for example, data sharing, the range of energy mindfulness, directing adjustment,
and information reserving empowers giving potential advantages to enhancing and
adjusting the asset utilization and recoveries the energy of altogether portable and
remote systems, consequently green.

5.2 Carbon Footmark Models and Conventions

Communication turns into the most extreme significant and promising exploration
point for future portable systems. The point of the exceptional issue is to spur inquires
about to distribute their most recent research cutting-edge problems and difficulties
in the green communication field for mobiles. Assessing versatile systems, which
incorporate sorts of hardware of the client, utilizing systems, projects of industry of
endorsers working. Use of appropriate systems as a result of the information traffic
performed by portable supporters [1, 1].
Carbon impressions used for portable interchange models can be divided into five
types:
(1) Mobiles production which is comparable to the creation including minimal
effort mobiles, Smartphone’s, and frameworks, on brief deals in the year and
covering all administrators recently year.
(2) Mobile gadget’s activities such as producing and then charging its battery and
hook. For this situation framework task for frameworks, additional screens and
different administrators are incorporated.
(3) RAN destinations task, which incorporates the productivity if there should be
an occurrence of the power to base station locations, control sites, and center
destinations.
(4) The operator exercises incorporate workplace tasks, supplies exercises, tasks in
the vehicle armada, and business venture out activity identified with all clients’
business exercises.
(5) Mobile system clients create server farms and transport. That is a versatile
system client creates the allotment of different systems [2] (Fig. 7).

5.3 Radio Approach Systems

Development for the radio access organize relies upon system type hardware overall
growth test. The measure of RAN power utilization in 2007 was about 20 kwh. The
ongoing examination gives us the thought regarding the development of new locales
and the expulsion of old site hardware. From the study, it is noticed that on a normal
there is a 10% diminishing of new base station when contrasted with the earlier year.
The base station model can be seen as a mix of the items offered.
Impact of Green Communication and Technology System 613

10
9 Cement Production
8
7 Other Fuels
6
Gt CO2
Gas
5
4 Oil
3
Coke
2
1 Cleaned Coal
0
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Raw Caol

Fig. 7 China’s main foundations Sejja et al. (2017)

We estimate that by 2022, the pattern progresses by as much as eight percent

each year over the study time and is considered to be in steady progress. The world-
wide normal of base station site power is anticipated to increase to approximately
1.8 kW in 2007 and decrease to approximately 1.2 kw in 2020, underestimates for
the installation of new and removal of old gear each year. Investigation accepts a
takeoff model expecting somewhere in the range of 600,000 and 675,000 destina-
tions recently conveyed and up to 300,000 locales removed from administration every
year [1].

6 Energy Competence Matrices

Energy proficiency measurements give data that can be utilized to survey and look
at the energy utilization of different segments of a cell organize and of the system
all in all. These measurements additionally help us to set a long haul to investigate
objectives for lessening energy utilization. With the expansion in research exercises
relating to green correspondence and because of the inherent contrasts and signifi-
cance of different correspondence frameworks and execution measures, it is hard for
one single measurement to get the job done.
While the meanings of vitality proficiency measurements at the part and gear levels
are genuinely direct, it is all the more testing to characterize energy effectiveness
measurements at the framework or system level. Measurements from the base of the
system evaluate energy effectiveness at the system level, thinking about the system’s
limitations and properties.
Objectives related to the green cell system:
• Enhancement of energy proficiency.
• Improving knowledge of the system through interchanges between energy
utilization and outer environments, that is a circulation capacity.
614 S. Dhariwal and A. Raipuria

Power Usage of Cellular Network

70

Power Usage (%)

60
50
40
30
20
10 Cellular Network Power
0 Consumption

Fig. 8 Effect of cellular networks power consumption

• Incorporation of system foundation and system administrations to empower the

system to remain gradually reactive what’s more, to require less capacity to work.
• Low carbon emissions [2, 3].
• Energy savings in base stations [9, 10].
• Improvements in PA: Linear PAs → 90% wastage.
DPD, Doherty, GaN based PA.
• Power saving mode: Sleep mode, discontinuous Tx/Rx.
• Optimization: BS placement, cell size (Fig. 8).
Renewable energy:
• Sustainable bio
• Fuel
• Solar energy
• Wind energy

7 Conclusion

This paper shows a diagram of energy utilization issues in green correspondence

systems and depicts organize energy sparing strategies. It is recognized that the
regular energy utilization issue green correspondence arranges and portrays the
systems to improve the energy efficiency of the system is used. This issue can be
arranged illuminated by time and repetition spaces energy sparing issue crosses the
various frameworks or systems is less comprehended. More endeavors are required
from the displaying to specific arrangements. Radio access systems are addition-
ally incorporated into this paper. Other significant issues incorporate the variety
of portable administrator inclusion and the kinds of administrations starting with
one territory then onto the next. As correspondence systems experience exponential
Impact of Green Communication and Technology System 615

development around the world, it is important to design as well as improve the remote
access system.

References

1. Thirumaran JJ, Dhinakaran S (2018) A review on green communications. Int J Eng Res Technol
(IJERT) 6(13):1–3
2. Alsharif MH, Nordin R, Ismail M (2013) Survey of green radio communications networks:
techniques and recent advances. J Comput Netw Commun, Article ID 453893, 13
3. Strinati EC, Domenico AD, Herault L (2011) Green communications: an emerging challenge
for mobile broadband communication networks. J Green Eng, CEA, LETI, MINATEC, 38000
Grenoble, France 267–301
4. Samdanis K, Rost P, Maeder A, Meo M, Verikoukis C (2015) Green communications:
principles, concepts and practice. 1st edn. Wiley, Ltd.
5. Zhang H, Gladisch A, Pickavet M, Tao Z, Mohr W (2010) Energy efficiency in communications.
IEEE Commun Mag, IEEE Xplore. 48(11):48–49
6. Mauri L, Rodrigues J (2012) Green communication and networking. In: Joel (eds) Second
international conference, GreeNets 2012, Gaudia, Spain, Oct 25–26
7. Chen T, Harjula I, Kiviranta M (2011) Energy saving techniques in layered structure of LTE
systems. In: Mobile VCE Wksp. Green Radio: In: conjunction with SDR’11-WInnComm-
Europe, Brussels, Belgium, June 2011
8. Chen T, Yang Y, Zhang H, Kim H, Horneman K (2011) Network energy saving technologies
for green wireless access networks. IEEE Wireless Commun 18(5):30–38
9. Hasan Z (2011) Green cellular networks: a survey, some research issues and challenges. IEEE
Commun Surv Tuts 13(4):524–40
10. Mancuso V (2011) Reducing costs and pollution in cellular networks. IEEE Commun Magz
49(8):55–62
11. Gandotra P, Jha R, Jain S (2017) Green communication in next generation cellular networks:
a survey. IEEE Access (99):1–1
12. Zhou X, Bai B, Chen W (2015) Antenna selection in energy efficient MIMO systems: a survey.
China Commun 12(9):162–173
13. Mowla MM, Ahmad A, Habibi D, Phung QV (2017) A green communication model for 5G
systems. IEEE Trans Green Commun Netw (99):1–1
14. Shaikh FK, Zeadally S, Exposito E (2017) Enabling technologies for green internet of things.
IEEE Syst J 11(2):983–994
15. Zhang D, Zhou Z, Mumtaz S, Rodriguez J, Sato T (2016) One integrated energy efficiency
proposal for 5G IoT communications. IEEE Internet of Things J 3(6):1346–1354
Recent Advancement in Fine-Grained
Access Control and Secure Data Sharing
Scheme for Distributed Environment

Vidhi Pandya , Atufaali Saiyed , and Khushi Patel

Abstract The demand for Big Data also increases with the increasing rate of usage of
Internet. The storage, usage, and management of Big Data are the biggest challenge
in today’s world. Cloud computing and fog computing are new eras of IT indus-
tries which provide various computing services, distributed storage and resources
over the internet which is available any-time-anywhere in the world using Internet
but still, the security and privacy of these data is the major concern for business
enterprises. The data shared among different users over distributed storage must be
configured with fine-grained access control mechanisms. The distributed platform
can use various encryption standards like DES, AES, Blowfish, Elliptic curve, RSA,
etc. but they only offer confidentiality and integrity but for the fine-grained access
control policies the mechanisms like identity-based encryption and attribute-based
encryption algorithm are the best possible solution for authentication and confiden-
tiality. Another option is homomorphic encryption which allows the processing of
encrypted data. In this paper, the identity-based encryption, attribute-based encryp-
tion, and homomorphic encryption techniques are analyzed with their use cases and
limitations which can be useful in the field of Industrial IoT, Industrial cloud, Big
Data generated manufacturing industries, etc.

V. Pandya
Department of Computer Science & Engineering, Devang Patel Institute of Advance Technology
and Research (DEPSTAR), Faculty of Technology and Engineering (FTE), Charotar University of
Science and Technology (CHARUSAT), CHARUSAT Campus, Changa 388421, India
A. Saiyed
Department of Information Technology, Devang Patel Institute of Advance Technology and
Research (DEPSTAR), Faculty of Technology and Engineering (FTE), Charotar University of
Science and Technology (CHARUSAT), CHARUSAT Campus, Changa 388421, India
K. Patel (B)
Department of Computer Engineering, Devang Patel Institute of Advance Technology and
Research (DEPSTAR), Faculty of Technology and Engineering (FTE), Charotar University of
Science and Technology (CHARUSAT), CHARUSAT Campus, Changa 388421, India
e-mail: khushipatel.ce@charusat.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 617
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_45
618 V. Pandya et al.

Keywords Identity-Based Encryption · Attribute-based encryption ·

Homomorphic encryption · Cloud computing · Distributed computing

1 Introduction

With the advancement of Internet Technology, millions of computing devices are

operated and connected through internet which generates 2.5 quintillion bytes of
data each day [1]. According to the statistics of Forbes [2], “the amount of the newly
created data in 2020 was predicted to grow 44X to reach 35 zettabytes (35 trillion
gigabytes) and it is predicted that 175 zettabytes of new data will be created around
the world in 2025”. Today every industry like finance, health care, retail, govern-
ment, etc. generates a large amount of data which is Big Data. Big data improves
efficiency of a business model. It is a core of Industry 4.0. Still, the storage, usage,
and management of Big Data are the biggest challenge in today’s world [3]. As
many IoT devices like sensors, actuators, RFID cards, Mobile devices are not able
to store the data on the local storage; it increases the demand of third-party storage
for sharing and processing the data. Nowadays many large companies like Amazon,
IBM, Microsoft provide various services through cloud platforms [3]. Currently,
cloud computing is the best option for the storage, management, and processing of
Big data. However, the cloud computing offers many benefits [4] like outsourcing of
storage, computation to servers using Internet using several fees, flexibility, scala-
bility, less maintenance, access anywhere–anytime, etc. major issues [5] associated
with cloud environment are (i) control and transparency rights of the data will be
in hand of service provider and (ii) Trust and dependency. Other security concerns
with respect to cloud computing are related to security end user machines and data,
security of data during communication, and file system security.
Secure Interface and storage, secure data transfer, separation of data, and user
access control are the major requirements for solving the issues related to trust and
transparency in cloud environments. The use of encryption and digital signature at
data producer side are the best ways to provide data confidentiality. Many strong
and efficient schemes like DES, Triple DES, AES, Blowfish, IDEA, RSA, Elgamal,
ECC, SHA-512, MD5, etc. are available for data encryption, digital signature, and
hashing which can be used to provide confidentiality and integrity but it can’t handle
the fine-grained access control policies.
In order to provide legal access to authorized users, various access control mech-
anisms can be used but the main problem occurs when a server got compromised or
a secret key got revealed by any of the users. Moreover, in order to accomplish the
needs of resources and computing power with respect to the current scenario, most
of the applications need to be migrated to the distributed and dynamic environment
in which during the lifetime new users need to be added, some access control rights
need to be changed and/or removing the current users. The cloud service providers
need to adopt efficient access control and encryption schemes which solve both the
issues related to authentication and confidentiality.
Recent Advancement in Fine-Grained Access Control and Secure … 619

In order to solve the issues associated with conventional Public Key Infrastruc-
ture (PKI) and access control schemes, Identity Based Encryption was presented by
Shamir in 1984. Attribute-Based Encryption (ABE) is another public key encryp-
tion technique which can be used to provide accurate access control to data stored
on a third-party server by restricting privacy leakage of data producers and users.
Attribute-based encryption works on the attribute the person has to manage access
control policies. Another scheme which can be used to share the data securely on
a third-party remote server is homomorphic encryption which leads to significant
growth in the security domain and specifically in cloud computing. Homomorphic
encryption provides the transmission securely.
In this paper, Sect. 2 covers Encipherment and Digital Signature schemes and
Sect. 3 covers fine-grained access control mechanisms suitable for cloud, fog, and
IoT-based applications which requires fine-grained access control, confidentiality,
and integrity.

2 Encipherment and Digital Signature Schemes

The encryption techniques are used to protect the private data from unauthorized
users. The user can simply encrypt the data before storing it in a cloud platform
and at the time of downloading, the data can be decrypted before usage but it is
not suitable at the time of data sharing among multiple users. In case of symmetric
encryption, the owner needs to share the secret key if data needs to be shared [6]. In
case of asymmetric encryption, data duplication will be the major issue as data needs
to be converted into ciphertext using user’s public key and shared through cloud and
if data is shared among multiple users multiple copies will be generated to share the
same data because data must be encrypted using receivers’ public key and then need
to be shared on cloud storage [6]. In this section, various techniques are covered for
offering confidentiality and integrity.

2.1 Data Encryption Standard (DES)

DES falls under the category of symmetric key algorithm which is used to encrypt
or decrypt the message/information, developed by IBM in 1977. DES algorithm is
developed based on Horst Fiestel [7]. It is made up of fiestel network which divides
the input into two parts—left and right part of 32-bit [8]. Besides 16 fiestel rounds,
it includes two permutations, i.e., initial permutation which is held at the beginning,
and the final permutation which takes place at the end. In the DES algorithm, round
key generator is used to generate a key for each round. To create round keys, it takes
input of 64 bits which is known as cipher key.
In the early days, DES was a comprehensively used algorithm in various appli-
cations like military, commercial, and communication systems [7]. According to the
620 V. Pandya et al.

avalanche principle, strength of an algorithm is directly dependent on key size. As

DES algorithm uses 56 bits key, the key length has become very small with respect
to the super computers or workstations [9] In addition to this, DES is vulnerable to
brute force attack, differential cryptanalysis attack [8] and man in the middle attack.
To overcome problems associated with DES, NIST tried to develop a new and
more secure DES algorithm [9] which is later on known as Multiple—DES [8].
Triple DES, the most commonly used multiple-DES algorithm is proposed by IBM
to overcome the disadvantages of DES algorithm [8, 9]. It is the same as DES but
the only change is that it is used thrice on the same data [7] and used 168-bit key. To
provide more confusion or to improve the security, later various modes of 3DES is
introduced, i.e., DES-EDE2, DES-EDE3, DES-EEE3, DES-EDE2, and DES-EEE2
[9]. However, the brute force attack can be applied on 3DES to threaten security
goals provided by 3DES [8].

2.2 Advanced Encryption Standard (AES)

As a trade-off of DES and Triple DES algorithms, National Institute of Standard

and Technology (NIST) built up an algorithm called Advanced Encryption Standard
[8]. AES was evolved in 2001 by Vincent Rijmen and Joan Daeman [10] based on
Rijndael cipher [7]. The primary issue in the DES algorithm is the small key size.
AES defeats this issue by giving three distinct key-size dependent on the quantity
of rounds utilized in the encryption procedure. In the view of the key size, AES
is isolated into three modes that are known as AES-128, AES-192, and AES-256
which uses 10, 12, and 14 rounds, respectively. A single round is comprised of four
transformations, i.e., SubBytes, ShiftRows, Mixing, and Add Round Key.
Instead of Feistel network, AES is a substitution—permutation network [10].
Unlike DES, AES is not vulnerable to brute force attack and has a strong avalanche
effect [8] so that it is more secure. There are many applications where one can use
AES such as monetary transaction, security applications, and in a small device to
send encrypted message or data over a network [7].

2.3 Blowfish

Blowfish is the symmetric algorithm that was developed in 1993 by Bruce Schneier.
Blowfish is used as a stand in for DES or IDEA. The key size of blowfish is 32–
448 bits. The data is encrypted at a rate of eighteen clock cycles per byte. Blowfish
algorithm is divided into 2 phases. The key is break is divided into subkeys in the
expansion process. A key size is not more than 448 bits [11]. There are 32 S boxes
and P arrays. In the blowfish, 16 round Feistal and large keys are used. CAST-128
is similar to blowfish where fixed S boxes are used. The input is of 8 bit and output
is of 32 bits. The input of the p array is used by the rounds, then after the last round
Recent Advancement in Fine-Grained Access Control and Secure … 621

half of the block is XOR [9]. The function divides the input of 32 bit into four eight-
bit quarters and then this is used as input in the S boxes. After XOR final output
is generated [11]. Considering it is a Feistal network, it can be reversed simply by
XORing the p array into a Ciphertext block, then the p entries are used in reverse order
[11]. In many applications blowfish is used because it consists of strong encryption.

2.4 RSA (Rivest–Shamir–Adleman)

The RSA [7, 12] algorithm is one of the first widely adopted public key encryption
and digital signature algorithms which uses positive integers theory. This algorithm
can be used for encrypting the data without exchanging the secret key and also
used for digital signature. The modular exponential theorem is used for converting
plaintext into ciphertext and vice versa in RSA. To achieve confidentiality, public key
can be utilized for converting plaintext to ciphertext and a secret key which is known
to user only, can be used to decrypt the data while in order to provide authentication,
keys will be used in reverse order where the document/message can be encrypted
by using the secret key which is only known to user that proves the identity of any
user [12, 13]. Key generation process and encryption are slower. It is only suitable
for encryption of small data units, not scalable, memory usage is high, vulnerable to
factorization attack, timing attack and chosen plaintext attack are the limitations of
RSA.

2.5 Elliptic Curve Cryptography (ECC)

Elliptic Curve Cryptography was developed in 1985 by Neal Koblitz and Victor S.
Miller [14] which is based on Elliptic curve theory. It is like RSA but it is faster
and provides many ways to the researcher of the algorithms. ECC is applied in
wireless communication like RFID, mobile phones, etc. Currently, it is accepted
commercially and many professional bodies have also adopted the algorithm. It
provides security with a smaller key size than other cryptography algorithms. It
covers algorithms like key exchange and digital signature. The building blocks of
ECC is point multiplication [15]. In ECC the smaller parameters in the algorithm are
used in comparison with RSA. The advantages of storage space, electrical power,
and bandwidth are important in smart devices. The efficiency of ECC depends on
the computations and fast algorithm for the multiplication.
622 V. Pandya et al.

3 Fine-Grained Access Control Mechanisms

for Distributed Environment

Security is the major concern because of digitalization over the world. The data
is store nowadays in the cloud while leads to security issues. The data which is
stored is encrypted in any storage medium but if the algorithm or the operation is to
be performed then the decryption is required. Moreover, if encryption is performed
with symmetric or asymmetric encryption techniques, keys must be shared with cloud
service providers which lead to privacy issues. Here the user must be authenticated
before accessing data on the cloud as well as data must be secured from cloud service
providers to solve the issues related to secure data sharing among multiple users. To
overcome these challenges some techniques have been introduced which are covered
in detail here.

3.1 Homomorphic Encryption

The homomorphic encryption was introduced to perform directly operation on the

data which is encrypted. The homomorphic encryption concept can be explained by
the example of the diamond shop. The shop owner wanted her workers in the shop
to access and fixed things such gold, diamond, ring, etc. However, the owner did not
trust the workers so she doesn’t want the worker to directly access any of the things
and so the owner decided to make the transparent impenetrable glovebox [16]. The
box was transparent so things inside it are clearly visible but the door is lock with
key. And only key access was with owner. Worker can only access the things with
gloves. Thus, the jewelry shop analogy represents the function of homomorphic.
First time in 1978 Ronald Rivest, Leonard Adleman, and Michael Dertouzos
[17] proposed the Homomorphic Encryption. In Homomorphic encryption specific
type of computation is performed on the ciphertext which gives encrypted result
and the result is also in the cipher text. If the enduser stores the private data on
the unknown party server and wanted that no one other user can access the data.
But in the traditional encryption, the data is protected when it is transmitting but
data is not secured at the time of computation is being performed. In comparison,
homomorphic encryption provides the data security from starting to end even while
in the computations. Network intrusion, browser attack, and DDOS attacks become
risks to cloud computing. Some of the issues in cloud computing are availability,
data security, third-party control, privacy, and legal issues. To solve these security
issues in the cloud computing, the solution is homomorphic encryption algorithm
[18]. The solution is fit for the data in cloud computing and secure transmission in
the cloud.
Example: In Fig. 1, the set of the integers are used. The first operation is performed
is addition and two further the halving and doubling of number is performed. In the
addition number 3 and 5 are taken as plaintext. The inputs are encrypted by doubling
Recent Advancement in Fine-Grained Access Control and Secure … 623

Fig. 1 Numeric value example using homomorphic encryption

the values 6 and 10, respectively. Then in the cipher text addition is performed
between 6 and 10, which resulting in 16. This result is decrypted by dividing 2 and
the value obtained is the actual value of plaintext. The Table 1 brief out the recent
research remarks on homomorphic Encryption.

Table 1 Recent research directions in homomorphic encryption

Source Year Technique Remark
[19] 2016 Homomorphic encryption To reduce the size of cipher text for
efficient data processing
[20] 2019 Fully homomorphic encryption To enhance the Capability for the data
by proposing the secure distributed
model
[21] 2019 Fully homomorphic encryption Main Focus on novelty of privacy
preservation
during the service selection process in
an untrusted cloud service
[22] 2019 Homomorphic encryption and fully Comparison of performance and
homomorphic encryption efficiency of simple Homomorphic
Encryption and Fully Homomorphic
Encryption
[23] 2019 Fully homomorphic encryption With the different key sizes, the
experimental analysis is done and it
gives the execution in millisecond for
security level
624 V. Pandya et al.

3.2 Identity-Based Encryption (IBE)

IBE was proposed by Shamir in 1984 [24] but Bohen and Franklin has implemented
the IBE in 2001 [25]. The IBE, public-key encryption suitable for large systems,
solves the issues related to key management and certificate authority of conventional
cryptosystem. The idea behind IBE was to protect important data from unauthorized
access. In IBE, Private Key Generator (PKG) provides secret keys to the users and
the identity of users like email, IP addresses, etc. can be used as a public key for
encryption. Basically, IDE is divided into four sections, i.e., Setup, Private Key
Extraction, Encryption, and Decryption.
As shown in Fig. 2, sender Alice uses receiver Bob’s identity and pKPKG to hide
the message. After encrypting the message, Alice sends a message to receiver Bob
via insecure channel. The receiver Bob can decrypt the message using sKIDBob . Thus,
IBE eliminates the need for PKI which was the heart of conventional approach [26]
and makes certificate free system [27]. For the small framework, this methodology is
exceptionally productive yet it isn’t valuable for a large system to give the private key
to all the users by utilizing secure channel. In this manner, to evacuate the utilization
of secure channel, Tseng and Tsai changed the algorithm and included the idea of
time update key with identity key (private key) [26]. PKG produces time update keys
which are mapped with the client’s private key occasionally and sends it to clients
by means of public channel [26]. Other scenarios where IBE can be applied is on
the cloud to provide confidentiality, fine-grained access control, backward secrecy,
and forward secrecy [28]. To provide forward and backward secrecy, PKG must be
online constantly and a secure channel must be set up among PKG and user while
giving the private key [25]. One downside is that for an extremely enormous number
of clients, PKG can be a bottleneck [25, 27] (Table 2).

Fig. 2 Identity-Based Encryption [29]

Recent Advancement in Fine-Grained Access Control and Secure … 625

Table 2 Recent research directions in area of Identity-Based Encryption

Ref Year Technique Remark
[30] 2020 Self-Updated Encryption (SUE) RS-IBE Scheme is used to encrypt the text
but the problem occurs with this technique
when a user wants to decrypt the ciphertext
to get the original message. The proposed
scheme of this paper uses RUE technique
with RS-IBE to overcome the problem
[31] 2020 RIBE with SD method In this paper, the proposed approach is
made up of IBE Scheme, IBR Scheme,
and two-level HIBE Scheme. As user’s
credentials changes dynamically in IBE,
there is the requirement to provide a
technique or method to handle key
revocation. Sub-Set Difference Method is
used by authors to solve the problem. It
provides the security against
chosen-ciphertext attack
[32] 2020 IBE under RSA (IBEET-RSA) Security is the major concern in any of the
fields. In this paper, security in Healthcare
is discussed. To provide the basics goals of
security, IBE under RSA mechanism is
used in proposed system of paper for
WBAN (Wireless Body Area Network).
The proposed scheme is efficient and
secure against chosen identity and chosen
ciphertext attacks
[33] 2020 HIBAF In this paper, Hierarchical Identity based
encryption mechanism is used to provide
security in fog computing. The burden of
PKG is also reduced in the proposed
approach but does not remove it
completely
[34] 2019 IBE with FBAC (Function-Based This paper mainly focuses on the
Access Control) authentication issue in IoT. Devices must
be authenticated to prevent over privilege
access. To overcome this kind of problem,
IBE with FBAC mechanism is proposed in
this paper. This paper also provides the
research direction for dynamic access
control schemes and to verify the validity
of IoT devices

3.3 Attribute-Based Encryption

Attribute-Based Encryption (ABE) [4, 35] is the public key encipher which was
first invented by Sahai and Waters in 2006 that provides both confidentiality and
fine-grained access control. In ABE, keys and ciphertext are based on the identity
626 V. Pandya et al.

of the consumer or set of attributes inherited with the consumer. The ciphertext
is generated in such a way that the decryption of the ciphertext is only possible
when the attributes associated with key and the attributes associated with ciphertext
matches. In ABE, attributes will be used at the time of encryption and secret keys
corresponding to access policy will be used at the time of decryption to achieve
revocation, collusion resistance, and scalability. The ABE also reduces the problem
related to data duplication on cloud storage and owners can simply modify the access
control policies to change the data sharing rights without using extra resources. The
actors [4] involved in ABE are Data Owner (DO), Cloud Storage Server (CSS), Data
User (DU)/Data Consumers (DC), and Trusted Authority (TA) (Key Generation
Center). The authority uses attributes to generate the user’s secret key along with the
master key. The authority distributes public keys along with attributes to the Data
Owner and private keys to Data users. For successful data decryption, at least some
m components of attributes in encrypted text should match with attributes in secret
key. In order to add a new system or change in an existing system, it is required to
redefine the keys. The basic architecture of the ABE scheme is as shown in Fig. 3.
There are different four algorithms [35] involved in ABE are Initial Arrangement,
Key generation process, Encryption, and Decryption. Initially, by using two random
numbers p and q from the Zp* domain, the third-party trusted authority generates
Public Parameters (PP) and master secret key (MSK). During the key generation,
the trusted authority selects a distinct random number r from Zp* for each consumer
and a distinct random number ri from Zp*for each attribute X from a set of attributes
(AS) to generate each consumer’s private key Kpt. The trusted authority generates
the attribute keys for each consumer by utilizing the Master Key (MSK), Consumer’s
indices, and set of attributes (AS). The TA communicates the attribute group AGi for
every attribute X from AS to CSS. The cloud storage generated encryption keys for
each Consumers and Access structure tree T. The data owner can use the encryption

Fig. 3 Architecture of Attribute-based Encryption [4]

Recent Advancement in Fine-Grained Access Control and Secure … 627

algorithm with public parameters (PP), set of attributes, and an access structure tree.
The plaintext can be retrieved by using secret key and access structure parameters
only if the set of attributes associated with ciphertext matches with the set of attributes
associated with the secret key.
The data privacy, accurate access control rights, prevention of combination of
different user’s attributes, prevention of accessibility after certain amount of period
elapsed, prevention of accessing the resources once user left the system are the
functional parameters to evaluate any of the ABE Scheme and the total amount
required for computing resources and storage are performance parameters [6, 36].
One limitation of ABE is the data owner needs to encrypt the data using each user’s
public key.
Generally, schemes can be divided into different classes: one is Key-Policy
Attribute-based Encryption (KP-ABE) and another is Ciphertext- Policy-based
attribute-based encryption (CP-ABE).
In the KP-ABE scheme, the public key received from the trusted authority and
different attributes are utilized to convert the plaintext to ciphertext. The encrypted
data can be decrypted only if the attribute associated with the user’s private access
policy matches with the attributes associated with ciphertext [37]. The KP-ABE
scheme is less expressive as compared to others. Moreover, it is not suitable for
synergic systems because access rights are incorporated with private keys instead
of ciphertext. In the CP-ABE, the encrypted data needs to be attached with access
policy and the user’s set of attributes can be used to describe a private key. It is like a
role-based access control scheme. The public key received from the key generation
authority is used by the data owner to specify the access plan which can be used
to encrypt the data. The data can be decrypted by the consumer using a private key
received from the trusted authority if the attributes associated with the private key
match with the associated attribute of ciphertext [37].
Many extended versions of ABE are Multi-authority, proxy re-encryption and
accountability, revocation, etc. Key-Policy ABE can be further categorized into two
different categories [4] like (i) Multi-Authority (ii) Revocable. Ciphertext-Policy
ABE (CP-ABE) can be categorized into categories like (i) Hidden- Policy (ii)
Hierarchical (iii) Revocable (iv) Multi- Authority (Table 3).

4 Use Cases: Health Care

The smart healthcare is the most promising area which uses cloud infrastructure to
share the data among stakeholders like hospitals, doctors, patients, etc. The privacy
of shared smart healthcare data is the most crucial requirement in current scenario
[44]. As shown in Fig. 4, the patient data must be shared among multiple stake-
holders through the cloud environment to offer the services like remote tele surgeries
and diagnosis, remote consultancy, intrahospital monitoring. The hospital authorities
store the patient data on the cloud infrastructure which can be accessed by stake-
holders anywhere–anytime but it attracts different security threats [45]. In order
628 V. Pandya et al.

Table 3 Recent research directions in the area of Attribute-based Encryption

Ref Year Type of Paper/Type of Proposed Remark
Scheme
[3] 2019 Key Generation Secure ABE key generation scheme for
Industrial cloud is proposed to solve the
issues related to privacy of users in terms
of attributed associated with some
products by separating attribute auditing
and key generation functions
[38] 2018 Authorization Framework for Cloud-based Electronic Heath record
Electronic Health Records scheme which uses the concepts of
semantic web like Web Ontology
Language (OWL) and Semantic Web
Rule Language (SWRL) is proposed.
Limitations: The proposed scheme may
open the door for another category of risk
which is not considered
[39] 2020 Survey on KP-ABE, CP-ABE, In this Paper, ABE schemes are analyzed
anti-quantum ABE, on the basis of security and performance.
generic ABE constructions, and their Future directions: (i) Existing Fully
variations secure ABE schemes are not up to the
mark in terms of efficiency, access policy
a usage of random oracle model. (ii)
Complexity of Bilinear pairing is higher.
(iii) anti-quantum ABE is the new
direction
[40] 2019 Linear Secret sharing based policy The Proposed scheme uses proxy cloud
update and file update CP-ABE service provider for policy updates and
Scheme secret value used for policy update and
file update is invariable for security
enhancement. Future Direction: Time
Complexity can be reduced for file
update. Blockchain technology can be
integrated for transparent policy update
[41] 2018 ciphertext-policy attribute-based The authors proposed the equality tests
encryption with equality test scheme which is useful for delegating a
cloud server for equality test without
having access to plain text. Future
Direction: Space Complexity can be
focused. Moreover, the model is tested
for indistinguishability- chosen plaintext
attack but still more security level can be
achieved
[42] 2019 Survey: Revocation in CP-ABE Future Directions: Revocation problem
needs to be solved. Fault tolerance,
single point of failure problem and user
accountability can be resolved by
adopting Decentralized Blockchain
based environment. Reducing the
computation overhead is another area
which attracts researchers
(continued)
Recent Advancement in Fine-Grained Access Control and Secure … 629

Table 3 (continued)
Ref Year Type of Paper/Type of Proposed Remark
Scheme
[43] 2020 Survey on CP-ABE for access control, Future Direction: Revocation problem
authentication, Signcryption and and adoption of CP-ABE in the area of
authorization fog computing and Edge computing is in
demand

Fig. 4 Requirement of fine-grained access control in healthcare [46]

to offer the fine-grained access control, the IBE and ABE are the best available
options while to process the encrypted data without decrypting the data homomorphic
encryption is the best available option.

5 Conclusion

This paper mainly focuses on confidentiality of data and the access control when the
data is stored at a third-party server. The various encryption technologies which fall
under broadly in two categories, i.e., traditional and modern are discussed in this
paper. Basically, the traditional encryption technologies like DES, AES, RSA, ECC,
etc. work with confidentiality and integrity but according to the current scenario
efficient access control is required in case of sharing data among multiple users in a
distributed environment. The IBE and ABE consider the same and instead of hiding
the data to provide confidentiality, it avoids letting out the stored information to
630 V. Pandya et al.

unauthorized parties by controlling the access to the data. Many researchers have
already proposed different versions of IBE and ABE but still limitations are there
which opens doors for attackers. Currently, IBE and ABE are the emerging area for
researchers for privacy preservation in cloud computing. Homomorphic is another
technique which is used to provide end-to-end communication security in distributed
environments by performing specific type of computation. Extended versions of
Homomorphic encryption schemes are also available for secret sharing but still doors
are open for researchers.

References

1. Schultz J (2019) How much data is created on the internet each day? 08 June 2019
2. Press G (2020) 6 predictions about data in 2020 and the coming decade, 6 Jan 2020
3. Song Y, Wang H, Wei X, Wu LH (2019) Efficient attribute-based encryption with privacy-
preserving key generation and its application in industrial cloud. Secur Commun Netw
4. Kumar P, Alphonse P (2018) Attribute based encryption in cloud computing: a survey, gap
analysis, and future directions. J Netw Comput Appl 108:37–52
5. Charanya R, Aramudhan M (2016) Survey on access control issues in cloud computing.
In: International conference on emerging trends in engineering, technology and science
(ICETETS)
6. Liu C-W, Hsien W-F, Yang CC, Hwang M-S (2016) A survey of attribute-based access control
with user revocation in cloud data storage. Int J Netw Secur 18(5):900–916
7. Maqsood F, Ahmed M, Ali MM, Shah MA (2017) Cryptography: a comparative analysis for
modern techniques. Int J Adv Comput Sci Appl 8(6):442–448
8. Verma A, Guha P, Mishra S (2016) Comparative study of different cryptographic algorithms.
Int J Emerg Trends Technol Comput Sci (IJETTCS) 5(2):58–63
9. Bhanot R, Hans R (2015) A review and comparative analysis of various encryption algorithms.
Int J Secur Appl 9(4):289–306
10. Harini M, Pushpa Gowri K, Pavithra C, Selvarani MP (2017) Comparative study and analysis
of various Cryptographic Algorithms. Int J Sci Eng Res 8(5):2229–5518
11. Forouzan BA, Mukhopadhyay D (2015) Cryptography and network security. McGraw Hill
Education (India) Private Limited
12. Arora R, Parashar A, Transforming CC (2013) Secure user data in cloud computing using
encryption algorithms. Int J Eng Res Appl 3(4):1922–1926
13. Bhardwaj A, Subrahmanyam GV, Avasthi V, Sastry H (2016) Security algorithms for cloud
computing. In: International conference on computational modeling and security (CMS 2016)
14. Li X, Shen X, Chen H (2011) Elgamal digital signature algorithm of adding a random number.
J Netw 6(5):774
15. Amara M, Siad A (2011) Elliptic curve cryptography and its applications. In: International
workshop on systems, signal processing and their applications
16. Martins P, Sousa L, Mariano A (2017) A survey on fully homomorphic encryption: an
engineering perspective. ACM Comput Surv (CSUR) 50(6):1–33
17. Parmar PV, Padhar SB, Patel SN, Bhatt NI, Jhaveri RH (2014) Survey of various homomorphic
encryption algorithms and schemes. Int J Comput Appl 91(8)
18. Zhao F, Li C, Liu CF (2014) A cloud computing security solution based on fully homomorphic
encryption. In: 16th International conference on advanced communication technology
19. Ojha S, Rajput V (2017) AES and MD5 based secure authentication in cloud computing.
In: 2017 International conference on I-SMAC (IoT in social, mobile, analytics and cloud)
(I-SMAC)
Recent Advancement in Fine-Grained Access Control and Secure … 631

20. Patel J (2017) Secure hashing algorithm and advance encryption algorithm in cloud computing.
Int J Comput Inform Eng 11(6):753–757
21. Kaur N, Singh H (2012) Efficient and secure data storage in cloud computing through blowfish,
RSA and hash function. Int J Sci Res (IJSR) 4(5)
22. Geng Y (2019) Homomorphic encryption technology for cloud computing. Procedia Comput
Sci 154:73–83
23. El-Yahyaoui A, ECH-CHERIF EL KETTANI MD (2019) A verifiable fully homomorphic
encryption scheme for cloud computing security. Technologies 7(1):21
24. Li F, Khan MK (2011) A survey of identity-based signcryption. IETE Tech Rev 28(3):265–272
25. Boldyreva A, Goyal V, Kumar V (2008) Identity-based encryption with efficient revocation.
In: Proceedings of the 15th ACM conference on computer and communications security
26. Tseng Y-M, Tsai T-T, Huang S-S, Huang C-P (2016) Identity-based encryption with cloud
revocation authority and its applications. IEEE Trans Cloud Comput 6(4):1041–1053
27. Li H, Dai Y, Tian L, Yang H (2009) Identity-based authentication for cloud computing. In:
IEEE international conference on cloud computing
28. Wei J, Liu W, Hu X (2016) Secure data sharing in cloud computing using revocable-storage
identity-based encryption. IEEE Trans Cloud Comput 6(4):1136–1148
29. Baek J, Newmarch J, Safavi-Naini R, Susilo W (2004) A survey of identity-based cryptography.
In: Proceeding of Australian Unix users group annual conference
30. Lee K (2020) Comments on "secure data sharing in cloud computing using revocable-storage
identity-based encryption". IEEE Trans Cloud Comput
31. Lee K (2020) A generic construction for revocable identity-based encryption with subset
difference methods. PLoS ONE 15(9):0239053
32. Ramadan M, Liao Y, Li F, Zhou S, Abdalla H (2020) IBEET-RSA: Identity-based encryption
with equality test over RSA for wireless body area networks. Mobile Netw Appl 25(1):223–233
33. Farjana N, Roy S, Mahi M, Nayeen J, Whaiduzzaman M (2020) An identity-based encryption
scheme for data security in fog computing. In: Proceedings of international joint conference
on computational intelligence
34. Yan H, Wang Y, Jia C, Li J, Xiang Y, Pedrycz W (2019) IoT-FBAC: Function-based access
control scheme using identity-based encryption in IoT. Futur Gener Comput Syst 95:344–353
35. Rajeev A, Kenu TS, Arun R, Babu SS (2016) A survey on attribute based encryption schemes
for data sharing. J Comput Eng 19–23
36. Garg H, Dave M (2019) Securing user access at IoT middleware using attribute based access
control. In: 2019 10th International conference on computing, communication and networking
technologies (ICCCNT)
37. Edemacu K, Park HK, Jang B, Kim JW (2019) Privacy provision in collaborative ehealth with
attribute-based encryption: survey, challenges and future directions. IEEE Access 7:89614–
89636
38. Joshi M, Joshi KP, Finin T (2019) Delegated authorization framework for EHR services using
attribute based encryption. IEEE Trans Serv Comput
39. Zhang Y, Deng RH, Xu S, Sun J, Li Q, Zheng D (2020) Attribute-based encryption for cloud
computing access control: a survey. ACM Comput Surv (CSUR) 53(4):1–41
40. Li J, Wang S, Li Y, Wang H, Wang H, Wang H, Chen J, You Z (2019) An efficient attribute-
based encryption scheme with policy update and file update in cloud computing. IEEE Trans
Ind Inform 15(12):6500–6509
41. Wang Q, Peng L, Xiong H, Sun J, Qin Z (2017) Ciphertext-policy attribute-based encryption
with delegated equality test in cloud computing. IEEE Access 6:760–771
42. Al-Dahhan RR, Shi Q, Lee GM, Kifayat K (2019) Survey on revocation in ciphertext-policy
attribute-based encryption. Sensors 19(7):1695
43. Belguith S, Kaaniche N, Hammoudeh M (2019) Analysis of attribute-based cryptographic
techniques and their application to protect cloud services. Trans Emerg Telecommun Technol
3667
44. Patel K, Vyas S, Pandya V, saiyed A (2019) IoT: leading challenges, issues and explication using
latest technologies. In: 2019 3rd International conference on electronics, communication and
aerospace technology (ICECA), pp 757–762.https://doi.org/10.1109/ICECA.2019.8821970
632 V. Pandya et al.

45. Patel A, Shah N, Ramoliya D, Nayak A (2020) A detailed review of cloud security: issues,
threats & attacks. In: 2020 4th International conference on electronics, communication and
aerospace technology (ICECA), pp 758–764
46. Archer D, Chen L, Cheon JH, Gilad-Bachrach R, Hallman RA, Huang Z, Jiang X, Kumaresan R,
Malin BA, Sofia H (2017) Applications of homomorphic encryption. HomomorphicEncryption.
org, Redmond WA, Tech. Rep
Context-Based Multi-System Integration
for Authorization Using Dynamic
Privilege Escalation

Nidhi Sinha, Meenatchi Sundram, and Abhijit Sinha

Abstract Since the inception of the Internet of Things (IoT), the question of Security
had been pertaining all along. The initial research was focused on encryption. Then,
focus on speed and something that takes less space over the network and storage;
multiple study findings also tried to address application-level Security like Authen-
tication. Despite advancements in Authentication, Authorization, and application of
context-aware Security, much of the research focus is still needed to address concerns
democratic Security, applicability, generic integration, and interoperability of hetero-
geneous systems. This paper addresses integrating with heterogeneous systems and
how the solution can be enhanced to implement more comprehensive interoperable.
The paper finds a trusted way to authenticate the device and ascertain its identity,
how to trust it, and how much to charge—Authorize all for the entity outside the
ecosystem keeping context as a crucial decision point.

Keywords Internet of Things · Authentication · Nonce · Privilege Escalation ·

Grant · Permission · Dynamic authorization · Context · Context-aware security

1 Introduction

Whenever transmission occurs, the central concern to prioritize is Security and

privacy [1]. When the talk is about the devices that interact, it does not know what
kind of security threat may occur before, during, and post connection. Humans have
intuitions about something wrong that can happen whenever they enter an unknown
and unsafe zone; however, as devices are non-living, they never fear getting hacked or
destroyed. They either keep working as programmed or stop, and in many scenarios,
they may slow down. Here, the concept is instrumented to make the device under-
stand or have knowledge about the threats that may cause physical damage to the
machines/devices and affect the transmission and compromise the network. Secu-
rity in IoT is one of the most important aspects to be handled with care, and all

N. Sinha (B) · M. Sundram · A. Sinha

Garden City University, Bengaluru, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 633
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_46
634 N. Sinha et al.

the precautions should be taken care of when there is talk about overall technology.
Security can be a cause of concern and treated with appropriate Authentication &
Authorization [2, 3]. It can be more fine-tuned when granting permission will be set
even when devices are packed together. The priority [4] setting of devices can be
changed, and the Authorization is also be shifted to more prior or less prior condi-
tions. From the context perspective, the level of approval can differ depending on
its situation (context). The device or back-end involved in decision-making behaves
smarter and is aware of the overall context. The main objective of all these activ-
ities is to make wise (smart) and intelligent devices like humans. The application
of the proposed solution will ensure smooth communication where the escalation
of privilege is required (a) emergency time calling an ambulance. (b) alerts by govt
agencies about some calamity like a tsunami (c) traffic clearance for VVIP move-
ment, (d) broadcasting health alerts to nearby people on their IoT enabled devices,
and similarly many more use-cases.
First comes Authentication, which indicates whether the operator/requesting
entity (a person or another device) is the one it claims to be or not. The creden-
tials provided by the organization play a vital role when the Authentication of a
person or device. Everything has its unique identity and critical password in every
case of accessing things or data, or network. Regarding Authorization, the entity
has permission to access a particular space. Therefore, it can be taken care of as
authorization states (access levels) where the entity can perform specific operations.
Based on context, the absolute priority of the device gets change to bring out the
appropriate action by the interconnected devices.

2 Review of Literature

2.1 Review of Literature Done Based

on Authentication survey Summary

Context-Aware Scalable Authentication (CASA) [5] makes Authentication easier

or harder based on some user parameters (e.g., a user’s location or time since the
last log in) comparatively than making it equally rigid for all cases. However, the
core idea of this paper is based on the current user location, and typing pin makes
Authentication much complex for the unauthorized person on run time, which brings
more Security at run time. The paper does not discuss Authorization and setting
permission dynamically.
The author discussed the context-awareness [6] techniques for authentication and
access control mechanisms. Context-awareness and context-security concepts high-
light contextual attributes used to support and enhance IoT’s authentication and
access control mechanism. An authentication framework was described by Goel
et al. [7]. To create a context-aware environment and encourage role-based decision-
making, the authors used a combination of destination access control and a user’s
Context-Based Multi-System Integration for Authorization … 635

background, authentication policies, and other factors tagging with a lightweight a

provision for extending is included in the framework. Malek et al. [8] proposed a
context-aware paradigm. For Context-aware environment and computing authenti-
cation services, the proposed design can allow users to take action in context-aware
computing. Domains establish trust and share secrets between parties based on their
ideal level of assurance. The context-aware authentication service uses Context-data.

2.2 Authorization Survey Summary

Hu and Weaver [9] proposed a flexible and dynamic context-aware system. For
healthcare applications, there is a security infrastructure. The control model extends
the accessibility the role-based access control mechanism by tying access privileges
to context-sensitive data restrictions. They listed their model’s capabilities as follows,
demonstrating how authorization decision-making is based on context Concerning
tasks details. Wullems et al. [10] applied context-aware computing to the environ-
ment to improve. The architecture of Authorization the architecture suggested facil-
itates the role-based access control process. Access control policy that is mindful
of the context. They explained the use of dynamic architecture to incorporate the
architecture. However, the study not addressed run-time Authorization and priority
setting.

2.3 Context-Based Security Survey Summary

Matos et al. [11] proposed a system that stated context information about any system,
applications, and devices’ (information) environment and events related to integrating
the different vertical domains. The proposed architecture in that paper supports Edge
to fog-centric context sharing, which reduces unwanted data exchange. The paper
deals with two challenges; the first is the heterogeneity, and the second is scalability
and real-time sharing. However, the paper stated that dynamic behavior is missing
in real-time sharing. Therefore, Context-aware Security is introduced but not used.
In this survey, Perera et al. [12] observations are based on context-aware on the
perspective Internet of Things. The paper analyzed the in-depth context lifecycle
over the IoT effect of context. The whole paper states that the challenges in context-
aware are handled in desktop, web mobile, sensor network, and pervasive computing.
Authors Filho et al. [13] stated that security breaches in a private network by IoT
edge devices are an extensive concern to IoT’s comprehensive adoption.
The author discussed four different technologies named Trust execution environ-
ment (TEE), Security by separation, Blockchain for trusted communication, Context-
aware Security to cover various vulnerabilities and threats in the same machine. The
author stated the research limitation is overhead imposed by the cryptographic oper-
ation required by CoT and Blockchain; to overcome this, run allotment and discard
636 N. Sinha et al.

run play a crucial role in overcoming the overhead problem. Author Ghali et al.
[14] proposed a new access control scheme, CATRAC: Context-Aware Trust- and
Role-Based Access Control for Composite Web Services’ higher-level control and
Security. In this semantic key and public key, cryptography is used for Authentica-
tion. CATRAC addresses the threats used for both single or composite web service
scenarios. Both role-based and trust-based access controls combine to provide the
enhanced scheme to provide a better framework. The role-based access control
method restricts the IoT environment limitation is going to be overcome in our
proposed design of run-time security consideration.
Author Perera et al. [15] proposed the Context-Awareness for Internet of Things
(CA4IOT) architecture to help users by automating selecting the sensors according
to the problems/tasks at hand. The main focus is automated filtering, fusion,
and reasoning applied to the collective data stream using sensors. The CA4IOT
architecture consists of four layers: Data, Semantics, and Context Dissemination
Layer (DSCDL), Processing and Reasoning Layer (CPRL), Context and Semantic
Discovery Layer (CSDL), and Sensor Data Acquisition Layer (SDAL) to provide
context-aware Security however paper lacking of context-aware Security. Author
Jagadamaba et al. [16] focuses on adaptive security schemes which use the secu-
rity parameters like context and trust. Many other security parameters are applied to
the different applications and evaluated by security credentials like control, privacy,
and context awareness. The findings are that context acquisitions make the system
adaptive to new changes context and trust are essential to developing an adaptive
and accurate security framework. Therefore, security schemes must be attended to
carefully by incorporating trust and context-aware middleware. The paper focuses on
Authentication, Authorization, and Access management; however, dynamic Security
was not addressed.
Jason et al. [17] identified three benefits for building context-aware applications.
First, the infrastructure can be neutral of hardware platform, operating system, and
programming language, so many different devices and applications can access the
infrastructure. Second, middleware decouples the individual pieces of infrastructure
toes upgraded independently and dynamically while the system is still running. Third,
devices use sensors, processing power services, and data in the infrastructure. Said
et al. [18] proposed a context-aware security controller. Using a context-aware secu-
rity controller to reduce security risks stimulates protection frameworks in response
to the current situation details such as the type of application and the device’s capa-
bilities. Context-aware access control was proposed by Zhang et al. [19]. A dynamic
role model for ubiquitous applications is a scheme for access control dependent on
the user’s location and access. However, it is restricted. Authentication processes that
are feasible to protect applications that are widely used. Thiago et al. [3] proposed
an extension of UPnP Universal plug-and-Play, which aims to connect consumer
electronics and smart devices from different makes and models (Hetrogenity and
Interoperability). UpnP-UP follows the Authentication and Authorization mecha-
nism for UpnP standards, maintaining backward compatibility with earlier versions
of UpnP. However, no talk about context-aware security dynamically.
Context-Based Multi-System Integration for Authorization … 637

2.4 Problem Introduction

As we see in the scenario depicted in Fig. 1 the ambulance needs to connect with
multiple devices to request for green signal/corridor from traffic signaling systems.
Requesting pass or warning vehicles those moving ahead to provide access and
warning to the hospital to ensure the emergency ward is ready, despite the underlying
fact that the ambulance in question is possibly not associated with any systems. Other
systems in the above description need to ensure that the request urged by ambulance
can be authenticated as someone it can trust and authorized to allow operation. It
cannot be done with typical certificate/signature validation alone. It can establish
identity (authenticate), but what if the system is hacked or the intended activity is
not approved, which must be validated authorization check.
This paper attempts to solve an IoT device trying to communicate with different
devices with their control (IAM) system—overall, in a heterogeneous environment.
Allowing, at the same time, successfully authenticate and authorize to allow execu-
tion. As this paper purely focuses on security aspects, few assumptions are made:
The IoT devices can connect using a shared medium like Bluetooth/GATT (Blue-
tooth Low Energy Generic Attributes)/ZigBee. The protocol to communicate is also
familiar, similar to CoAP. All devices implement and provide a standard interface for
communication using CoAP like authenticating. To do an operation like asking to
open the gate, send an alarm, or in the case of a vehicle, it will flash a message to the
driver, and a hospital will ring warning for medical staff to be ready. The system will
provide a green signal blocking other traffic to ensure a fast pass across the crossing
for the ambulance.

Fig. 1 The ambulance needs to connect to multiple heterogeneous devices/systems to get

authorized and communicate [18, 20]
638 N. Sinha et al.

3 Solution

Summarizing the problem statement—this paper intends to solve Authentication

and authorize the action to be performed by requesting an IoT device. It needs to
be achieved under-listed constraints such that the devices can be heterogeneous, for
example: of different makes and models. Moreover, the ecosystem they are controlled
in is heterogeneous and need not know each other or be connected. Example: traffic
system, hospital.

3.1 Overview

Figure 2 below depicts a high-level sequence of action. In the below scenario, the
ambulance hereafter will also be called Requestor. It needs privileged access to send
messages or requests, attention to various Requestees like vehicles moving ahead
to get past, or requestee like traffic signal to get green signal blocking others for
speedier transport or requestee like a hospital for alerting staff be ready and open the
gates. The challenge for the Requestee devices is two-fold, viz:
• Find a trusted way to authenticate the device and ascertain its identity.
• How to trust and how much to charge—Authorize all for the entity outside the
ecosystem.
The high-level flow is explained in steps at this moment.
Requestor like an ambulance connects with its own IAM server (belonging to
the same ecosystem to which the ambulance belongs, like a hospital or ambulance
control system like 108 in India) to get a unique token. It needs to be passed to other
devices to establish identities and request authorization.
The IAM system where the ambulance belongs registers the token and sets
parameters (detailed further in the paper) to help establish identity and purpose.
At the same time, the device (Requestor) on the move (or not) can try to connect
with other devices in the vicinity (radius of reach like in case of Bluetooth 10 m, or
in case of Zigbee up to 100 m) and share the token.
The Requestee (device receiving the request) connects with their own IAM server
for Authentication and Authorization.
The IAM server, upon receiving a request, the IAM server identifies that the token
does not belong to their system, instead of from Central Identity and Access Registry
hence requests it to establish identity.
Once identity is established, and purpose is resolved, the Requestee IAM server
evaluates the context and may authorize the action. Once approved, the Requestee
device allows acting as a message to the driver or raise an alert in the hospital
emergency ward to be better prepared.
The solution assumes the ambulance indiscriminately keeps broadcasting about
the emergency. The context is that the ambulance ecosystem evaluates the need to
Context-Based Multi-System Integration for Authorization … 639

raise the mode to the emergency level. At the same time, it tries to connect to relay
its need of emergency, the other systems (within their ecosystem) try to evaluate the
condition with fitment in their context. For example, the moving vehicles on receiving
distress signal post successful Authentication and Authorization will assess‘ the
context. They may deem fit to show the message to the driver (which can be intrusive
at other times). On a comparable basis, a security camera-based IoT device in a nearby
cafe getting distress signal from ambulance even post successful Authorization of
device, when evaluating the context of the fitment of the message, may deem unfit
to process and act any.

3.2 Details

The detailed steps involved are broken into two significant steps. However, before
the actual Step kicks in. First, the role of Central Identity and Access Registry and
Requestor IAM needs to be clarified. The Central Identity & Access Registry acts
as a central governing and authorizing identity system, managed by a government
body or consortium of industry bodies as a self-regulated and trusted party. Any
organization or industry that needs to have its devices with escalated privilege must
be registered with the central identity system. The registration and trust granted need
to be mutual and require a manual approval process.
Step 1. In Step 1 above, the device needs complete access. In this case, the
“Requestor device” comets with its own IAM server. The IAM server validates the
context (evaluate need), generates a Nauthorization pass, and registers the same with
the Central Identity server. Then, the Central Identity server initiates a mock test to
validate the Nauthorization pass with the recorded IAM service before approving for
requesting from other systems. Once registered successfully with the Central Identity
server, the generated pass is returned to a device for further usage (Figs. 3 and 4).
Step 2. The steps involved are detailed below, where the requesting device
(Requestor) needs access while broadcasting or trying to connect to multiple devices
en-route. Requestor devices keep broadcasting or connecting to multiple devices in
route, indicating its intention and passing its identity embedded in Nauthorization
Pass to all devices trying to connect to and get validated.
It is more of async communication, where the Requestor generally does not wait
for success or response (however, it can if needs to be). Instead, the Requestee devices
can react once it finds action authorized in ways it can, like show distress message
on driver’s head-mounted display screen to give pass or traffic signal can block other
ways except for the requestor device’s path is arriving.
The Requestee device, upon receiving the request, the Requestee device would
validate the token for two purposes, identify the device, and get it authorized of
action that the device needs to perform. In this paper, we intend to have a genuine
intention of action that can be taken wisely by the Requestee device itself (as in case
of grabbing attention during an emergency). However, specific activities can also be
chipped in.
640 N. Sinha et al.

Fig. 2 Solution overview [20] depicts the high-level flow of devices connecting authenticate and
authorize

Fig. 3 A sequence of steps involved in generating Nauthorization pass

Requestee IAM, in turn, would try to validate the token, which is not registered
locally and hence would fail. It would, in turn, get it authenticated from Central
Identity and access registry. The Central Identity and Access Registry system is
authenticated locally. However, it double checks with issuing organizations regarding
Context-Based Multi-System Integration for Authorization … 641

Fig. 4 The sequence depicting validation of requesting device and getting authenticated through a
chain of trust mechanism

validity. The issuing organization can indicate revocation of Nauthorization pass as

context (need) is no longer valid or purpose solved.
Once approved, the requestee IAM can re-evaluate the context of what it can
do with the emergency request pending, like display message or give signal or ring
alarm.
The optional last success message from the Requestee device to Requestor can be
async and need not necessarily indicate the success or failure of the request. However,
it can be devised to return a message for better communication.

4 Implementation and Evaluation

This section will detail the implementation (setup) and evaluation against various
parameters.

4.1 Implementation

Two server setups were programmed: (i) the Central Identity and Access Registry and
(ii) the Requestee/Requestor IAM. The code for Requestee and Requestor IAM are
clubbed as one web service since reason in the real-world that the same organization
can act like Requestee or to evaluate Requestor Identity server and are essentially
bare IAM server which integrates with Central Identity and Access Registry to either
register it is a device or being called upon for evaluation.
The web services are written in Java 8, run within a docker container supporting
1 CPU and 1 GB of RAM. As in many IAM server cases, the database used is
relational SQL DB, which we used PostgreSQL. The container runs on a desktop
642 N. Sinha et al.

with a processor Intel i5-1020U with four cores with a base processing speed of
1.6 GHz. The desktop has an overall RAM of 16 GB DDR4, which clocks at a speed
of 2666 MHz. The host Operating system used is desktop Ubuntu 20.04, while the
docker container uses CentOS using Sun JAVA. The Requestor and requestee device
are simulated using Postman as their primary job is to hit the API endpoint exposed
by either Requestor/Requestee IAM.

4.2 Evaluation

Below we look at various parameters and evaluate how the solution is against them.
Finally, we evaluate against selected OWASP TOP 10 security threats applicable to
evaluate Security.
Broken Authentication. The Nauthorization pass generated is stored in a central
server. Any device passing the Nauthorization Pass to the Requestee device (and
hence Requestee IAM), the pass is re-evaluated with Requestor for validity apart
from local verification in Central Identity and Access Registry. It makes the solution
grim for being hacked of manipulation or extended use.
Sensitive data exposure. The communication assumed is encrypted as in SSL
over COAP or ZigBee, making it indifferent to raw text reading or sniffing. Also, the
Nauthorization Pass critical systems; this just a token that does not reveal anything.
However, it can also be implemented based on JWT with signed content.
Broken Access Control. The access control (Authorization) cannot be used in a
prolonged fashion. Whenever a Requestee gets a request for operation, it connects
to Central Identity and Access Registry, which connects with the originator to re-
evaluate the applicability. Thus, it Makes it secure against prolonged unwarranted
use and keeps the usage under check.
Action Replay. Action Replay is possible by hackers gaining the entire packet
stream in transit and replaying the request payload. The net result would be over-
whelming for requestee device sets (devices & IAM) and may eventually lead to
system failure or serve other genuine requests. However, the duration of usage is
limited and hence the impact. It can also be mitigated by intelligent monitoring
systems employed by the current IAM system to distinguish between genuine and
hacked repeat offenses.
Storage Overhead. The device needs to store the Nauthorization Pass and
discover the device while passing its pass alone. The Nauthorization Pass, like other
tokens, is lightweight and contains UUID in a gene. However, it can be extended to
have JWT like structure for extended use. The UUID recommended solution does
not need more than few additional bytes and is hence very lightweight to store and
manage.
Communication Overhead. The suggested solution is like token passing over
HTTP(S) or COAP and no different. The back-end integration does the heavy lifting
or authenticating and Authorization. In this case, the frontend is devices with abso-
lutely no additional protocol or bytes to transfer other than authorization access
Context-Based Multi-System Integration for Authorization … 643

Fig. 5 Graph showing throughput with 50 concurrent users

token, which is Nauthorization Pass, making this solution compatible with existing
IoT device limitations (Fig. 5).
Performance. The observed performance is noted below when run using JMeter
calling endpoint to grant access to Requestee IAM (assuming being called from
Requestee device). As we see from Fig. 3 and Fig. 4, the throughput hovers around
250 requests/sec on average. Furthermore, the latency stays pretty swift to respond
within ~ 60 ms, keeping it highly responsive and scalable at par with any IAM
solution (Fig. 6).

Fig. 6 Graph depicting latency observed with increased concurrent users over time
644 N. Sinha et al.

4.3 Benefits of This Proposal

Benefits of this proposal are summarized below.

• The solution provides a way to integrate the various heterogeneous systems to
trust each other and still not with leaky doors to ensure those devices of different
origin and make from the altogether different ecosystem (IAM control) can be
trusted to operate and gain access.
• The context is evaluated numerous times to validate the applicability repeatedly
and, hence, the chance to retain or withdraw the emergency status of the pass issued
to the Requestor device. The context is also evaluated by giving the organization
(origin) and the requestee system suitable applicability and rejecting the request
if deemed unfit.
• The fundamentals of token for authenticating and Authorization remain intact and
used in the same philosophy. The back-end does all the integration and problem
solving with the trust-based mechanism.
• The Nauthorization pass can be revoked easily. The requestee system is under
no obligation to fulfill the request despite authenticated and authorized giving
autonomy of operation at its discretion.
• As we see in Latency and throughput comparison charts, the result data sets
worked well compared to any standard IAM solutions in the market like Key
cloak.

5 Conclusion and Future Work

The solution serves to Authentication and Authorization are dynamically driven by

context and chance for heterogeneous devices to interconnect and execute operations.
However, the current solution does not address few concerns that are evident and
summarize below for further study and solutions:
• The specific action is not presented in this paper being requested by the Requestor
device. Therefore, for this paper, it is assumed that it is a generic emergency and
the requestee device can take care of action it needs to perform, which would be
primarily monotonous like traffic signal giving way for a particular direction, or
car dashboard display would show the message to the user to provide brighter
side.
• The trust between Central Identity and Access Registry and Requestor organiza-
tion is high and implicit. The current solution does not consider if the Requestor
organization is compromised or hacked. There can be separate monitoring systems
built into Central Identity and Access Registry to block or get the organization
re-trusted temporarily.
Context-Based Multi-System Integration for Authorization … 645

We conclude our current work with acknowledgment to extend this solution further
for better integrity and fine-grained control. The current solution addresses connec-
tivity limited to general emergency and action mostly leftover to Requestee device
rather than Requestor identifying and needing specific action to be performed.

References

1. Zhang G, Parashar M (2004) Context-aware dynamic access control for pervasive applica-
tions. In: Proceedings of the communication networks and distributed systems modelling and
simulation conference, pp 21–30
2. Hong JI, Landay JA (2001) An infrastructure approach to context-aware computing. Human-
Comput Interact 16(2–4):287–303
3. Icons used in Fig. 1 and Fig. 2 viz: truck, CT-scan, gate, hospital building and server made by
https://www.flaticon.com/authors/smashicons. Last Accessed 19 May 2021
4. Malek B, Miri A, Karmouch A (2008) A framework for context-aware authentication. In: 4th
International conference on intelligent environments, pp 1–8
5. Thaker R, Yusera Farooq Khan YF, Mughal S (2018) Fog approach in the internet of things: a
review. IJSRCSEIT, 4(1b)
6. Sinha N, Sundaram M, Sinha A (2020) Authorization secure dynamic privileged escalation.
RTEICT, Bangalore India, pp 110–117
7. Sales T, Sales L, Almeida H, Perkusich A (2021) Towards UPnP-UP: developing context-aware
UPnP applications (position paper)
8. Sinha N, Sundaram M, Sinha A (2021) Instinctive and effective authorization for internet of
things. In: Proceedings of second (IC4S). Springer, Singapore, pp 567–579
9. Hayashi E, Das S, Amini S, Hong J, Oakley I (2013) Casa: context-aware scalable
authentication. In: Proceedings of the 9th SOUPS ‘13. ACM, vol 3. NY, USA, pp 1–3:10
10. Habib K, Leister W (2015) Context-aware authentication for the internet of things
11. Goel D, Kher E, Joag S, Mujumdar V, Griss M, Dey A (2010) Context-aware authentica-
tion framework. In: Mobile computing, applications, and services, vol 35. Springer, Berlin,
Heidelberg, pp 26–41
12. Hu J, Weaver AC (2004) A dynamic, context-aware security infrastructure for distributed
healthcare applications. In: Proceedings of the (PSPT), pp 1–8
13. Wullems C, Looi M, Clark A (2004) Towards context-aware security: an authorization
architecture for intranet environments. In: Proceedings of the (PERCOMW’04), pp 132–137
14. De Matos E, Tiburski R, Amaral L, Hessel F (2018) Context interoperability for IoT through
an edge-centric context sharing architecture. In: 2018 IEEE symposium on computers and
communications (ISCC). Natal, Brazil, pp 439–440
15. Perera C, Zaslavsky A, Christen P, Georgakopoulos D (2014) Context-aware computing for
the IoTs: a survey. IEEE Comm Surv Tutorials 16(1):414–454
16. Filho S, Tiburski R, Moratelli C, de Matos E, Hessel F (2019) Privacy and security of internet
of things devices
17. Ghali C, Chehab A, Kayssi A (2010) CATRAC: context-aware trust- and role-based access
control for composite web services. In: 2010 10th IEEE ICCIT, pp 1085–1089
18. Perera C, Zaslavsky A, Christen P, Georgakopoulos D (2020) CA4IOT: context-awareness
for internet of things. In: Presented at: IEEE international conference on internet of things
(iThings 2012). Besancon, France, pp 20–23 Nov 2012. Green Computing and Communications
(GreenCom), IEEE/ACM Int’l Conference on & Int’l Conference on Cyber, Physical and Social
Computing (CPSCom). IEEE, 775–782, (2020)
19. Jagadamaba G, Babu BS (2017) Adaptive security schemes based on context and trust for
ubiquitous computing environment: a comprehensive survey. Indian J Sci Technol
646 N. Sinha et al.

20. Said SBH, Guillouard K, Bonnin J-M (2013) On the benefit of context-awareness for security
mechanisms in LTE/EPS networks. In: Proceedings of the (PIMRC), pp 2414–2428
Next Generation Computing Technologies
Machine Learning-Based Security
Solutions for Healthcare: An Overview

Pallavi Arora, Baljeet Kaur, and Marcio Andrey Teixeira

Abstract The eruption of chronic diseases such as COVID-19 has re-emphasized

the need for people all over the world to have access to urgent healthcare. The latest
pandemic has shown the flaws in the conventional healthcare system, namely that
hospitals and clinics alone are incapable of dealing with such a crisis. Smart and inte-
grated wearables are one of the main technologies that favor modern healthcare solu-
tions (Godi, B., Viswanadham, S., Muttipati, A.S., Prakash Samantray, O., Gadiraju
student, S.R.: E-Healthcare Monitoring System using IoT with Machine Learning
Approaches. In: 2020 International Conference on Computer Science, Engineering
and Applications (ICCSEA). pp. 1–5. IEEE, Gunupur, India (2020). https://doi.org/
10.1109/ICCSEA49143.2020.9132937). These wearables can now gather data on an
unparalleled scale thanks to advancements in the Internet of Things (IoT). Health-
care is among many fields that have been transformed by IoT, with the Internet of
Medical Things (IoMT) systems being introduced as an IoT branch. Patients with
chronic diseases can be monitored remotely using IoMT systems. As a result, it can
offer prompt diagnostics to patients, potentially saving their lives in the event of an
emergency. However, protection in these vital systems is a major issue that has arisen
as a result of their widespread use. This paper presents an overview of the technolo-
gies that are being used on IoMT as well as some security problems found in the
literature. This survey will give an insight to the readers regarding the importance
of security in healthcare and the different machine learning methods used to address
that issue.

Keywords Machine learning · Healthcare · Security · Internet of Things

P. Arora (B)
IK Gujral Punjab Technical University, Kapurthala, Punjab, India
B. Kaur
Guru Nanak Dev Engineering College, Ludhiana, Punjab, India
e-mail: baljeetkaur@gndec.ac.in
M. A. Teixeira
Department of Informatics, Federal Institute of Education, Science, and Technology of São Paulo,
Catanduva, SP 15808-305, Brazil
e-mail: marcio.andrey@ifsp.edu.br

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 649
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_47
650 P. Arora et al.

1 Introduction

IoT has become an essential part of our daily lives in the twenty-first century. IoT has
many applications in various fields ranging from healthcare, Smart Homes, Smart
Grids, and Connected Cars to Industrial Internet. Healthcare is one of the most basic
human needs, but with a growing population and an increasing number of diseases,
the cost of healthcare facilities is also skyrocketing. In such cases, the advancement
of IoT has paved a new path in the healthcare industry that can aid in the discovery of
new treatment options [1]. IoT has enabled people to provide and receive healthcare
services through a network of interconnected networks and IoT devices. This paper
provides a thorough review of the literature on the various types of models proposed
by researchers over the years involving machine learning and deep learning [2, 3] in
IoT healthcare.

1.1 Contribution

The following are the primary contributions of this paper:

• We gave an insight into IoMT system architecture and different types of IoMT
devices.
• We emphasized the traditional healthcare system’s security and privacy issues.
• We discussed the usage of machine learning algorithms in healthcare security.

1.2 Organization

The remainder of the paper is organized as follows: Sect. 2 acquaints with the princi-
ples of IoT in healthcare. Section 3 provides a comprehensive description of IoMT-
related assaults at different stages. Section 4 provides a review of the literature
concerning the use of machine learning techniques for IoMT device security. Finally,
Sect. 5 provides the paper’s concluding thoughts.

2 IoT in Healthcare

2.1 IoMT Types

IoMT systems focus on providing essential or enhanced assistance for a wide range
of health conditions [4]. Ingrafted devices, such as pacemakers treating cardiovas-
cular disease, are required for particular healthcare disorders. In contrast, helping
Machine Learning-Based Security Solutions for Healthcare … 651

gadgets are mostly wearables for better healthcare experiences, such as smart-
watches. Because of these distinctions, IoMT systems are divided into two categories:
Implantable medical devices (IMDs) and Internet of wearable devices (IoWDs) [5].
• Implantable Medical Devices (IMDs): An IMD itself is embedded equipment
that is used to substitute, augment, or reinforce a tissue function. A pacemaker,
for instance, seems to be an IMD that regulates irregular heartbeats by assisting
the heart to pump at a standard rhythm whether pumping very quickly or slowly.
IMDs are tiny as well as own a decent energy duration. As a result, enabling
such gadgets to remain within the living organism over a longer period of time,
minimal energy usage little memory capability with compact cells which endure
a prolonged time was indeed required. Instances of embedded gadgets include
cardiac pacemakers, implantable cardiac defibrillators (ICDs), coronary stents,
hip implants, interocular lenses, and implantable insulin pumps (Fig. 1).
• Internet of Wearable Devices (IoWDs): Such are wearable gadgets that monitor
biomarkers such as pulse rate and might even positively affect people’s general
wellbeing. Some examples include wristbands, telemonitoring bands, electrocar-
diogram (ECG) screens, and bloodstream pressure screens. Wearables however
are presently the most recognized IoWDs for monitoring biomarkers such as
pulse rate and mobility. So, when the individual is not active, the monitoring can
identify slow and fast heartbeats. They are now commonly used for non-critical
patient monitoring. However, due to detector precision and energy power are both
limited, these devices are unlikely to replace IMDs in critical situations.

Fig. 1 Examples of IMDs and their associated body parts

652 P. Arora et al.

2.2 IoMT System Architecture

The vast majority of existing IoMT setups are organized into four tiers [6, 7], as illus-
trated in Fig. 2. Such levels encompass various information phases, beginning well
with individual’s biomarkers collection, and continuing with information keeping
as well as interpretation through the doctor enabling examination. Furthermore, any
client may access individual medical status via the website. IoMT architecture is
organized in various layers as follows:
• Sensor layer: Such a stratum has a collection of tiny sensors that are embedded or
donned by the patient that gather biometric data. Communication protocols such
as Wireless networking, Bluetooth, or the MedRadio available bandwidth, which
is reserved for IMDs, are being utilized towards transmitting data to the next level
[8].
• Gateway layer: IoMT sensors have storage and processing constraints and hence
transfer the data forward to the next layer, i.e., the gateway layer, without being
processed. The devices in this layer might have been the client’s cellphone or
indeed a dedicated access point (AP), which seem to be usually greater powerful
than perhaps detectors. It can carry out some data preparation tasks including veri-
fication, relatively brief storing input, and basic artificial intelligence-based inter-
pretation. Devices also use the Internet to communicate biosensor information to
the server.
• Cloud layer: The cloud layer is in charge of receiving data from the gateway and
storing, analyzing, and securely accessing it. Information analysis may detect vari-
ations arising in a person’s condition and display these for healthcare professionals
as well as clients for further action could be part of the analysis.
• Visualization/Action Layer: This layer provides information to physicians and
clients in order for them to take care of their wellness. Additionally, it encompasses
the doctor’s recommendations based on the patient’s medical issues. Prescriptions
and dosage adjustments for various medications are illustrations of actions.

Fig. 2 IoMT system architecture

Machine Learning-Based Security Solutions for Healthcare … 653

3 IoMT Security Model

3.1 Different Stages of IoMT Threats

IoMT platforms shall safeguard medical information throughout various phases of

gathering, transfer, as well as storing [5].
(1) Attack during data collection: The initial phase of an IoMT protocol is
the collection of patient data in the sensor layer. At such a point, assaults
could either be software-based (information manipulation) or hardware-based
(i.e., detector instrument alteration). Unless the detector equipment, as well as
programming, is compromised, these attacks can endanger patients’ lives. As a
result, for maintaining the mechanism operating, it is vital to secure information
from such threats.
(2) Attack during data transit: It involves equipment communication across all
four levels, such as IoMT devices with in device tier and APs in the gateway
levels. At this step, threats can change or prevent sensing information from
being communicated. While a result, safeguarding from data in transit threats
could prohibit information being getting distorted as it was transmitted between
the four levels.
(3) Attack during data storage: The client’s information is stored in the cloud
layer after it has been gathered and disseminated between both sensor and
gateway layers. Attacks in this layer range from account theft to DoS (denial
of service) attacks. It is very important to protect data at this stage since most
of the time data is resting thereby imposing a greater risk (Fig. 3).

Fig. 3 IoMT security threats and attacks at various layers

654 P. Arora et al.

4 Literature Review

4.1 Machine Learning in Healthcare

Machine learning is used to find trends in medical data and has excellent disease
prediction capabilities. Authors in [9] evaluate a variety of machine learning algo-
rithms for developing effective decision support for healthcare systems. This paper
discusses various machine learning methods for predicting diseases such as heart
disease, breast cancer, diabetes, and thyroid disease. It contributes to addressing the
research gaps in the development of effective decision support systems for medical
applications.
The research in [10] seeks to develop a model capable of properly predicting the
probability for hyperglycemia in people. This project incorporates three machine
learning classification methods to diagnose hyperglycemia during a preliminary
phase: Decision Tree, SVM, and Naive Bayes. The studies make use of the Pima
Indians Diabetes Database (PIDD), which is provided through the UCI machine
learning repository. Precision, Accuracy, F-Measure, and Recall are all used to test
the performance of the three algorithms. The classification system’s precision is
measured by the multitude of incidents that have been actually observed vs those
that are mistakenly categorized. According to the data, Naive Bayes outperforms
with a 76% precision.
In [11] the authors proposed wearable IoT for health tracking. In order to enable
real-time health tracking, it uses the BASN (body area sensor network) platform.
Among the worn sensors that have been incorporated are pulse, body heat, and arterial
pressure detectors. Secondly, most of the worn fitness tracking communication is
based on a smartphone as an information processing, display, as well as dissemination
portal, that has an impact on the smartphone’s typical everyday usage. Information
is instantly transferred into web data servers, and a compact wrist-worn display is
being placed like a substitute strategy for real-world content viewing.
The study showed in [12] does not stop at simple machine learning algorithms; it
also digs deeper into the idea of deep learning. Deep Neural Networks were contrasted
to Gradient Boosting Decision Tree, logistic regression, and support vector machine
methods for diagnosing infarction based on a dataset acquired from a large-scale
population database. Deep Neural Networks had an accuracy of 87.3%, while the
others performed similarly, except for SVMs. Out of all the algorithms, support vector
machines had the lowest accuracy. As a result, Deep Neural Networks are the most
appropriate, while SVMs are the least suitable. Although Deep Neural Networks
have been shown to be the best algorithm, it is still more complex than the others, so
one must consider the algorithm’s complexity before using it in real life.
Authors [13] emphasized the conventional healthcare system’s security and
privacy issues and also spoke about how blockchain can be used and integrated into
the healthcare system. They proposed a healthcare architecture for remote patient
monitoring and also stressed the usage of ML in blockchain to improve security.
Machine Learning-Based Security Solutions for Healthcare … 655

Table 1 A brief comparison of the works described in Subsection 4.1

Author Objective ML technique used Dataset Disease
[9] Developing J48, SVM, Naïve UCI Heart disease, breast
effective decision Bayes, Decision Diabetic Research cancer, diabetes, and
support for Tree Institute thyroid disease
healthcare systems in Chennai Swami
Vivekananda
Diagnostic Centre
Hospital
[10] To create a model Decision Tree, The Pima Indians Diabetes
that can accurately SVM, and Naive Diabetes Database
predict the risk of Bayes (PIDD)
diabetes in patients
[11] To develop SVM and neural Data from wearable Predicts disease that
wearable IoT for network sensor to cloud will emerge in near
real-time health Server directly future
monitoring
[12] To investigate the Deep Learning and Customized Stroke
viability and Gradient Boosting
efficacy of Deep Decision Tree,
Neural Networks logistic regression,
(DNN) and other and support vector
ML approaches for machine algorithms
predicting stroke
[13] To boost safety and Blockchain along Real-time data from Earlier prediction of
confidentiality of with Naive cloud server disease
patient’s data as Bayesian, k-Nearest
well as usage of ml Neighbor, and
with blockchain Decision Tree
technology for
healthcare

In the end, they discussed various research issues regarding blockchain and also
suggested solutions for those challenges (Table 1).

4.2 Security in Healthcare

Despite the various advantages of using mobile devices, there are significant privacy
and security issues with mobile health [14]. The use of successful encryption tech-
niques to prevent data leakage is crucial for the growth of mHealth, which is funda-
mentally based on trust. Researchers have looked at the privacy and security concerns
surrounding mHealth from a variety of angles. Because of its current inadequate secu-
rity support, health care data security has a loophole in that it is seldom supported by
direct implementation of Database Management System (DBMS) security features.
656 P. Arora et al.

Healthcare technology has the potential to protect, prolong and upgrade someone’s
life [15]. Techniques range beyond storing electronic health records (EHRs) to
tracking wellbeing and delivering medicines, as well as telemedicine technology
that allows treatment to be delivered remotely—even across countries. Because of
interconnected, easily accessible access points, obsolete systems, and a lack of focus
on cybersecurity, healthcare technologies are vulnerable to security threats. While
the focus has appeared to be on patient care, healthcare innovations contain large
quantities of useful and sensitive information. Because healthcare recognition has
been of greater significance unlike alternative forms of identification, personal profit
is often the motivation behind assaults. Certain attacks, such as information warfare,
could well be driven by sociopolitical considerations. People’s lives are jeopardized
if important health systems are targeted.
The advancement of information technology has sparked a transformative phase
in all sectors in terms of integrating these technologies in order to take advantage
of new functionality and capabilities [16]. Healthcare is a target because of the
lack of security mechanisms, as well as the importance of the data collected or
stored. We believe it is necessary to start on the procedural front preferably only at
the country’s scale having the goal of guaranteeing that the distributors of medical
systems satisfy specific consented security standards. The security standards for
computers, or similar devices used in the industry, maybe implemented and governed
by national regulations.
The aim of the paper proposed in [17] is to examine different forms of security
threats, as well as their depth and effect on organizations. Various IoT attacks are
discussed in this article, along with current solutions. Furthermore, these IoT attacks
are characterized based on the attack’s vulnerability to compromise the network.
Since IoT employs a network architecture that is similar to traditional network archi-
tecture for communication among various devices, it inherits the flaws of conven-
tional network architecture. To combat these attacks, researchers have suggested
a variety of solutions. Implementing all of these protection measures and strate-
gies at the same time consumes computer computation and battery power, which is
incompatible with IoT technology and its devices.
The Internet connects us to the actual world through individual medical sensors,
vicinity systems, connected residences, smart automobiles, as well as automation
networks [18]. All of this cutting-edge IoT creation creates new security problems and
research gaps that must be tackled. Despite the fact that TCP/IP (Transport control
protocol/Internet protocol) is the most common Internet communication protocol,
to connect to a centralized module through which information is transmitted to
the database, IoT devices may consider a brief network communication. There are
research opportunities for developing, implementing, and testing new lightweight
algorithms to secure data in IoT networks cluster information is encrypted in the
server, on the internet, and perhaps even on smartphones. Even secured connections
will also not secure customer information if an attacker acquires accessibility of the
database through the use of the internet, storage, or portable devices. The safety of IoT
secret information, on the internet, and even on smartphones is a difficult problem
to solve. The use of artificial intelligence and machine learning to safeguard IoT
Machine Learning-Based Security Solutions for Healthcare … 657

devices is a relatively latest field of research. Another growing subject is network

automation and intent-based networking, in which AI-based and ML-based tech-
nology handle network access and components. To protect your network, you can
use network function virtualization (NFV) and software-defined networking (SDN).
Usage of intend-based networking in conjunction with SDN for IoT devices as well
as security is still a work in progress.
Authors in [19] presented a cellphone established medical procedure for the
Healthcare 4.0 period in this study. Patients may self-identify and use personal smart-
phones and smartwatches, generating a communication channel that can be used
across owned devices. Following mutual authentication, the web gateway checks
individual clients. The AVISPA tool is used to describe the approach’s security. They
investigated the inherent security and performance qualities that improve health care
and calculated the associated connection as well as computing expenses. The find-
ings indicate the proposed methodology delivers higher protection than competing
cutting-edge alternatives.
In [20], the authors designed a Blockchain-based architecture for the effective
storage and management of electronic health records (EHRs). This also allows
patients, providers, and third parties to have safe and efficient access to medical data
while protecting the patient’s private information. The objectives of this study are to
examine how their suggested framework serves the demands of patients, providers,
and third parties, as well as to comprehend how the framework addresses security
and safety problems in healthcare 4.0 (Table 2).

Table 2 A brief comparison of the works described in Subsection 4.2

Author Objective Merits Demerits
[14] To conduct a comprehensive Theoretically discussed No practical examples of
exploration of the security importance of safety of types of attacks on these
and safety elements of mHealth gadgets gadgets
mHealth gadgets
[15] Assuring the safety of A systematic analysis of Because of the nature of
healthcare services and patterns, dangers, and this research, only English
patient data solutions for electronic papers were included for
healthcare assessment
[16] To investigate cyberspace Insisted on setting bare No discourse on healthcare
security concerns in minimum security security solution
healthcare industry standards for healthcare
gadgets
[17] Examines numerous IoT In-depth survey of Lacks discussion on usage
attacks, comparison and attacks on IoT and of techniques such as ML,
their defenses comparing them on basis blockchain, etc.,
of their effectiveness and
severity of damage
(continued)
658 P. Arora et al.

Table 2 (continued)
Author Objective Merits Demerits
[18] To highlight the problems Security solutions using Didn’t mention about
and open concerns around emerging technologies implementation of these
IoT security are discussed techniques in IoT
[19] Offer a mobile-based Suggested methodology The topic of privacy is not
solution for protecting ensures a high level of discussed in depth
electronic medical records security while incurring
lower computing and
transmission costs
[20] To present a Discussed framework –
Blockchain-based solution and it’s implementation
for effective EHR storage for preserving security
and management and privacy of all related
to health care

5 Conclusion and Future Scope

We provided an overview of IoT technological advances in healthcare in this paper.

A number of research challenges have been identified, with the expectation that they
will become major research trends in the coming years. We would like to highlight
the importance of protecting IoMT gadgets and medical technologies as a whole
(hereafter referred to as IoMT systems) is, however, a significant challenge. IoMT
systems are distinct from other systems as they have the ability to impact patients’
lives and raise privacy issues if their identities are disclosed. We hope that this study
will be beneficial for researchers and professionals in the field, assisting them in
grasping the tremendous potential of IoT in the medical domain and recognizing
major IOMT challenges. Based on the IoMT system architecture shown in Fig. 2, as
future work, we are developing a security framework to strengthen the security of
the IoMT system between the Gateway Layer and the Cloud Layer. Our framework
will be composed of an intrusion detection system (IDS) based on machine learning
algorithms.

References

1. Godi B, Viswanadham S, Muttipati AS, Prakash Samantray O, Gadiraju student SR (2020)

E-healthcare monitoring system using IoT with machine learning approaches. In: 2020 Inter-
national conference on computer science, engineering and applications (ICCSEA). IEEE,
Gunupur, India, pp 1–5. https://doi.org/10.1109/ICCSEA49143.2020.9132937
2. Durga S, Nag R, Daniel E (2019) Survey on machine learning and deep learning algorithms
used in internet of things (IoT) healthcare. In: 2019 3rd International conference on computing
methodologies and communication (ICCMC). IEEE, Erode, India, pp 1018–1022. https://doi.
org/10.1109/ICCMC.2019.8819806
Machine Learning-Based Security Solutions for Healthcare … 659

3. Qayyum A, Qadir J, Bilal M, Al-Fuqaha A (2021) Secure and robust machine learning for
healthcare: a survey. IEEE Rev Biomed Eng 14:156–180. https://doi.org/10.1109/RBME.2020.
3013489
4. Joyia GJ, Liaqat RM, Farooq A, Rehman S (2017) Internet of medical things (IOMT): appli-
cations, benefits and future challenges in healthcare domain. JCM. National University of
Sciences and Technology, Islamabad, Pakistan. https://doi.org/10.12720/jcm.12.4.240-247
5. Ghubaish A, Salman T, Zolanvari M, Unal D, Al-Ali AK, Jain R (2020) Recent advances in
the internet of medical things (IoMT) systems security. IEEE Internet Things J 1–1. https://
doi.org/10.1109/JIOT.2020.3045653
6. Mawgoud AA, Karadawy AIE, Tawfik BS (2020) A secure authentication technique in internet
of medical things through machine learning. 1538814 Bytes. https://doi.org/10.6084/M9.FIG
SHARE.13311479.V2
7. Hathaliya JJ, Tanwar S (2020) An exhaustive survey on security and privacy issues in Healthcare
4.0. Comput Commun 153:311–335. https://doi.org/10.1016/j.comcom.2020.02.018
8. Medical Device Radiocommunications Service (MedRadio), https://www.fcc.gov/medical-dev
ice-radiocommunications-service-medradio. Last Accessed 13 Apr 2021
9. Shailaja K, Seetharamulu B, Jabbar MA (2018) Machine learning in healthcare: a review. In:
2018 Second international conference on electronics, communication and aerospace technology
(ICECA). IEEE, Coimbatore, pp 910–914. https://doi.org/10.1109/ICECA.2018.8474918
10. Sisodia D, Sisodia DS (2018) Prediction of diabetes using classification algorithms. Procedia
Comput Sci 132:1578–1585. https://doi.org/10.1016/j.procs.2018.05.122
11. Wan J, AAH Al-awlaqi M, Li M, O’Grady M, Gu X, Wang J, Cao N (2018) Wearable IoT
enabled real-time health monitoring system. EURASIP J Wireless Commun Netw 298. https://
doi.org/10.1186/s13638-018-1308-x
12. Hung C-Y, Chen W-C, Lai P-T, Lin C-H, Lee C-C (2017) Comparing deep neural network
and other machine learning algorithms for stroke prediction in a large-scale population-based
electronic medical claims database. In: 2017 39th Annual international conference of the IEEE
engineering in medicine and biology society (EMBC). IEEE, Seogwipo, pp 3110–3113. https://
doi.org/10.1109/EMBC.2017.8037515
13. Hathaliya J, Sharma P, Tanwar S, Gupta R (2019) Blockchain-based remote patient monitoring
in healthcare 4.0. In: 2019 IEEE 9th International conference on advanced computing (IACC).
IEEE, Tiruchirappalli, India, pp 87–91. https://doi.org/10.1109/IACC48062.2019.8971593
14. Bhuyan SS, Kim H, Isehunwa OO, Kumar N, Bhatt J, Wyant DK, Kedia S, Chang CF, Dasgupta
D (2017) Privacy and security issues in mobile health: current research and future directions.
Health Policy Technol 6:188–191. https://doi.org/10.1016/j.hlpt.2017.01.004
15. Coventry L, Branley D (2018) Cybersecurity in healthcare: a narrative review of trends, threats
and ways forward. Maturitas 113:48–52. https://doi.org/10.1016/j.maturitas.2018.04.008
16. Buzdugan A (2020) Integration of cyber security in healthcare equipment. In: Tiginyanu I,
Sontea V, Railean S (eds) 4th International conference on nanotechnologies and biomedical
engineering: proceedings of ICNBME-2019, Sept 18–21, 2019, Chisinau, Moldova. Springer
International Publishing, Cham, pp 681–684. https://doi.org/10.1007/978-3-030-31866-6
17. Deogirikar J, Vidhate A (2017) Security attacks in IoT: a survey, vol 6
18. Sadique KM, Rahmani R, Johannesson P (2018) Towards security on internet of things: appli-
cations and challenges in technology. Procedia Comput Sci 141:199–206. https://doi.org/10.
1016/j.procs.2018.10.168
19. Hathaliya JJ, Tanwar S, Evans R (2020) Securing electronic healthcare records: a mobile-based
biometric authentication approach. J Inform Secur Appl 53:102528.https://doi.org/10.1016/j.
jisa.2020.102528
20. Vora J, Nayyar A, Tanwar S, Tyagi S, Kumar N, Obaidat MS, Rodrigues JJPC (2018) BHEEM:
a blockchain-based framework for securing electronic health records. In: 2018 IEEE Globecom
Workshops (GC Wkshps), pp 1–6. https://doi.org/10.1109/GLOCOMW.2018.8644088
Fake News Detection Techniques:
A Survey

Sahitya Diwakar, Divyansh Bansal, Raghav Jagya, Poonam Tanwar,

and Shweta Sharma

Abstract Lately, on account of the thriving development of online relational asso-

ciations, fake news for various businesses, what is more, political purposes have
been appearing in tremendous numbers what is more, extensive in the online world.
The development of beguiling information in standard access news sources, for
instance, electronic media channels, news web diaries, and online papers have made
it attempting to perceive reliable news sources, therefore growing the prerequisite for
computational contraptions prepared to give pieces of information into the faithful
nature of online substance ( Poovaraghan et al. (2019) Fake news accuracy using
naive bayes classifier. Int J Recent Technol Eng ). This paper considers past and
current systems for fake news distinguishing proof in text-based associations while
determining how and why fake news exists regardless.

Keywords Fake news · News media · Social community

1 Introduction

As most of the time in our lives is spent interacting online through social media
platforms like Instagram, Facebook, more and more people around the world prefer
to see news via social media rather than traditional news platforms. The reason
for this change in platforms is easy to understand as it is faster and cheaper to
spread the news through social media rather than news platforms like newspapers
and television. Moreover, it is easy to debate some news on social media among other
readers and discuss it. In the US 62% of readers get news from social media in 2016,
while in 2012 only 49% of readers got news from social media [1]. It is also said
that social media now outperforms television as the major news source. However,
despite the advantages of social media, the quality of news on social media is lower
than traditional news organizations because it is cheap to provide news online and

S. Diwakar · D. Bansal · R. Jagya · P. Tanwar (B) · S. Sharma

Computer Science and Engineering, FET, Manav Rachna International Institute of Research and
Studies, Haryana, Faridabad, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 661
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_48
662 S. Diwakar et al.

much faster and easier, large volumes of fake news are produced online for a variety
of purposes.

18% Never
38% Often
Sometime
26%
Hardly ever
18%

News Use Across Social Media Platforms 2016. Source Survey conducted Jan 12 Feb, 2016

The extensive spread of fake news can have a serious negative impact on individ-
uals and society. It has a major impact on locals who have no idea how to distinguish
between fake or real news. For example, the most popular real news spread through
television did not spread as widely as the most popular Fake news was spread on
Facebook. Fake news is mostly manipulated by protagonists with a hidden agenda
of conveying some political message. Fake news also changes how people see real
news. To help reduce the negative impact caused by fake news people must know
how to differentiate between fake or real news and we must develop algorithms that
automatically detect fake news on social media. Detecting fake news on social media
opens a variety of challenges and research problems which make it very challenging
for automated detection. Fake news is written with an intention of confusing readers
and misleading them, which makes it not so simple to detect fake news based on its
content.
The key motivation of this survey is that Fake news detection on social media is
still at the early age of development, and there are still many challenging issues that
need further attention. It is necessary to discuss potential research directions that can
improve fake news detection and its capabilities. To guide the future of fake news
detection research, appropriate clarifications are necessary. We give an overview of
three existing fake news detection methods Style-based, Content-based, Naïve Bayes
Classifier in detail from various research till now.
Fake News Detection Techniques: A Survey 663

2 Literature Survey

This research seeks to bring together a series of similar works in the field of social
media identification of fake news. As a result, I conducted a survey using different
platforms like Facebook, Twitter, and others. To determine the potential of worldly
experiences to be relied upon. Style-based methods focus on capturing the writing
style of the news content and try to classify articles based on it [2]. Most of the
works were based on the fact that there were several similarities between fake news
and spam emails like both having a huge number of grammatical mistakes, trying to
manipulate the reader’s opinions, and using a somewhat similar word set. Although
these approaches can be very effective at times, they fail to explain what exactly is fake
in the news content. While on the other hand, content-based methods for detecting
fake news, also known as fact-checking [3], take in use of the content of fake news
rather than the style. It is the most straightforward way and aims at checking the truth
value of the claims in the news content. The contemporary approaches majorly focus
on finding a path in an existing knowledge graph for a certain triple. We will focus
more on content-based approaches in this paper since it looks more promising in the
current status quo. Because there has been a significant growth in the distribution of
fake news via social media, there have been changes in the percentage of efficiency
of algorithms in detecting fake news. Changes are good for some approaches and
negative for others. For example, the Naive Bayes method has an efficiency of 0.92,
which is 80% efficient in the market, while others have an efficiency of 0.73–0.82,
which is 60% effective [4].

2.1 Domain

We learn from the papers that there are several social media domains, such as Face-
book (Tucker J and Savyan), Twitter, and others. People will spread false news using
this method. Automatic real-time monitoring of harmful posts on Facebook and using
opinion mining to consider user activity on Facebook.

2.2 Dataset

The datasets ( Zhou and Monti) that were used to propose new models that aid in the
analysis of false news included Snopes, PolitiFact, BuzzFeed, and many others. For
the identification of false news in their respective works, datasets from models such
as the PHEME dataset [5], BuzzFeed [2], Epinions, and others were gathered.
664 S. Diwakar et al.

2.3 Issues

Owing to shortcomings in the databases and the length of the news story, the key prob-
lems illustrated include difficulty in classification accuracy with inaccurate content
on social media. The article would demonstrate how the model failed to work well
in a particular data case, but it will not say that their dataset is representative of the
Facebook population (p. Diwan).

3 Previous Method

Different types of methods to detect fake news:

• Fake News Detection from Style Based
• Fake News Detection from Content-Based
• Naïve Bayes Classifier

3.1 Fake News Detection from Style Based

The style-based fake news detection method analyses the news report based on the
writing style of people who spread fake news. In this method, it is easily detectable
that is this news is fake or not because in spam emails and fake news there are so
many grammatical errors that are used to manipulate the ideology and thinking of
people. Fake news authors used to use self-oriented pronouns rather than oriented
pronounce more sensory-based words [6]. It generally uses deep syntax, data repre-
sentation, sentiment analysis, and semantic analysis to understand fake news. The
main assumption is that the fake news author has no previous experience regarding
an event that is why they end up contradicting themselves or may leave important
facts which are very important for the investigation [6]. But the biggest problem with
this method is it does not tell us which part of the news is fake.
Fake news teller is generally using unethical plans to spread useless baseless fake
news by misdirecting the data and impact enormous networks of buyers, requiring
specific composing styles important to interest and convince a wide extent of shoppers
that is not found in evident news stories. Style-based methods can easily identify fake
news by identifying the composition style of news. Style-based method is further
divided into two main categories: Deception-arranged and Objectivity situated.
Stylometric strategies of fraud type capture articulation or beguiling cases from
news content. Inspiration of the pre-signing location begins from legal brain research
(i.e., the undeutsch hypothesis) [7], and various criminological instruments including
content-based content investigations [8] and scientific-based content analysis (Tony
Lesce, 1990) has been made. Even more than the end-model of late normal language
Fake News Detection Techniques: A Survey 665

handling, developing applied at the duplicate spot stage from the perspective accom-
panying: deep grammar and rhetorical construction. The profound grammar model
has been done utilizing the probabilistic settings for free grammar (PCFG), with
which sentences can be converted into decisions that describe the grammar struc-
ture. Considering PCFG, various guidelines can be produced for scaly locations,
for example, the rules of creation that are not lexicalized and grandparents’ rules
[9]. The expository design hypothesis can be used to capture the contrast between
misleading and honest sentences [10]. In-depth organizational models, for example,
convolutional nerve organizations (CNN), have also been applied to the truth of fake
newsgroups [11].
• Methodologies arranged to objectivity the signs of capture style that can demon-
strate the decrease in the objectivity of the content of the information and, therefore,
the possibility of deceiving buyers, such as Hyper Partisan styles and sensationalist
reports. Hyper partisan styles address an outrageous behavior for a specific ideolog-
ical group, which corresponds regularly with solid inspiration to make counterfeit
news. Featured based on linguistics can be applied to distinguish hyper partisans
[12]. Sensationalist reports address those items that do not contain well-informed
news, but depending on the characteristics of obtaining eyes (ie, deceptive content)
with an inclination for beautification, sensationalization, alarm achievement, etc.
Frequently, the news titles will summarize the significant prospects of the article that
the writer needs to transmit, and consequently deceptive and cheated the deceptive
content titles can complete as a decent marker to recognize counterfeit items [13, 14].

3.2 Fake News Detection from Content-Based

The content-based method of finding fake news is not like the style-based method.
As in the style-based method, it does not check the styling, but it checks the content
of the news report. It checks the content of the news report with previous research,
report, content and checks the knowledge graph, and then tells if the news is fake or
not. But here is also a problem that if there is no knowledge graph available on the
internet or there are no previous reports on the internet.
There are some key features that tell us about the basic language and sentence
structure of news. The semantic features of the content are acceptable pointers and
help to discover suspicious news, which is focused to change the mindset of the
reader. The writing style feature tells us different writing methods of the author of
Counterfeit news to change the mindset of the reader [15].
Counterfeit news makers expect to emulate a similar method of composing of
genuine news writers to delude the pursuers and urge them to accept their cases
yet, there are a few loopholes that are still left out. These loopholes help to help
to differentiate between the genuine news writer and fake news writers such as the
use of several syntactical structures and keystroke features that gives clarity on the
sentence structure of the news, we cannot predict the semantics of the sentences from
666 S. Diwakar et al.

the writing style which plays an important role in recognizing nature of the news
(fake or real) [16].
Semantic features are vital highlights which are used to uncover fake news. There
are few studies which had used this semantic feature to detect counterfeit news [17].
Results of studies have shown that after using this semantic feature we can easily
recognize counterfeit news, but it needs more hands-on and examination [18].
It is a necessity to use semantic features in counterfeit news detection, just a couple
of research have used these highlights in the research [17]. The aftereffects of the
research have shown that utilizing semantic features to recognize counterfeit news
is vital and successful and wants more examination [18]. To uncover the fake client
accounts and suspicious clients, sentimental features are a way to deal [19]. Authors
who write about these kinds of news generally overdraw the facts and mislead the
reader. With the help of sentiments, it is easy to detect false and the psychological
keywords can uncover the feelings, judgements [20, 21]. For the utilization of the
sentiment of news so many methods have been proposed some of them are arousal
valence, happiness detection and analysis, and many more [15, 22].
It has been demonstrated as an adequate technique to identify counterfeit news
by joining a few scores of sentiments and can recognize and can differentiate fake
bot records and genuine human records [19]. For the analysis of visual content such
as images, videos, popular images, multi-image frequency, and so on, the visual-
based feature analysis is proposed. These are the cues for detecting fake information
[3, 16, 14, 23, 24, 25].

3.3 Naïve Bayes Classifier

This method is proposed from the mathematics theorem that is known as the Bayes
theorem. This theorem is used to calculate the “probability that something will
happen, given that something else has already occurred” [2]. It is a machine learning
class, so it is predicted by “membership probabilities” and for each record, or data
that is belong to that one class. And the class with maximum chances is considered as
“most likely class” [23]. In many cases, this method was proven wrong because some-
times when two things are dependent on each other it takes them to be independent
of each other. Let us say there is news related to Donald Trump but if in the scanning
of Naïve Bayes classifier, it reads Donald in many cases and the same article there
could also have Trump, which Naïve Bayes classifier treat as two different people
which change the result of news flakiness. So, the biggest flaw of method is that it
considers every one of the features to be discrete, which may not generally be the
situation [24].
Naive Bayes classifiers are part of straightforward machine learning. For checking
whether any news is fake utilizing pipelining ideas, or it is genuine Naïve Bayes is
a well-known method. There are many calculations that zero in on regular standard,
so that is not the solitary calculation for preparing such classifiers. For checking
whether the news is fake or genuine Naive Bayes can be utilized. Naive Bayes
Fake News Detection Techniques: A Survey 667

classifiers are part of straightforward machine learning. Naive Bayes is well-known

calculation which is utilized to discover the exactness of the news regardless of
whether its genuine or phony utilizing multinomial NB and pipelining ideas. There
are many calculations that zero in on basic standard, so Naïve Bayes is not the only
algorithm for preparing such classifiers. This method is derived from the mathematics
theorem which is known as the Bayes theorem. This theorem is used to calculate
the “probability that something will happen, given that something else has already
occurred” [6].
Naïve Bayes is a probabilistic classifier based on independent classification of
features.
P( B|A).P( A)
P( A|B) = (1)
P(B)

These Naïve Bayes classifiers are the pairs of clear probabilistic classifiers that
depend on applying Bayes theory with strong independence doubts between the
features. For the development of models that are used to appoint the class marks
to issue occasions, in which the name of class is taken by some limited sets, this
method is used [2]. It is a singular computation for the preparation of these kinds
of classifiers, however, the pairs of calculations are dependent on a regular rule: all
Naive Bayes classifiers are expected to be that the approximation of the specified
component is undertaken of another component where the class variable is given.
Naive Bayes classifiers are a mainstream measurable procedure of email filtering.
They arose in the 90 s and were one of the principal endeavors to handle spam
separating issues. Naive Bayes normally use pack of words highlights to distinguish
spam email; a methodology regularly utilized in text order. Innocent Bayes Classifiers
works by connecting the syntactic or not syntactic development with the spams and
without spam messages and after applying this method it figures out whether the
email is a spam message or not.
As it is a part of machine learning class, it can be performed by predicting “mem-
bership probabilities” for every class, data, or any of the record that belongs to that
class. The class which has the maximum chances shall be determined as the “most
likely class” [6]. In many cases, this method was proven wrong because sometimes
when two things are dependent on each other it takes them to be independent of each
other. Let us say there is news related to Donald Trump but if in the scanning of Naïve
Bayes classifier, it reads Donald in many cases and the same article there could also
have trump, which Naïve Bayes classifier treat as two different people which change
the result of news flakiness [24]. This method does not always classify the data, so
there is a higher risk of matching with unrelated data, which ultimately affects the
reliability of the news [26]. The advantages and disadvantages of all are shown in
Table 1.
668 S. Diwakar et al.

Table 1 Comparative analysis of Fake news detection methods

S.no Methods Advantages Disadvantages
1 Fake News Detection from This process studies the This process doesn’t study
Style Based writing style of the author the content of the author and
and then gives results verify from online sources
It generally uses deep and then gives results. This
syntax, data representation, process crashes when there is
sentiment analysis, and no previous papers or data on
semantic analysis to the internet regarding a
understand fake news particular topic
2 Fake News Detection from This process studies the This process crashes when
Content-Based content of the author and there is no previous papers or
verifies from online sources data on the internet regarding
and then gives results a particular topic
3 Naïve Bayes Classifier This process involves Sometimes in this domain
counting the number of people use hybrid cloud
times a word appears in the algorithm to detect fake
headline or stanza, news, which is less accurate
converting that to a and required large amount of
probability, and then space. Sometimes, it takes
calculating the probability human input
that it is a fake or not

4 Challenges

News is always provided by some journalist or blog writer via websites or articles.
The problem arises in 3 different ways:
• Misinformation—This is the interface between worlds and journalists.
Disinformation- This is the interface between journalist and their content.
• Misleading—This is the interface between content provided by journalists and
readers or viewers.
Misinformation arises when a journalist does not have a clear understanding of the
target system that he or she is discussing. Misinformation is characterized by a lack
of veracity, that is, Information about the target system is not accurately conveyed
in the journalist’s message. Fake news writer uses minor mistakes done by honest
journalists to create a big issue.
Disinformation is a bit different from misinformation, disinformation is character-
ized by the journalists’ wrong intentions and arises when a journalist wants to change
the mindset or mislead the audience via their message Wardle [27]. A journalist like
this promotes false information, perhaps even true information that implies some-
thing false. Disinformation consists of deception of both the content and authorship.
They can, for example, present content as though it were the product of genuine
journalistic practices when, in fact, it was not [28, 29]. Misleading content is char-
acterized by its ability to mislead the audience. News can an audience for several
reasons. A vague or omitted article may mislead the audience. Sarcastically writing,
Fake News Detection Techniques: A Survey 669

for example, misleads readers who do not realize the author was trying to entertain
rather than informing them about something [30].

5 Conclusion

In this paper, we have examined the fake news story, the maker also, subject identifi-
cation issue. Because of the news increased heterogeneous informal organization, a
bunch of expressed and dormant highlights can be removed from the text-based data
of information articles, makers, and subjects individually. Fake news interferes with
the capacity of a client to observe valuable data from the internet benefits particu-
larly when news gets basic for dynamic. Thinking about the changing scene of the
advanced business world, the issue of fake news has gotten something other than
a showcasing issue as it warrants genuine endeavors from security scientists. Fake
news identification has many open issues that require the consideration of scien-
tists. For example, to lessen the spread of fake news, distinguishing key components
associated with the spread of information is a significant advance.

References

1. Gottfried J, Shearer E (2016) News use across social media platforms

2. Granik M, Mesyura V (2017) fake news detection using naive bayes classifier. UKRCON IEEE
3. Shu K, Sliva A, Wang S, Tang J, Liu H (2017) Fake news detection on social media: a data
mining perspective. Computer Science & Engineering, Arizona State University, Tempe, AZ,
USA
4. Sharma U, Saran S, Patil SM (2021) Fake news detection using machine learning algo-
rithms. Department of Information Technology Bharati Vidyapeeth College of Engineering
Navi Mumbai, India, IJERT
5. Zhang X, Ghorbani AA (2020) An overview of online fake news: characterization, detection,
and discussion. Inform Process Manage
6. Stahl K (2018) Fake news detection in social media. In: Candidate BS (ed) Department of
Mathematics and Department of Computer Sciences, California State University Stanislaus, 1
University Circle, Turlock, CA 95382
7. Undeutsch U (1967) Beurteilung der glaubhaftigkeit von aussagen. Hand- buch der psychologie
8. Vrij A (2005) Criteria-based content analysis: a qualitative review of the first 37 studies. Psychol
Public Policy Law
9. Wang WY (2017) Liar, liar pants on fire: a new benchmark dataset for fake news detection.
arXiv preprint
10. Potthast M, Kiesel J, Reinartz K, Bevendorff J, Stein B (2017) A stylometric inquiry into hyper
partisan and fake news. arXiv preprint
11. Rubin VL, Lukoianova T (2015) Truth and deception at the rhetorical structure level. J Assoc
Inform Sci Technol 66(5):905–917
12. Chen Y, Conroy NJ, Rubin VL (2015) Misleading online content: recognizing clickbait as false
news. In: Proceedings of the ACM on workshop on multimodal deception detection, ACM
13. Saxena R (2017) How the naive bayes classifier works in machine learning. Retrieved 2017
14. Kaustav et al (2020) A facial expression recognition system to predict emotions. In:
International conference on intelligent engineering and management (ICIEM)
670 S. Diwakar et al.

15. Zhang J, Dong B, Philip SY (2019) FAKEDETECTOR: effective fake news detection with deep
diffusive neural network. IFM Lab, Department of Computer Science, Florida State University,
FL, USA BDSC Lab, Department of Computer Science, University of Illinois at Chicago, IL,
USA
16. Hussein A, Ahmad F, Kamaruddin S (2019) Content-social based features for fake news
detection model from Twitter. Int J Adv Trends Comput Sci Eng
17. Jia S et al (2018) Fake reviews detection based on LDA. In: 4th International conference on
information management (ICIM). IEEE
18. Xu K et al (2019) Detecting fake news over online social media via domain reputations and
content understanding. Tsinghua Sci Technol
19. Dickerson JP, Kagan V, Subrahmanian V (2014) Using sentiment to detect bots on twitter: Are
humans more opinionated than bots? In: IEEE/ACM international conference on advances in
social networks analysis and mining
20. Thakur P, Shrivastava DR (2018) A review on text based emotion recognition system. Int J
Adv Trends Comput Sci Eng
21. Zhao J et al (2014) FluxFlow: visual analysis of anomalous information spreading on social
media. IEEE Trans Vis Comput Graphics
22. Oueslati O, Khalil AIS, Ounelli H (2018) Sentiment analysis for helpful reviews prediction.
Int J
23. Tanwar P, Rai P (2020) A proposed system for opinion mining using machine learning, NLP
and classifiers. IAES Int J Artif Intell (IJ-AI) 9(4):726–733, ISSN: 2252-8938. https://doi.org/
10.11591/ijai.v9.i4.pp726-733
24. Petkar PB, Sonawane SS (2020) Fake news detection: a survey of techniques. Int J Innov
Technol Exploring Eng (IJITEE)
25. Konagala V, Bano S (2020) fake news detection using deep learning. IGI Global
26. Poovaraghan RJ, Priya MK, Vamsi PS, Mewara M, Loganathan S (2019) Fake news accuracy
using naive bayes classifier. Int J Recent Technol Eng
27. Wardle C, Derakhshan H (2018) Thinking about ’information disorder’: formats of misinfor-
mation, disinformation and mal-information. In: Ireton C, Posetti J (eds) Journalism, ’fake
news’, & disinformation: handbook for journalism education and training
28. Pepp J, Michaelson E, Sterkin RK (2019) What’s new about fake news? J Ethics Social Philos
29. Fallis D, Mathiesen K (2019) Fake news is counterfeit news. Routledge Taylor and Francis
group
30. Garrett RK, Bond R, Poulsen S (2019) Too many people think satirical news is real. The
Conversation Academic rigour, journalistic flair
Impact of Image Classification
in Dermatology: Automated Diagnosis

Anukriti Singh, Devam Kumar Sinha, and Namrata Dhanda

Abstract The implementation of AI and ML in health care sector is majorly in

cardiology, radiology, and ophthalmology but its employment scope in dermatology
is minimal. Early diagnosis of skin diseases may prevent the disease to grow worse
in condition. Machine-Learning has the caliber to improvise the dermatologist’s
practices by making the diagnosis process less tiresome and enable people to get a
more personalized treatment through the use of mobile applications etc. Numerous
technologies based on machine learning algorithms have emerged in different areas
of medicine. Dermatological diseases are usually easily visible and diagnosis are
carried on by visual screening of the abrasion, recognizing, and comparing with the
patterns of known skin diseases. Three commonly occurring skin diseases namely
Acne Vulgaris, Lichen Planus, and SJS (Stevens-Johnson syndrome) & TEN (toxic
epidermal necrolysis) have been studied and worked upon in this paper. The purpose is
to classify correctly these three diseases by implementation of machine-learning algo-
rithms such as SVM, CNN, logistic-regression, Random-Forest, and naïve-Bayes.
After analysis, it was found that CNN gives the highest accuracy among the discussed
models in the prediction of the disease.

Keywords Dermatologist · Lesions · Visual Screening · SVM · CNN · Logistic

regression · Random forest · Image classification

1 Introduction

Being the largest organ of the body with an approximate area of 20 square feet, skin
is vital to produce the sense of touch, heat, cold, regulates the temperature of the
human body, and safeguards us from microbes present in the environment. Skin is
susceptible to both external and internal conditions.
As it protects the vital organs of the body from harmful pathogens, its proper care
is extremely crucial. The skin is exposed to an environment full of pathogens is prone

A. Singh (B) · D. K. Sinha · N. Dhanda

Amity University Uttar Pradesh Lucknow, Lucknow, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 671
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_49
672 A. Singh et al.

to various kinds of diseases which may range from something as harmless as acne
to something as dreadful as skin cancer.
The dermatologist is required to track the symptoms accurately and create a proper
diagnosis as any discrepancy in judgment regarding critical skin diseases might prove
deadly for the patient. Thus, an efficient approach must be established for the clear
diagnosis of the symptoms of skin diseases in the early stage. Since 1955, AI has
been emerging at a fast pace, revolutionizing every major field. Artificial intelligence
and machine learning cater to the need for automation, deeper insights, predictive
analysis, and increased accuracy.
With this fast-paced advancement in AI tech, skin diagnosis procedures can be
technically designed for the initial diagnosis of skin infections. A variety of innova-
tive ideas can be accessed for imagery and pattern-based detection for various skin
diseases [1]. Since the employment of machine learning in the field of dermatology
is still relatively new, this research aims at introducing the readers to the prospective
technological methods driven by machine learning to obtain a more accurate and
easily accessible diagnosis.
Machine learning can be a benefactor in the early identification of distinguished
classes of skin infections. Skin disease symptoms are easily visible and often exhibit
patterns. Through image classification, these diseases might be categorized and diag-
nosis can be fast-tracked and be made more personalized. Image-classification is
a supervised-learning model in which a bunch of objective classes is defined and
models are trained for the detection of correct class [2]. For the emerging technolo-
gies to be successful in the desired motive, it is required that medical professionals
have a basic understanding of machine learning. The research provides a comparative
study of different algorithmic models to find the most accurate image classification
model.

2 Literature Review

The inter-merge and relation development between the technology and health-care
services have led to frequent and rapid development in the field of image processing
methods for the benefits and progress in the field of medicines. For an accurate
diagnosis, many applications of digital image-based mechanisms like Computed
Tomography (CT), Magnetic-Resonance-Imaging (MRI) and Digital-Subtraction-
Angiography (DSA) are in use and have gained a lot of popularity.
Many organizations are working to build more efficient mechanisms for bene-
fiting the healthcare field with modern technical aspects. Quite a few researchers
have done work to detect skin diseases through different algorithms. A vivid litera-
ture survey is mentioned as follows. Ercal et al. [3] adapted efficient color metrics
using the RGB planes. That serves in differentiating the tumor from its background.
Image-segmentations are carried out using the appropriate coordinate transforma-
tions. Boundary shadows are sketched out by taking out the tumor portion of the
classified image. That’s a very efficient method to detect tumors while diagnosis.
Impact of Image Classification in Dermatology: Automated Diagnosis 673

Sigurdsson et al. [4] arrayed skin laceration using the in vitro Raman spectroscopy.
This technique made use of a nonlinear-neural-network classifier for the work.
Distinctive brands of diversity express unambiguous lipids & proteins that give
the facts and figures and relevant facts for the detection of skin lesions. Aberg [5]
applies electrical bio-impedance for the assessment of skin cancers and lacerations.
Multi-frequency impedance spectrum is made use to differentiate the skin-cancer
and benign-nevi. Wong et al. [6] put forward a novel looping stochastic-region
merging technique for segmentation of skin lesion infected area out of the macro-
scopic images. Initially, in the approach mentioned, stochastic-region merging is
applied on the pixel level.
After this pixelated level, it upgrades to the region level till the convergence
takes place. Demyanov et al. [7] made use of deep-convolutional-neural-networks,
for the image-classification algo combined with data-augmentation for the successful
investigation to autodetection of the dermoscopic-patterns and skin abrasion analysis.
Grana [8] gave a proficient mathematical perspective for the lesion border assess-
ment. This method takes illumination values alongside a normal way to any contour
at any instance point in consideration. Sigurdsson and Wighton [9] published auto-
mated skin lesion diagnosis. It’s a model-based on the usage of supervised learning
and makes use of MAP-estimation for portraying the results for identification. Emre
Celebi et al. [10] make use of ensembles from the thresholding techniques for the
detection of lesion boundaries in dermoscopy pictures. Ganster et al. [11] devised a
computer-based system for analyzing images that are acquired via ELM.
General segmentation algorithms using fusion approach are applied to obtain the
binary mask of skin lesion. The danger of injury is determined dependent on the
shape and radiometric highlights. The neighborhood and worldwide boundaries are
additionally taken under consideration for better outcomes. The framework serves
to improve the early recognition of harmful melanoma. The procedure determined
psoriasis vulgaris to have high affectability and particularity. The strategy determined
psoriasis vulgaris to have high affectability and particularity. An inventive approach
for auto-segmentation and classification of skin lesions was published by Sumithra
et al. [12].
The discussed approach makes use of SVM and k-N neighbor algorithm for the
detection of lesion. Lu et al. [13] utilize 2D computerized picture division and read-
justing the size to arrange smooth-pixels consolidating the discussed procedures with
Markov arbitrary field_(MRF). A solid division procedure is set up. Salimi et al. [14]
differentiated various skin diseases with the use of a pattern recognition technique.
Kolkur et al. [15] presented a novel skin identification calculation that improves
the discovery of skin pixels that includes RGB, HSV, and YCbCr shading models.
Kotian and Deepa [16] read auto conclusion framework for skin illness. Strategies,
for example, picture line ID and highlight information mining are carried out utilizing
Matlab programming. Oyola and Arroyo [17] gathered and characterize a picture of
varicella through Hough change and applied the shading change, adjustment, and
edge discovery procedures of picture handling. It helps in a better conclusion of
varicella identification. Hung and Sapiro [18] recommended a technique for skin
674 A. Singh et al.

injury recognition based on a halfway differential condition. In view of the morpho-

logical sifting through PDE, a form model of sores was figured out, that helps in
precisely recognizing the illness. 3-dimensional computed tomography (CT) imago
logical technique was applied by Zhong et al. [19]. Kumar and Singh [20] related
skin-cancer imagery across various types of neural networks. A bunch of skin cancer
images was trained then tested using MATLAB. This helps in the characterization
of skin-malignant growth.
According to research done in Medline, Embase (in December 2019), and WoS
DB of articles retaining AI & ML in the field of dermatology. Articles from 2000 to
2020 are analyzed and included to concentrate on newer and better methods. A great
deal of examination and applications are as of now trending in the field of clinical
imaging and determination. Still, we need to focus on giving skin checking logically
definite systems, minimal effort, and dependability.

3 Overview of Skin Diseases

The skin safeguards us from dangerous bacteria and other pathogens present in the
environment. It’s quite common for it to get infected and develop symptoms. The
symptoms may be harmless but they can be as life-threatening as skin cancer. The
reasons can be external environment-based or even internal because of the gene
abnormality or any hormonal misbalance.
Major studies making use of ML in dermatology create have a clear focus on
classification of skin lesions for a different disease. These studies generally use
CNN for image recognition and classification. First, an already trained CNN (i.e.,
AlexNet) was employed to take out the workings, and these characteristics were
then categorized by a simpler Machine-Learning algorithm like k-nearest neighbors
or SVMs. In the current time, most of the CNNs take out characteristics and classify
images by end-to-end learning [2]. This paper classifies aims at categorizing three
kinds of skin diseases namely lichen planus, acne and Stevens-Johnson syndrome
(SJS), and toxic epidermal necrolysis (TEN).

3.1 Acne

Acne is usually called acne-vulgaris. It is an immedicable inflaming infection. It

happens over the body parts like neck, face, upper chest and back, and so forth. As
per examines, acne is one of the most common and predominant sicknesses on the
planet. It influences the patient both mentally just as psychosocially. The irritation
of the skin happens because of the oil emitted from the sebaceous organ or that is
usually known as oil-organs of the skin [21]. Skin inflammation is rarely a perilous
situation, yet it influences the confidence of an individual. Individuals with the age
Impact of Image Classification in Dermatology: Automated Diagnosis 675

Fig. 1 Acne vulgaris

Fig. 2 Lichen planus

bunch going from 12–24 are more inclined to skin inflammation and around 85% of
the populace is being influenced [21] (Fig. 1).

3.2 Lichen Planus

Lichen planus is a genuinely unmistakable, constant, immunologically interceded

mucocutaneous sickness, of dubious etiology that was first depicted as an infection
of the skin that can likewise influence mucosal surfaces, including the oral cavity
lining. The commonness of LP in everyone is in the range of 0.9–1.2%. It is seen
that roughly 40% of lesions happen on both oral and cutaneous surfaces, 35% on
cutaneous, and 25% on mucosal surfaces alone. [22] Oral lichen planus is known to
influence all genders between the ages of 30 and 70 years [22] (Fig. 2).

3.3 Stevens-Johnson Syndrome and Toxic Epidermal

Necrolysis (SJS and TEN)

Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS-TEN) are

uncommon, possibly dangerous, extreme mucocutaneous antagonistic responses
described by broad epidermal separation, the disintegration of mucosae, and serious
indications. The medical highlights of SJS-TEN are trademark and the analysis is
clinical. The first indications usually involve fever, conjunctivitis mimicking febrile-
illness of infective origin and upper-respiratory-tract symptoms. This is accompa-
nied by the disengagement of mucous-membranes (conjunctival, anogenital, oropha-
ryngeal and nasal) [23]. Generally, more than two mucous-membranes are asso-
ciated. Cutaneous-lesions, in the appearance of dusky erythematous-purpura/flat
676 A. Singh et al.

Fig. 3 SJS and TEN

typical-atypical target abrasions, break-out with ache and burning commotion

(Fig. 3).

4 Proposed Work and Algorithms Used

In this comprehensive paper, the implementation of 5 machine-learning models that

could detect 3-kinds of skin diseases namely lichen planus, acne and Stevens-Johnson
syndrome and toxic epidermal Necrolysis (SJS-TEN) is discussed. The dataset for
this research has been extracted through Bing Downloader. To further carry out the
image classification process, the skin dataset is first preprocessed. While prepro-
cessing, we first read the images using ‘imread’ function and loaded the folders
containing the images into an array. Next, we resized them to the dimension (150,
150, 3) using ‘resize’ function. The images were segregated into three classes namely
0, 1, and 2. The dataset has been divided into the training–testing data: 80% is used
for training purpose and rest 20% is utilized for testing. Thereafter, the models are
trained using five variant classification algorithms namely naive Bayes classifier,
logistic-regression, SVM, random-forest, and CNN for the classification of three
mentioned skin diseases.

4.1 Logistic-Regression

Logistic-regression is statistical-model which utilizes employs logistic-functions to

model up a binary-dependent variable, though plenty more complicated extensions
exist. A logistic-model contains a dependent-variable which has 3-values which are
possible that are indicated via an indicator-variable, where the 3-values are labeled
as 0, 1 and 2. [24] In the logistic-model, the log-odds for the value tagged as “1”
is a linear concoction of 1 or more than 1 predictor; the independent-variables can
individually be a binary variable (2 classes, coded through the indicator variable) or
a continuous-variable (any real value). The corresponding probability for the value
tagged “1” can be valued somewhere from 0 to 1. The dependent variable in a binary-
logistic-regression model has 2-levels (categorical) [24]. Outputs that have > 2 values
Impact of Image Classification in Dermatology: Automated Diagnosis 677

are included in a model by multinomial logistic regression. The logistic-regression

model plainly models the probable outcome of output in respective of input and
doesn’t undergo statistical-classification as this isn’t a classifier, but it still can be
employed well to use as a classifier, for example by selecting a cutoff-value and
categorizing inputs which have the probability-index > the cutoff as 1 class, lesser
the cutoff as the other, that’s a different method to employ a binary-classifier.

4.2 Kernel SVM

The SVM is a supervised-learning algorithm that is usually implemented for classi-

fication although it can also be made use of for regression. The primary idea focuses
on the labeled data (training data) by which the algorithm attempts to find the optimal
hyperplane that can be employed to classify new data points. In two dimensions the
hyperplane is considered a simple line.
Generally, a learning algorithm attempts at learning the most common charac-
teristics, i.e., what distinguishes one class from another class, and the classification
is based on those representative characteristics learned, so classification is based on
differences between various classes. The SVM has its work in the other way around.
It detects the most similar examples between classes. Those will be the support
vectors.

4.3 Naive Bayes

Naive Bayes is an extremely easy approach used for the construction of classi-
fiers/models that provide class labels to problems that are presented as vectors of
feature values, in which case the class labels are extracted from the finite sets.
There isn’t any other algorithm that trains these types of classifiers, but a family-of-
algorithms follows on a common principle that each and every naive Bayes classifier
follows the assumption that the value of a specific feature isn’t dependent over the
value of any other feature.
For few kinds of probability models, naïve-Bayes classifiers can be effec-
tively trained under a supervised learning atmosphere. In real-time uses, parameter
estimation for naïve-Bayes models makes use of the way of max likelihood.

4.4 Random Forest

Random-forest, as the name infers, comprises an enormous number of individual

choice trees that operate as a group. Each tree in the random forest provides a class
prediction and the class having the largest number of votes takes the form of our
678 A. Singh et al.

model’s outcome. In the language of data-science, the primary reason for the random
forest model to work this well is that:
"A large number of relatively uncorrelated models (trees) operating as a committee will
outperform any of the individual constituent models [24]."

The low interrelation among models is the reason. Uncorrelated models are
capable of producing ensemble outcome that is more correct than any other separate
prediction outcome. The reason behind this is that the trees guard each other from
their respective errors just as long as all of them don’t persistently make errors in the
same perception.

4.5 Convolutional Neural Networks

A CNN is a Deep-Learning algorithm that accepts an image as input, allocate impor-

tance, i.e., learnable weightage and biases to different prospect and objects of the
image, and having ability to distinguish one from other. The pre-processing needed in
CNN is much less in comparison to other classification algorithms. While in primeval
methods, filters were hand-engineered, by proper training, Convolutional Networks
possess the capability to accumulate these filters/characteristics. [12] The architec-
ture of a CNN is homologous to the way of the association pattern of neurons and was
greatly motivated by the establishment of the Visual Cortex. Each neuron replies to
stimuli only in a constrained region of the ocular field known as the Receptive-Field.
A cumulation of likewise fields overlaps to enfold the entire visual area.

5 Result Analysis

We had three different classes of skin diseases namely lichen planus, acne, and
SJS and TEN in the suggested skin disease classification scenario. The data was
then segregated into these three classes. We segregated the dataset between the
training and testing dataset into the ratio of 80:20. We then employed the five models
using Sklearn and Tensorflow and obtained the below-mentioned accuracies using
r2_score metric. Among all these Naïve Bayes algorithms provides the least accuracy
percentage, counting 47.4 and 49 as in testing and training accuracy respectively. The
best mechanism that can be adopted for skin lesion detection is the convolutional
neural network, with the most efficient working percentage. Its training accuracy
is 86.77 and the training accuracy shows to be 89.05. So, the order of accuracy is
as follows: Naïve Bayes < Kernel SVM < Random Forest < Logistic Regression <
convolutional neural network (Fig. 4).
Impact of Image Classification in Dermatology: Automated Diagnosis 679

Fig. 4 Comparison of various classification models for skin disease diagnosis

6 Conclusion

Our early discussion started with brief factual information on skin diseases and their
current treatments. In the field of skin disease treatment, the most crucial and major
issue that the medical industry faces, is the detection of the lesion. The skin laceration
can be diagnosed, retrieved, and cured efficiently, but only if analyzed beforehand at a
primal stage. Literature survey clarifies that various skin disease inspection methods
are being implemented. However, there is still a dire requirement to differentiate skin
diseases at a real primary stage. Machine-learning algorithms hold the capacity to
offer quite a few impactful methods for initial detection for skin laceration. It can also
help people to take real-time measures for their skincare. If invested upon well for
futuristic approaches, these techniques can provide pertinent assistance for certain
and can also come up with many unified approaches for skin lesion issues preven-
tion. The development in these mechanisms will also lead to assistance to patients
and dermatologists to cure the skin lesion speedily and effectively. Many Machine
Learning can be effectively used and applied to detect the skin lesion boundary very
effectively. Naïve Bayes, Kernel SVM, Random Forest, Logistic Regression, and
convolutional neural network are some of the used algorithms mentioned above in
the discussion and have been given a mention in the literature survey. The CNN
Model though had the most accurate among all the models.
Researches and implementations of medical information and figures are accessible
in a very limited manner. We need more real-time info’s to be made available in the
coming future so that the detection of skin lacerations can be better tested and inves-
tigated with the latest advances in Artificial Intelligence and the pros of diagnosis
assistance led with Artificial Intelligence. Machine learning offers an outstanding
potential in the field of dermatology, ranging from efficient diagnosis to putting
forward much better, effective, and safer treatments. As the Machine Learning and
Artificial intelligence mechanisms advance, dermatologists will need to obtain at
680 A. Singh et al.

least a basic understanding of how these mechanisms work, and also to figure out
that with what, when and how it should be applied effectively in real-time medical
set-up. While ML methods are powerful, they still have quite a few similarities to the
past clinical equipment, in which doctors’ interaction is very important for the proper
application of the tools in a real-world set-up. We are also required to be acquainted
with the techniques of how budding biases can interact and make changes with the
black-box feature of all the algorithms. One more crucial and important development
needed is to make all these mechanisms which include the skin tones and color. Future
researches in Machine Learning need to be crystal clear by providing algorithms and
datasets to the public for more futuristic validation and implementations. Before
launching into the market, careful and diligent peer-reviewed anticipated clinical
proceedings should be conducted. At last, involving more and more dermatologists
in the process development and trials of Machine Learning is of vital importance for
producing efficient and clinically effective technologies.

References

1. Juyal P, Sharma S (2020) Machine learning algorithms based skin disease detection
2. Chan S, Reddy V, Myers B et al (2020) Machine learning in dermatology: current applications,
opportunities, and limitations. Dermatol Ther (Heidelb) 10:365–386
3. Erçal F, Moganti M, Stoecker WV, Moss RH (1993) Detection of skin tumor boundaries in
color images. IEEE Trans Med Imaging 12(3):624–626
4. Sigurdsson S, Philipsen PA, Hansen LK, Larsen J, Gniadecka M, Wulf HC (2004) Detection
of skin cancer by classification of Raman spectra. IEEE Trans Biomed Eng 51(10):1784–1793
5. Aberg P, Nicander I, Hansson J, Geladi P, Holmgren U, Ollmar S (2004) Skin cancer identifica-
tion using multifrequency electrical impedance-a potential screening tool. IEEE Trans Biomed
Eng 51(12):2097–2102
6. Wong A, Scharcanski J, Fieguth P (2011) Automatic skin lesion segmentation via iterative
stochastic region merging. IEEE Trans Inf Technol Biomed 15(6):929–936
7. Demyanov S, Chakravorty R, Abedini M, Halpern A, Garnavi R (2016) Classification
of dermoscopy patterns using deep convolutional neural networks. In 2016 IEEE 13th
International symposium on biomedical imaging (ISBI). IEEE, pp 364–368
8. Grana C, Pellacani G, Cucchiara R, Seidenari S (2003) A new algorithm for border description
of polarized light surface microscopic images of pigmented skin lesions. IEEE Trans Med
Imaging 22(8):959–964
9. Wighton P, Lee TK, Lui H, McLean DI, Atkins (2011) Generalizing common tasks in automated
skin lesion diagnosis. IEEE Trans Inform Technol Biomed 15(4):622–629
10. Emre Celebi M, Wen Q, Hwang S, Iyatomi H, Schaefer G (2013) Lesion border detection in
dermoscopy images using ensembles of thresholding methods. Skin Res Technol 19(1):e252–
e258
11. Ganster H, Pinz P, Rohrer R, Wildling E, Binder M, Kittler H (2001) Automated melanoma
recognition. IEEE Trans Med Imaging 20(3):233–239
12. R Sumithra R, Suhil M, Guru DS (2015) Segmentation and classification of skin lesions for
disease diagnosis. Procedia Comput Sci 45:76–85
13. Juan Lu J, Kazmierczak E, Manton JH, Sinclair R (2012) Automatic segmentation of scaling
in 2-d psoriasis skin images. IEEE Trans Med Imaging 32(4):719–730
14. Salimi S, Sabbagh Nobarian M, Rajebi S (2015) Skin disease images recognition based on
classification methods. Int J Tech Phys Problems Eng 22(7):78–85
Impact of Image Classification in Dermatology: Automated Diagnosis 681

15. Kolkur S, Kalbande D, Shimpi P, Bapat C, Jatakia J (2017) Human skin detection using RGB,
HSV and YCBCR color models. arXiv preprint arXiv:1708.02694
16. Kotian AL, Deepa K (2017) Detection and classification of skin diseases by image analysis
using MATLAB. Int J Emerg Res Manage Technol 6(5):779–784
17. Oyola J, Arroyo V, Ruedin A, Acevedo D (2012) Detection of chickenpox vesicles in digital
images of skin lesions. In: Iberoamerican Congress on pattern recognition. Springer, pp 583–
590
18. Chung DH, Sapiro G (2000) Segmenting skin lesions with partial-differential-equations-based
image processing algorithms. IEEE Trans Med Imaging 19(7):763–767
19. Zhong LS, Jin X, Quan C, Zhi-ping WEI, Yan-qun LIU (2011) Diagnostic applicability
of confocal laser scanning microscopy in psoriasis vulgaris. Chin J Dermatovenereology
25(8):607–608
20. Kuila S, Dhanda N, Joardar S, Neogy S, Kuila J (2019) A generic survey on medical big data
analysis using internet of things. In: First international conference on artificial intelligence and
cognitive computing. Springer, Singapore, pp 265–276
21. Janani SK, Raman S (2019) A comprehensive review on acne, its pathogenesis, treatment,
in-vitro and in-vivo models for induction and evaluation methods
22. Madalli V (2013) Lichen planus-a review. IOSR J Dent Med Sci 12:61–69. https://doi.org/10.
9790/0853-1216169
23. Gupta LK, Martin AM, Agarwal N, DSouza P, Das S, Kumar R, Pande S, Das NK, Kumaresan
M, Kumar P, Garg A, Singh S (2016) Guidelines for the management of Stevens–Johnson
syndrome/toxic epidermal necrolysis: an Indian perspective. Indian J Dermatol Venereol Leprol
82:603–625
24. Patnaik SK, Sidhu MS, Gehlot Y, Sharma B, Muthu P (2018) Automated skin disease
identification using deep learning algorithm. Biomed Pharmacol J 11(3)
IPFS-Based Image Sharing Using
Perceptual Hash Code with Digital Right
Management

D. Geethanjali, R. Priya, and R. Bhavani

Abstract In the current scenario, a large amount of information is exchanged over

the Internet. An unauthorized user access may cause the information to be modified,
disseminated, misused, detected, and used. Digital rights management (DRM) is a
method of preserving copyrights in digital media. It often requires more elements
than simply restricting or prohibiting copies of the content. This strategy includes
the deployment of technologies that limit the copying and use of copyrighted works
as well as proprietary software. The proposed approach has two modules. They are
hash code generation of an image and Store and retrieve the data from IPFS. First,
select the watermarked image. The different types of perceptual hashes are average
hash, perceptual hash, difference hash, and wavelet hashing. The perceptual hash
codes are generated for the selected watermarked image. Then, the watermarked
image and created copyright text files are shared through the IPFS system. The goal
of this proposed system is to compare the perceptual hashes by using the hamming
distance, size of the image, and calculation time. Based on the evaluation, difference
hash algorithm provides a better result. So, it is used to check whether the send
and retrieve files of IPFS are similar or not. In the IPFS file system, hash codes are
used to find the location of a file effectively and securely. This paper proposes, how
IPFS and perceptual hashes are preventing unauthorized user access to the system
while sharing the images over the Internet. Genuine image authors’ copyrights are
protected by using perceptually identical images on a decentralized blockchain IPFS
store.

Keywords Perceptual hashes · Blockchain · IPFS · Digital right management ·

Watermark

D. Geethanjali (B) · R. Priya · R. Bhavani

Department of Computer Science and Engineering, Annamalai University, Chidambaram,
Tamilnadu, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 683
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_50
684 D. Geethanjali et al.

1 Introduction

The rapid advancement of technology has resulted in a significant increase in multi-

media content, particularly in recent years. The total amount of digital content has
risen at an exponential rate and continues to do so. We live in a world where almost
everything can be digitalized and sent from one end of the globe to the other in
seconds. On the Internet, modern technology allows for fast data transfer, data content
alteration, and the creation and retransmission of unauthorized copies. However, in
the network media, security, robustness, and content protections remain a difficult
task. To solve those problems, the proposed system provides copyright verification
with some concealed data.
The contribution of the work is the secure sharing of data with the help of hash
algorithms. This may help for various applications like digital garment design works,
research record-keeping systems, medical applications, online education, and so on.
The work aims to discuss the applications of perceptual image hashing which is
better than cryptographic hash functions. This proposed work is divided into two-step
processes. First, generate and compare the perceptual hash code for the watermarked
image. The authorized users’ blockchain list was created. Next, the watermarked
image and copyright text file are shared through IPFS. The following is a breakdown
of the paper’s structure. The related works of the existing systems are listed in Sect. 2.
The proposed system is described in Sect. 3. The implementation results and design
are presented in Sect. 4. The conclusion and future efforts are described in Sect. 5.

1.1 Hashing

The essential qualities of the perceptual hash algorithm are differentiability, colli-
sion tolerance, summarization, stability, and uni-direction. Hashing is a file-based
algorithm for calculating a fixed-size bit string value. A file is essentially a set of
data blocks. Hashing reduces the length of the data to a fixed-length value or key that
represents the original string. When compared to the cryptographic hash, perceptual
hashes are not affected by the avalanche effect and are less susceptible to perceptual
changes in images such as rotation, cropping, re-sizing, grayscale conversion, salt and
pepper noise, exposure or saturation changes, and so on. The hash function returned
bit value. They are called hash values, hash codes, digests, or simply hashes. The
perceptual hashes are used in various applications like digital garment design works,
research record-keeping systems with digital right management (DRM), similarity,
and dissimilarity of an image.
IPFS-Based Image Sharing Using Perceptual Hash Code … 685

1.2 IPFS (InterPlanetary File System)

The images and hashes are stored using the IPFS architecture. It is a peer-to-peer
hypermedia distribution protocol that is decentralized. By ensuring that files with
identical content are only kept once, the IPFS system eliminates content duplication.
This creates the Internet world as a faster, secure, and more open environment. Data
warehoused on IPFS can be scrambled with any usual encryption techniques used to
retrieve the stored information. The registered authenticated users can only access
the system with the proper encryption keys. The diagrammatic representation of the
IPFS system based on blockchain is shown in Fig. 1.
The IPFS system works based on the following three fundamental principles.
They are (1) Unique identification through content addressing. (2) Content linking
by using directed acyclic graphs (DAGs). (3) Content discovery through distributed
hash tables (DHTs).
IPFS operates by tying all networked devices to the same file structure. To ensure
the immutability, IPFS file structure combined with the Merkle trees. A hash tree,
also called the Merkle tree, is used to efficiently and securely encrypt blockchain
data. It enables the peer-to-peer blockchain network to quickly verify blockchain
data as well as transfer vast volumes of data from one computer node to the next.
The directed acyclic graphs are permitting the user to know the versions of content
on IPFS. InterPlanetary version control is a distributed version control framework
based on IPFS, similar to git (InterPlanetary File System). It can handle any type of
data, not just human-readable text. It is also mainly suitable for versioning large files.

Fig. 1 Diagrammatic representation of IPFS system based on blockchain

686 D. Geethanjali et al.

Distributed hash table (DHT) serves as a catalog and navigation system, assisting
the network in keeping track of and finding data. The IPFS system performs the
file searching operation in two steps. First, IPFS identifies the file with content
addressing through the hash value. Next, the network connection establishment of
the corresponding node is identified; then, the file is retrieved.
The most significant benefit of employing IPFS in this manner is that nodes
become individual servers capable of serving content to others. IPFS is widely utilized
as a data-sharing storage platform. It offers great availability and performance, but it
lacks the capacity to trace access and authentication, making it difficult to investigate
illicit access and authorship.

1.3 Blockchain

A blockchain is a decentralized software network that serves as a digital record and

allows for the secure transfer of property without the use of a third party. Blockchain
is a technology that facilitates the digital exchange of units of value. It is the current
technology that enables the digital flow of information. On a blockchain network,
anything from money transfers to land titles to votes can be tokenized, stored, and
traded.

2 Related Work

Geethanjali et al. [1] proposed a secure transaction through blockchain. It is achieved

with the help of watermarked image and generated image hash code. It is passed to the
blockchain for the secure sharing of data. Dedge et al. [2] proposed image copyright
protection using blockchain by a digital watermark added on the image. Blockchain is
used to store the data in a protected manner, which also provides proof-of-ownership
for the copyright holder. This scheme also affords a cross-check functionality for
commercial users which leads to avoiding copyright issues by checking whether
the image is copyrighted or not. Mehta et al. [3] proposed automatic detection and
rejection of tampered images using perceptual hashes for stock photo Web sites
displayed in the market places which also establishes robust smart contracts with
the help of the Ethereum blockchain. Lu et al. [4] presented a blockchain-based
approach for managing digital rights for digital garment design works. A new proof-
of-delivery technique is proposed for the sake of trade fairness. The design effects
were displayed with privacy and compatibility with the registered users. Nyaletey
et al. [5] proposed a novel approach (BlockIPFS) to generate a clear audit trail.
BlockIPFS allows for increased data trustworthiness and user security, as well as a
great solution to track all activity linked to a specific file utilizing blockchain as a
service. Zhu et al. [6] proposed a combination of the perceptual hash algorithm and
SIFT features, which increase the exactness of the image retrieval. It also eliminates
IPFS-Based Image Sharing Using Perceptual Hash Code … 687

the perceptual hash algorithm’s influence on geometric attacks, imagescale changes,

and other abnormal conditions. Rajalakshmi et al. [7] proposed a mixing of IPFS and
blockchain technologies and the traditional encryption methods login id, biometric
authentication, and secrete question are asked to check the authenticity. They used
to maintain the academic research records in a secure manner with tamper-proof.
This system improves the privacy, authenticity, and access control of the registered
users. Steichen et al. [8] suggested an InterPlanetary File System (IPFS). It may be
used to efficiently store and share enormous volumes of data. This work described a
reformed version of the InterPlanetary File System (IPFS) that uses Ethereum smart
contracts to enable access-controlled file sharing. The smart contract is used to keep
track of the access control list. When a file is uploaded, downloaded, or moved, it also
allows interaction with the smart contract. Tang et al. [9] proposed medical image
sharing in a secure manner built through blockchain smart contracts and credit scores.
This provides the solution for the following problems, for cross-domain distribution
and patient privacy protection problem while sharing the medical image informa-
tion. Meng et al. [10] offered a design scheme based on digital watermarking and
blockchain to increase the security and transparency of information in the copy-
right management system. Zhang et al. [11] proposed a DRM mechanism based on
blockchain technology. This blockchain technology is used to keep track of copy-
right transactions and authorizations in a transparent and secure manner. The smart
contract establishes the copyright transaction’s trustworthy automated licenses. Ma
et al. [12] proposed a technique, based on watermark and blockchain, a compre-
hensive and high-level security artwork picture digital right management strategy
for online abusing detection was presented. Zhaofeng Ma, et al. [13], proposed, a
trustworthy, protected, effective and tamper resistance digital content services using
DRM chain scheme. Li Weng, et al. [14], Image content authentication is achieved
through wavelet watermarking with perceptual hashing.

3 Proposed System

The available technologies are becoming more accessible and open source in the
current scenario. This has posed significant challenges in recent years in terms of
data management, authentication, and security of sharing data. This proposed work
was implemented in two steps.
They are (1) perceptual hash code generation. (2) datastore and retrieval from
IPFS. Figure 2 illustrates the block diagram of the proposed system.
688 D. Geethanjali et al.

Watermarked Image

Hash code Generation using Hash Algorithms

Average Difference Perceptual Wavelet

Hash Hash Hash/ DCT Hash

Sender sends the Watermarked file with

respective Perceptual Hash

Initialize
and Add the

Connect the
Pin the File IPFS to the Net- If Connection
to IPFS work not Established

Check the loaded file on the IPFS

Receiver download
the respective file Stop the
with the Hash code IPFS Process

Fig. 2 Block diagram of the proposed system

3.1 Watermark Process

Get the cover image. Get the watermark image. The discrete wavelet transform
technology was used to incorporate the watermark image on the cover image. Create
a copyright text file for the watermarked image.

3.2 Perceptual Hash Code Generation

Perceptual hashing functions are not the same as cryptographic hashing functions.
In contrast to cryptography hashes, especially, perceptual hashes are similar if their
inputs have similar characteristics. The distinction (Hamming distance) metric of
IPFS-Based Image Sharing Using Perceptual Hash Code … 689

Fig. 3 Average hash code

generation process

dissimilarity is used between perceptual hashes. The perceptual hashes give effective
encryption and are more accurate and more compact.
To calculate the hash value, perceptual hashes usually follow the methods below.
The input image was given to the hash function. The image hashing function calcu-
lates the image hash value and displays the hash code for the respective image. This
is performed based on the visual content of an image. The image hashing is used
to perform some of the following applications like content-based image retrieval,
near-duplicate detection, and reverse image search engines. The four most popular
perceptual hash algorithms are average hash (AHA), difference hash (DHA), percep-
tual/DCT hash (PHA), and wavelet hash (WHA). Each perceptual hash algorithm
works based on the given below steps.

3.2.1 Algorithm Steps for Average Hash

(i) Resize the input image size into 8 × 8.

(ii) Convert the resized image into a grayscale image.
(iii) Calculate gray average value, and it is denoted as M.
(iv) According to the comparison with the gray average to get the binary map.

0 if A(i) pM
h(i) = (1 ≤ i ≤ 64) (1)
1 otherwise

where h(i) is denoted as average hash, A(i) represents the normalization of

each pixel value based on the comparison of average value M.
(v) Combine binary maps into 64-bit hash values (fingerprints) in order.

Figure 3 shows the general hash code generation process for the average hash.

3.2.2 Algorithm Steps for Difference Hash

(i) Resize the image size into 9 × 8.

(ii) Convert the resized image into a grayscale image.
(iii) Calculate the difference value (compare the size of two adjacent values in
each row) to get a binary map

dpxr,c = pxr,c+1 − pxr,c (2)

690 D. Geethanjali et al.

Fig. 4 Difference hash code generation process

where pxr,c is denoted as the gray value of the resized image at r-row and c-
column, dpxr,c specifies the difference value. Like this, the pixels per row yield
8 differences between adjacent pixels, and 8 differences of 8 rows become an
8 × 8 image.
(iv) Combine binary maps into 64-bit hash values (fingerprints) in order.

Figure 4 shows the general hash code generation process for difference hash.

3.2.3 Algorithm Steps for Perceptual Hash

(i) Resize the image size into 8 × 8.

(ii) Convert the resized image into a grayscale image.
(iii) Calculate the DCT transform of the image.
(iv) Reduce DCT.
(v) Calculate the mean (m) of the DCT coefficients.

1 
8 8
m= dr,c (3)
64 r =1 c=1

where r, c are rows and columns of the DCT coefficients. d refers to the mean
difference between the DCT coefficients.
(vi) Compare and judge to get a binary map.

0 Ci pm
hi = , i = 0, k 63. (4)
1 Ci ≥ m

where h i denotes bit of the perceptual hash value at the i position; Ci is the
array of DCT coefficients; m specifies the mean value of the DCT coefficients.
(vii) Binary graphs are combined into a 64-bit hash value (fingerprint) in sequence.

Figure 5 shows the general hash code generation process for perceptual/DCT
hash.
IPFS-Based Image Sharing Using Perceptual Hash Code … 691

Fig. 5 Perceptual hash code generation process

3.2.4 Algorithm Steps for Wavelet Hash

(i) It generates a gray value image sized 8 × 8.

(ii) The image is then transformed into a two-dimensional wavelet transform.
(iii) Removes the lowest Haar frequency LL (max). The lowest frequency has only
one data point/pixel, which shows the contrast image and is not very useful
for hashing.
(iv) Binary graphs are combined into a 64-bit hash value (fingerprint) in sequence.
Figure 6 shows the general hash code generation process for wavelet hash.

3.3 Data Store and Retrieval from Interplanetary File System

(IPFS)

The centralized servers used to store and retrieve the users data. It also resolves any
issues that may arise in centralized server. The watermarked image was stored and
distributed using IPFS. The following steps are performed to add and retrieve the
data through IPFS.
(i) Download and install IPFS for your system.
(ii) Check whether the path of the file is stored in a proper location.
(iii) Initialize and add the file to IPFS. Pin the file to IPFS.
(iv) Connect the IPFS system to the network.
(v) Check the loaded file on the IPFS by using the https://ipfs.io/ipfs/Hashcode
of a file/file directory

Fig. 6 Wavelet hash code generation process

692 D. Geethanjali et al.

Cats.jpg Resized image Gray scale image Calculation of Average hash

Binary Representation 0000000000000000000111000010000001000010100001000111111011111110
Hexadecimal Representation: 00001c6042847efe

Fig. 7 Representation of the generated average hash code for Cat.jpg image

(vi) Finally, the daemon is stopped by using the Ctrl + C keys.

4 Implementation Results and Discussion

4.1 Perceptual Hash Code Generation

4.1.1 Average Hash

To obtain an average hash, execute the following steps: First, the size of an image
was reduced (for example, to an 8 × 8 image, to further shorten the number of
computations); then, the image was converted to gray scale, 3rd average the resulting
colors (for an 8 × 8 image, we average 64 colors), 4th calculate the bits by comparing
if each color value is greater than or less than the mean, 5th build the hash. The
following Fig. 7 represents the process of average hash code generation.

4.1.2 Differential Hash

The differential hash works in the following way: The difference hash algorithm
primarily the image was scaled down to 9 × 8 pixels, then creates a grayscale image
from the input image. Next, from each row, the first 8 pixels are reviewed serially
from the left side to the right side and compared to their right hand neighbor, which,
equivalent to the average hash algorithm, outcomes in a 64-bit hash. The following
Fig. 8 represents the process of difference hash code generation.

4.1.3 Perceptual/DCT Hash

The perceptual hash works in the subsequent manner: Initially, this algorithm calcu-
lates the gray value of an image and scales it down. In this case, to desire a factor of
4, which is why we scaled down to 8 * 4 × 8 * 4, that is, a 32 × 32 image. Apply
IPFS-Based Image Sharing Using Perceptual Hash Code … 693

Cat.jpg
Resized image Gray scale Image calculation of Difference hash
Binary representation 0111100001011011100111110011000001111001111010110111001100010111
Hexadecimal Representation: 785b9f3079eb7317

Fig. 8 Representation of the generated difference hash code for Cat.jpg image

Cat.jpg
Resized image Gray scale image Calculation of DCT hash
Binary Representation: 1011010000001000110100000011010001100001100000110001101100110000
Hexadecimal Representation: b408d03461831b30

Fig. 9 Representation of the generated perceptual hash code for Cat.jpg

a discrete cosine transform on this image, first per row and afterward per column.
The pixels with high frequencies are now located in the upper-left corner. Because
of this reason, the image cropped to the upper left 8 × 8 pixels. The following Fig. 9
represents the process of perceptual hash code generation.

4.1.4 Wavelet Hash

The wavelet hash function as follows: This wavelet hash algorithm also generates a
gray value image sized 8 × 8. A two-dimensional wavelet transform is applied to the
image. To improve the test results to set the top row to 0, that is, to black and re-apply
the wavelet transform three times. It is implemented by using the image hash library.
Next, similar to the perceptual hash algorithm, each pixel is compared to the median,
and the hash is generated. The following Fig. 10 represents the process of wavelet
hash code generation.
Perceptual hash codes are implemented based on their algorithms. The binary form
hash code is generated and is converted to a hexadecimal representation. Usually, this
hexadecimal format is used for further operations for easy understanding of humans.
The following Fig. 11 shows the sample images are taken from the open database
for testing the proposed system.
The following Table 1 shows the generated perceptual hash codes for sample
images.
694 D. Geethanjali et al.

Cat.jpg Gray scale image Calculation of wavelet hash

Resized image
Binary Representation: 0000010110111001111110011111101111111010000000000000011100000101
Hexadecimal Representation: 05b9 f9fb fa00 0705

Fig. 10 Representation of the generated wavelet hash code for Cat.jpg image

a) Lena.jpg b)Artima c) Fruits.jpg d)ctimage.jpg e)Greenbamb e)Cat.jpg

ge.jpg oo.jpg

Fig. 11 Sample test images for the proposed system

Among the four perceptual hashes based on the hamming distance, calculation
time, the difference hash (DHA) provides a minimal error and better result. So, DHA
is used for further simulation process of IPFS.

4.2 Storing and Retrieval of an Image from IPFS

The go-ipfs is an open-source software. It is provided by the IPFS Web site. Figure 12
shows the downloaded IPFS installed into the system. And also, it was checked
whether the path of the file is stored in a proper location.
The given Fig. 13 shows the initialization of the IPFS system.
The given Fig. 14 represents the creation of a new folder and adding the
watermarked image file and copyright text file into that folder using IPFS.
Figure 15 represents that the file folder is pin to the IPFS, using the hash code
Figure 16 shows the IPFS system connection that is established to the network
File/File directory is searched by using the hash code of a respective folder. The
file/file folder is retrieved from IPFS by using the following link. https://ipfs.io/
ipfs/QmVPzqYHb16nc4iRJB62gv5WxYG5sMZK5YzP99R1ascKzE. This shows
the folder file names with their and size of each file.
Figure 17 shows the loaded file on the IPFS by using the file/file directory.
Figures 18 and 19 shows the downloaded folder content of an image and copyright
text file from an IPFS.
The following Fig. 20 shows the downloaded folder of a Lena image which
contains the two files, watermarked-lena.jpg and lena.txt.
Table 1 Generated perceptual hash codes for sample images
Sample images Average hash Difference hash DCT hash Wavelet hash
Lena.jpg ac6969eb61e0a080 58dbdbdbdb414832 fb819e56b0d55612 ae69e9ebe1e0a888
Artimage.jpg eeed6721208bf300 c8c9cdc5c91a86c2 a392f5f946a439c2 eeef67352003fb00
Fruits.jpg fff5e1c004032323 8d4989096df64e47 e3f2a51d1e318966 fff5e1c0040b23a7
Ctimage.jpg c7c667e7e723c18 71cc8e96b696d471 c4d30b4c2dcb6c3e 387c667e56723c18
Greenbaboo.jpg fcfcfcf4f8743400 e1f1c1edc1e5e4fc fcfcfcf4f8743400 fcfcf8f478743000
IPFS-Based Image Sharing Using Perceptual Hash Code …

Cat.jpg 00001c6042847efe b408d03461831b30 b408d03461831b30 05b9 f9fb fa00 0705

695
696 D. Geethanjali et al.

Fig. 12 Installation and file storing location of IPFS

Fig. 13 Initialization of the IPFS

Fig. 14 Adding the image file and copyright file to IPFS

Fig. 15 Pin the file folder to IPFS

Fig. 16 Connect the IPFS system to the network

IPFS-Based Image Sharing Using Perceptual Hash Code … 697

Fig. 17 Loaded folder of Lena image on the IPFS

Fig. 18 Downloaded watermarked-lena.jpg image

Fig. 19 Downloaded copyright text file of a watermarked image Lena from the IPFS

Fig. 20 Downloaded folder of a Lena image

The following Table 2 shows the similarity matches between the sent watermarked
image and retrieved image from the IPFS.
Hamming distance is used to compare the two images. It refers to the number of
dissimilar bits between two hashes. The hamming distance (HD) value is 0; then, the
source and retrieved images are similar. If the HD value is between 1 and 10, then the
two images are hypothetically variance, and the HD value is greater than 10; then,
698 D. Geethanjali et al.

Table 2 Similarity match between the sent and retrieval image of IPFS
Image name Hash value of sent Hash value of Hamming Size of Result
watermarked image retrieved image from distance the
IPFS image
(in kilo
bytes)
Lena.jpg 898d82a4466ca583 898d82a4466ca583 0 30 Same
Artimage.jpg 372752a9364429f6 372752a9364429f6 0 42 Same
Fruits.jpg 709c34b6920528b0 709c34b6920528b0 0 40 Same
Ctimage.jpg 452b514949492b05 452b514949492b05 0 28 Same
Greenbamboo.jpg 0d061a13360a1b03 0d061a13360a1b03 0 25 Same
Cat.jpg 70539f3371eb7357 70539f3371eb7357 0 26 Same

the images are different. As per the result, the hamming distance value and size of
the images are similar while sent and retrieved data. So, the hash code performs the
better retrieval of data from IPFS with good protection.

4.3 Performance Evaluation

4.3.1 Peak Signal-To-Noise Ratio (PSNR)

It is a measure of the peak error between the two images. The ratio between the
maximum possible power of a signal and the power of corrupting noise affects the
fidelity. The formula for the PSNR value of an image is given in Eq. (5).

2552
PSNR = 10 log10 (5)
MSE

4.3.2 Normalized Correlation (NC)

It evaluates the performance of the watermarking scheme. To estimate and check

the quality of the extracted watermark image and the original watermark image. It is
given in Eq. (6)
m n
i=0 j=0W (i, j)W  (i, j)
NC =    (6)
m n 2 m n 
i=0 j=0 W (i, j) i=0 j=0 W (i, j)
2
IPFS-Based Image Sharing Using Perceptual Hash Code … 699

Table 3 Attacked images

Salt and pepper noise (0.015) Crop Rotation (5°) Rotation (10°)

4.3.3 Bit Error Rate (BER)

It is defined as percentage of bits that have errors relative to the total number of bits
received in a transmission. It has been calculated by using Eq. (7)

No. of Error Bits

BER = (7)
Total no. of bits sent

4.4 Watermark Attacks

To check the robustness of the proposed system, is tested by applying the following
attacks like salt and pepper, crop, and rotation attacks on the sent image. Table 3
shows the attacked image.
The PSNR of the embedding value of the Lena image is 36 dB. The retrieved
image NCC values of all the test images are 1 for both before and after attacks. At
the same time after the attack, the bit error value varies from 0.8816 to 0.9952. The
retrieved images are recognizable and have reasonable correlation coefficients.

5 Conclusion and Future Enhancement

The proposed system can find the similarity between sent and retrieval images using
the perceptual hash. As per the results, the identical data or image may be identi-
fied, with a different watermark images. The perceptual hashes have the following
better properties. It is a useful technique for rapidly accessing data based on a key
value. It also provides image protection. IPFS provides a better result of storing and
retrieval of information from IPFS. In future, it can be used to evaluate the perfor-
mance measure of the perceptual hashing and apply the transformation attacks on
the image. Perceptual hash functions are commonly used in detecting cases of online
copyright infringement as well as in digital forensics. The effectiveness of perceptual
hashing and IPFS file storage will be applied in real-time application digital garment
700 D. Geethanjali et al.

design work. The smart contract will be established through blockchain with the
authenticated user.

References

1. Geethanjali D, Priya R, Bhavani R (2020) Blockchain-based protected digital copyright

management with digital watermarking. In: Hemanth DJ et al (eds) Intelligent systems and
computer technology. IOS Press. https://doi.org/10.3233/APC200113
2. Dedge O, Shingade R, Jagtap A, Yadav A, Kamble A (2020) Image copyright protection system
using blockchain. Int J Future Gener Commun Networking 13(3s):37–43
3. Mehta R, Kapoor N, Sourav S, Shorey R (2019) Decentralised image sharing and copy-
right protection using blockchain and perceptual hashes. In: 11th international conference on
communication systems and networks (COMSNETS). IEEE. 978-1-5386-7902-9/19$31.00
©2019
4. Lu Z, Shi Y, Tao R, Zhang Z (2019) Blockchain for digital rights management of design works.
IEEE. 978-1-7281-0945-9/19/$31.00©2019
5. Nyaletey E, Parizi RM, Zhang Q, Raymond Choo K-K (2019) BlockIPFS—blockchain-enabled
interplanetary file system for forensic and trusted data traceability. In: IEEE international
conference on blockchain (blockchain). IEEE. 978-1-7281-4693-5/19/$31.00 ©2019. https://
doi.org/10.1109/Blockchain.2019.00012
6. Zhu W, Chang J (2019) Research and implementation of perceptual hash algorithm for image
retrieval. AIP Conf Proc 2073(020083):2019. https://doi.org/10.1063/1.5090737PublishedOn
line:13February
7. Rajalakshmi A, Lakshmy KV, Sindhu M, Amritha P (2018) A blockchain and IPFS based
framework for secure research record keeping. Int J Pure Appl Math 119(15):1437–1442.
ISSN: 1314-3395 (on-line version)
8. Steichen M, Fiz B, Norvill R, Shbair W, State R (2018) Blockchain-based, decentralized access
control for IPFS. In: IEEE conferences on internet of things, green computing and communica-
tions, cyber, physical and social computing, smart data, blockchain, computer and information
technology, congress on cybermatics. IEEE. 978-1-5386-7975-3/18/$31.00 ©2018. https://doi.
org/10.1109/Cybermatics_2018.2018.00253
9. Tang H, Tong N, Ouyang J (2018) Medical images sharing system based on blockchain and
smart contract of credit scores. In: Proceedings of 2018 1st IEEE international conference on
hot information-centric networking (HotICN 2018). IEEE. 978-1-5386-4870-18/$31.00©2018
10. Meng Z, Morizumi T, Miyata S, Kinoshita H (2018)Design scheme of copyright management
system based on digital watermarking and blockchain. In: IEEE 42nd annual computer software
and applications conference (COMPSAC), Tokyo, pp 359–364
11. Zhang Z, Zhao L (2018) A design of digital rights management mechanism based on blockchain
technology. In: Chen et al (eds) ICBC 2018. Springer International Publishing, pp 32–46. https://
doi.org/10.1007/978-3-319-94478-4_3
12. Ma Z, Huang W, Gao H (2018) A new blockchain-based trusted DRM scheme for built-in
content protection. EURASIP J Image Video Proc
13. Ma Z, Jiang M, Gao H, Wang Z (2018) Blockchain for digital rights management. Futur Gener
Comput Syst. https://doi.org/10.1016/j.future.2018.07.029
14. Weng L, Darazi R, Preneel B, Macq B, Dooms A (2012) Robust image content authentication
using perceptual hashing and watermarking. In: Lin W et al (eds) PCM 2012. pp 315–326
Mid-Price Prediction Using Online
Kernel Adaptive Filtering

Shambhavi Mishra, Tanveer Ahmed, and Vipul Mishra

Abstract Stock price prediction is an important and popular area of time-series pre-
diction. Although numerous methods have been proposed for stock price prediction
in literature, a good model with excellent performance is still a long way to go. In
addition, earlier prediction methods were based on offline classification and regres-
sion, thereby requiring independent and large batch-oriented samples. We would
emphasize on one point: a financial time-series data is non-stationary in nature.
Hence, traditional regression and classification-based techniques are not ideal for
predicting stock prices. In this paper, we propose an “online” kernel adaptive fil-
tering (KAF)-based approach for stock price prediction. We experiment with ten
KAF methods and analyze their predictive capability on nine different time windows
such as: one day, sixty minutes, thirty minutes, twenty-five minutes, twenty minutes,
fifteen minutes, ten minutes, five minutes, and one minute. The process is repeated
for all 50 stocks of the National Stock Exchange’s main Index (NIFTY-50). The
empirical results from the experiment show that KAF has several advantages over
the current state-of-the-art. For instance, the probability that we can predict the next
up or a down movement of a stock is 66%. Further, their low latency makes them a
preferred choice for high-frequency trading.

Keywords Stock price prediction · Financial time-series · Kernel adaptive

filtering · Online learning

S. Mishra (B) · T. Ahmed · V. Mishra

Bennett University, Greater Noida, India
e-mail: sm7835@bennett.edu.in
T. Ahmed
e-mail: tanveer.ahmed@bennett.edu.in
V. Mishra
e-mail: vipul.mishra@bennett.edu.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 701
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_51
702 S. Mishra et al.

1 Introduction

In economics and investments research, time-series has always been a prominent

area of investigation. Time-series has a multitude of applications, with stock price
prediction being one of the popular domains. A good stock price prediction model
can minimize the risk in trading and can optimize asset management. It can reduce
overhead costs and mitigate market risks [1]. Although popular, stock price predic-
tion at any instance of time has remained one of the non-trivial problems, earlier
research has demonstrated that the stock price prediction is challenging [2]. Hence,
significant efforts have been devoted in addressing the issue; however, the results in
past research showed a little improvement in achieving the goal. A number of studies
were conducted where various methods have been developed for the improvement in
the stock price prediction. These methods vary from support vector machines (SVM)
[3], fuzzy logic [4], particle swarm optimization [5], neural networks [6], and genetic
algorithms [7] to name a few. Despite the vast body of work, we found that the authors
have often overlooked the potential of kernel adaptive filtering (KAF) for stock price
prediction. Though there are few studies [8, 9] that utilized KAF methods for stock
price prediction, a large-scale evaluation of KAF methods for stock price prediction
was limited from the previous research. It is also important to mention here that KAF
methods can play a vital role in stock price prediction because they require fewer
iterations to converge. Along with that, KAF methods have shown outstanding per-
formance on non-stationary time-series prediction [10, 11]. Moreover, in time-series
prediction, KAF methods have been applied extensively [10, 12, 13]. Hence, it is
logical to apply the paradigm to predicting stock prices.
To address the challenges and as a potential solution, we utilized the paradigm
of online-KAF for stock’s mid-price prediction. Since, the issue of financial series
forecasting is non-trivial, we work with 50 stocks of the main Index at the National
Stock Exchange (NSE): Nifty-50. In contrast to the existing methods, the data was
collected at multiple time windows such as one minute, five minutes, ten minutes,
fifteen minutes, twenty minutes, twenty-five minutes, thirty minutes, sixty minutes,
and one day. We then apply the prediction algorithms to each of these windows
and try to show the predictive capabilities of KAF algorithms. In addition to this,
we experiment with ten different algorithms and try to find the best algorithm that
would produce good results.
The main contributions of the paper are:
• A novel KAF-based method for stock’s mid price prediction. The mid price is
calculated using two aspects; (high + low)/2 and (open + close)/2.
• Experimentation with ten KAF algorithms on nine different time windows and
fifty stocks.
To the best of our knowledge, this article is the first wherein we apply and test the
performance of ten different algorithms of the KAF class on multiple time windows.
The paper has been structured as follows. Section 2 summarized the details of the
existing work. The proposed method for stock’s mid price prediction using KAF-
Mid-Price Prediction Using Online Kernel Adaptive Filtering 703

based methods is discussed in Sect. 3. The experiments performed and their results
are included in Sect. 4. Finally, the conclusions and future directions are given in
Sect. 5.

2 Related Work

This section elucidates the detail of the existing work in the domain of stock price
prediction. Stock prediction is a tedious task, and the early methods were focused on
past prices using indicators [14, 15]. It has also been shown that the stock prediction
is noisy, chaotic, and follows nonlinearity [16–18]. The use of linear methods like
correlations, autoregressive integrated moving average (ARIMA) models, and mov-
ing averages were also explored in the past [19, 20]. Due to the chaotic behavior of
the stocks, in these methods, even short-time prediction needs heavy computation
[17]. Much of the literature also focused on machine learning (ML) techniques due
to their ability to handle complex patterns [21, 22].
Moreover, there is a great scope for sequential learning-based online methods as
the time-series data follows nonlinearity. Sequential learning techniques have the
ability of generating faster and acute results [23]. Neural network and support vector
regression were also proposed in sequential learning techniques [11, 24]. However,
support vector regression and neural network-based techniques have several dis-
advantages that make them less suitable for stock prediction task. In spite of the
powerful generalization ability, support vector regression is unable to perform well
on the large datasets. Neural network methods also showed slow convergence issue
and needs high computational power during training and inference [25]. Another
popular class of online learning-based method is kernel adaptive filtering (KAF)
[10]. For sequential learning technique, kernel-based online learning is efficient and
reduces the computational burden [12, 26]. In contrast to the other existing methods
of the literature, KAF techniques have gain more popularity as these methods have
several advantages like low computational complexity, non-parametric, and faster
converges [10].
In one of the researches [26], the authors compared multi-filter neural network
(MFNN) with random forest, SVR, and other NN-based approach for stock pre-
diction. The results demonstrated that MFNN outperformed other methods. In [27,
28], long short-term memory (LSTM) was also used to predict stock prices. [29]
combined SVR and kernel principal component analysis (KPCA) for improving
short-term prediction performance. In addition to this, [30] proposed the idea of
multi-step prediction using meta-cognitive recurrent kernel online learning. In one
of the researches [31], the authors used a simple LSTM network with single layer
and 200 nodes. In addition, the network used by [32] also uses a single layer LSTM
with 140 nodes. In contract to [31–33] used a deep architecture with four LSTM
layers and 96 nodes in the hidden layers. Each LSTM layer was further followed by
a dropout layer.
704 S. Mishra et al.

Table 1 Time window of 60-min (Stock-Adaniports)

Day High price Low price (High price + low % Change in
price)/2 mid-price
1d 387.65 384.05 385.85 −0.1555
2d 386 384.5 385.25 0.0324
3d 386 384.75 385.375 −0.0194
4d 386.9 384.7 385.8 0.4671
5d 388.9 385.3 387.1 0.5683
6d 390.75 387.85 389.3 −0.4880
7d 388.65 386.15 387.4 0.1484
8d 388.95 387 387.975 NA
If we choose M = 4, then Input = {−0.1555, 0.0324, −0.0194, 0.4671}.Output = {0.5683}

The authors of another research [29] found that in high dimensional space, train-
ing ML-based methods for stock prediction requires longer processing time and high
computing resources. To alleviate this, [34] proposed convolutional neural network
(CNN) for automatic feature selection and market prediction. Recently, the use of
deep reinforcement learning has also been explored for stock prediction [35] for
adaptive stock trading strategy. The authors proposed gated deterministic policy gra-
dient trading strategy (GDPG) and gated deep Q-learning trading strategy (GDQN)
for extracting financial features.
Furthermore, there are some studies where the authors made use of KAF for stock
prediction, i.e., [36]. In another method [37], the authors proposed two important
aspects of kernel filtering; step-size and kernel-size. Whereas, [36] introduces the
concept of the local model. Although the wok by Sergio et al. [36, 37] showed
some preliminary work on KAF for stock prediction, the effectiveness of the on
the large-scale dataset eludes literature. We also believe that the potential of the
KAF methods has not been utilized and analyzed thoroughly for stock prediction.
Moreover, in literature, no one thoroughly investigated stock price prediction with
multiple time windows. To the best of our knowledge, we are the first to test multiple
KAF algorithms on a large-scale dataset with various time windows.

3 Methodology

In this section, we have discussed the online KAF-based learning. KAF works with
self-tuning, and input–output mapping is formulated according to an optimization
criterion determined by an error signal. The main objective is to learn the input–output
mapping f : V → R, based on a well-known sequence of data ((v1 , d1 ), (v2 , d2 ), …,
(vi , di )), where V ⊆ R L is the input space, vi , i = 1, . . . , n is the system input at
sample time, and di is the desired response. The filter adaptively adjust weights,
ω(i − 1), where i denotes the discrete time interval. KAF has the following advan-
Mid-Price Prediction Using Online Kernel Adaptive Filtering 705

Fig. 1 Proposed stock price prediction framework

tages compared to the other existing techniques of regression and classification:

KAF follows the idea of universal approximators. It does not suffer from local min-
ima because it uses gradient decent and follow the concept of online learning and
also fits on non-stationary conditions well.

(||v − v  ||2 )
κ < v, v  >= exp . (1)
σ2
In stock price prediction, we predict stock’s future values depending upon the histor-
ical values. In the proposed work, the objective is to predict stock’s mid-price using
online KAF. Here, we enforced the concept of auto-regression of order n to predict
future change in the stock price. Table 1 shows the sample regression formulation
and considering the 60-minute time window for one stock (Adaniports). We pre-
dict and analyze the stock’s mid-price using two different scenarios (high + low)/2
and (open + close)/2 for all nine time window. The proposed stock price prediction
framework is shown in Fig. 1. There are various methods following the real-time
online prediction concepts [38]. Therefore, in the proposed work, we have worked
with the following algorithms:
1. Least mean square (LMS)
2. Kernel least mean square (KLMS)
3. Kernel affine projection algorithm (KAPA)
4. Leaky kernel affine projection algorithm (LKAPA)
5. Normalized online regularized risk minimization algorithm (NORMA) [12]
6. Quantized kernel least mean square algorithm (QKLMS) [13]
7. Fixed budget quantized kernel least mean square algorithm (FBQKLMS) [39]
8. Kernel adaptive filtering with maximum crossentropy criterion (KMCC) [40]
9. Multi-kernel normalized least mean square (MKNLMS) [41]
10. Probabilistic least mean square filter (PROB-LMS) [42].
706 S. Mishra et al.

430

420

410
Mid-Price

400

390

380

370

360 Original
Prediction

350
0 50 100 150 200 250
Instances

Fig. 2 Prediction for one stock (Adaniports) using KAPA (high + low)/2

4 Result

4.1 Dataset Description

To perform the experiments, the stock data from the National Stock Exchange of India
was used. We archived data for one year between January 01, 2019, and December 31,
2019, from 9:15 a.m. to 3:30 p.m. The dataset used in the experiment is available at the
link.1 The original dataset contains only one-minute time window data and includes
four different fields open, high, low, and close (OHLC) prices. In our research, data
for different time windows was prepared by merging OHLC quotes from the one-
minute data. More specifically, the dataset was created and preprocessed as per the
nine prediction windows (one minute, five minutes, ten minutes, fifteen minutes,
twenty minutes, twenty-five minutes, thirty minutes, sixty minutes, and one day). In
addition to this, while creating the dataset, some practical assumptions were made:
the trading hours were between 9:15 a.m. and 3:30 p.m., weekends data was not
considered, and the values were normalized between zero and one. As discussed in
Sect. 1, our aim was to predict the stock’s mid-price by taking two different scenarios
into consideration; (high + low)/2 and (open + close)/2. For this, we calculated the
percentage change in mid-price. Then, we applied ten different KAF algorithms on
the processed data. The empirical results obtained by applying KAF algorithms on
the NIFTY-50 dataset shows the superiority of the proposed work and sets a new
benchmark for future research.

1 https://drive.google.com/drive/u/0/folders/0B8e3dtbFwQWUZ1I5dklCMmE5M2M.
Mid-Price Prediction Using Online Kernel Adaptive Filtering 707

420

410

400
Mid-Price

390

380

370

360 Original
Prediction

350
0 50 100 150 200 250
Instances

Fig. 3 Prediction for one stock (Adaniports) using KAPA (open + close)/2

Table 2 Parameter description for close-price using ten different KAF algorithms
Algorithms σ σ2 n σ2 d η  Λ tco f f τ mu0 P nu β
LMS – – – – – – – – 0.2 – – –
FBQKLMS 5 – – 0.3 0.4 – – – – – – 0.85
KLMS 7.0 – – 1.7 – – – – – – –
KAPA 3.0 – – 1.4 1E−2 – – – 0.2 20 – –
LKAPA 5 – – 0.03 – 1E−2 – – – 20 – –
NORMA 3.0 – – – 1.5 1E−2 0 5000 – – – –
QKLMS 4 – – 0.2 0.5 – – – – – – –
KMCC 6.0 – – 1.7 – – – – – – – –
KNLMS 3.0 – – 1.5 1E−4 – – – 2 – – –
PROB-LMS – 2 3 – – 0.9 – – – – – –
σ = Kernel width , σ2 n= Variance of observation noise , σ2 d = Variance of filter weight diffusion,
η = Step-size,  = Regularization parameter, Λ = Tikhonov regularization , tcoff = Learning rate
coefficient, τ = memory size (terms retained in truncation), mu0 = Coherence criterion threshold,
P = Memory length, nu= Approximate linear dependency (ALD) threshold

4.2 Evaluation Criterion

We used standard evaluation metrics and analyzed the performance of different KAF
algorithms. These evaluation metrics use are mean squared error (MSE), and direc-
tional symmetry (DS). The different metrics are defined as:
Minimum Square Error (MSE)
It is also called the mean squared deviation (MSD), and it measures the average
squared difference between the actual and predicted observation, i.e., the average
prediction error, and it is given by:
708 S. Mishra et al.

n 
 2
MSE = yi − di (2)
i=1

Directional Symmetry (DS)

In terms of time-series analysis, DS predicts the performance using positive and
negative trends, from one time to the next.

100 
n
DS = Di (3)
n i=1

where 
0 otherwise
Di =
1 (yi − yi−1 )(di − di−1 ) ≥ 0

In above equation, di represents the predicted output, yi represents the actual values,
and n is the time-step.
Measuring the evaluation metrics with Nifty-50
1. Table 2 shows the details of the different hyperparameters and their values for
each of the algorithms.
2. With the help of standard evaluation metrics, we calculated error values and tried
to quantify the best predictive performance. Overall, we get 50 × 2 (one for each
stock) error values.
3. Finally, Tables 5 and 6 shows the predictive potential of the algorithms on all 50
stocks. We considered the average of all 50 error metrics to come up with the
final number for a single time window and a single stock.

4.3 Prediction, Convergence Analysis

In this section, we discuss the predictive performance of the KAF class of algorithms
using the convergence and prediction graphs. For this, we have shown the predic-
tion graphs for one stock (Adaniports) with the KAPA algorithm. Figure 2 shows the
prediction result for (high + low)/2 and that for (open + close)/2 is shown in Fig. 3.
It should be noted here that the graphs show the result for an individual stock (Ada-
niports) and one prediction window (sixty minutes). It can be seen from the graphs
that the predicted curve fits strongly against the original curve. Similar results were
obtained for other stocks. We can clearly see that the predictions are not accurate;
nevertheless, they are close. The prediction result that we achieved, therefore, shows
the excellent capability of the proposed work.
Figures 4 and 5 show the convergence results for mid-price using (high + low)/2
and (open + close)/2, respectively. In this case, similarly to prediction graphs, we
have taken individual stock (Adaniports) and one prediction window (sixty minutes)
Mid-Price Prediction Using Online Kernel Adaptive Filtering 709

-1

-2

-3
MSE

-4

-5

-6

-7
0 200 400 600 800 1000 1200 1400 1600 1800
Instances

Fig. 4 Error convergence for one stock (Adaniports) using KAPA (high + low)/2

-1

-2

-3
MSE

-4

-5

-6

-7
0 200 400 600 800 1000 1200 1400 1600 1800
Instances

Fig. 5 Error convergence for one stock (Adaniports) using KAPA (open + close)/2

using the KAPA algorithm. Convergence graphs show that the algorithm converges
quickly, i.e., at the 1000th data point. This is because algorithms adaptively adjusts
its weight and learn quickly from the data. Sometimes fluctuations were present in
the graphs, it is nevertheless acceptable because there will be some noise in the new
data and minor changes are inevitable. We also analyzed the performance in terms
of movement of a stock using directional symmetry. To do this, we used ten different
algorithms listed in Sect. 3. Table 6 shows the probability that we can predict the next
710 S. Mishra et al.

up or a down movement is 66%; using KNLMS, we obtained the best result at the
twenty-minute time window in terms of directional symmetry.

4.4 Comprehensive Evaluation of Algorithms

We experiment with two different aspects using ten different aspects. First, the results
are analyzed using (high + low)/2 to rule out the best algorithms. Second, we tried
mid-price using (open + close)/2. The final analysis results are present in Tables 5 and
6. Since we have used ten algorithms in our experimentation; therefore, it becomes
essential to compare their performance. We can see in terms of MSE, KAPA gives
the best performance compared to other algorithms. From Tables 5 and 6, we can
see that KAPA performed well in minimizing the error, the lowest error obtained
at one-minute time window is order of 10−4 . We got a minimum error value at the
one-minute time window. The DS results shows a contradiction. For DS, KNLMS
performed superior on all time windows. Overall, the results shows the predictive
superiority of KNLMS and KAPA algorithms.

4.5 Experimentation with Dictionary Size

The experiments were also performed by varying the dictionary size, and the results
are summarized in Table 4. We have chosen KMCC because it is insusceptible to
outliers. While experimenting with dictionary size, KMCC algorithms and sixty-
minute time windows were selected with one stock (Adaniports). The dictionary
size experiments demonstrate that if the dictionary size increases, then the system’s
performance degrades. It should be noted here that the prediction of an individual
stock with dictionary size 1000 takes only 0.41 s execution time, and it shows the of
advantage of using KAF for high-frequency trading.

4.6 Comparison with Other Methods

We compare our result with [31–33]. These are some of the popular deep learning-
based stock prediction methods. All the methods were implemented based on the
architecture details and hyperparameters setting found in the articles. Further, all
authors of the articles used LSTM-based deep learning models to predict the stock
prices. We trained all the methods on the Nifty-50 dataset. For experimentation, we
selected fifty stocks for the sixty-minute time windows to have consistency across
various methods. The result of all the methods were then compared with the proposed
KAF method. Table 3 shows the comparison of the proposed work with other existing
stock prediction methods. The 80:20 splits was used for training and testing the
Mid-Price Prediction Using Online Kernel Adaptive Filtering 711

Table 3 Comparison of the proposed work with other state-of-the-art stock prediction methods
Method MSE RMSE Execution time (s)
[31] 0.4001 0.5626 351.31
[32] 0.0148 0.0985 259.72
[33] 0.0325 0.1253 556.92
[43] 0.0163 0.1102 6392.85
Proposed method 0.0043 0.0633 4.27

Table 4 Effect of dictionary size with execution time, algorithm chosen KMCC (60-min)
Dictionary size MSE DS Execution time (s)
1000 0.00467 0.56 0.41
5000 0.00477 0.63 0.45

Table 5 Result in terms of MSE and DS for mid price (high + low)/2
Time MSE Best algorithms out of ten DS Best algorithms out of ten
window discussed (According to MSE) discussed (According to DS)
1-Day 0.0156 KAPA 0.5638 NORMA
60 min 0.0043 KAPA 0.5667 KNLMS
30 min 0.0025 KAPA 0.5582 KNLMS
25 min 0.0022 KAPA 0.5537 KNLMS
20 min 0.0018 KAPA 0.5497 KNLMS
15 min 0.0014 KAPA 0.5446 KNLMS
10 min 0.0010 KAPA 0.54447 KNLMS
5 min 0.0005 KAPA 0.5299 KNLMS
1 min 0.00012 KAPA 0.5516 KNLMS

deep learning methods. Time taken to train & test was recorded, and the results are
summarized in the last column of Table 3. The results shown in Table 3 show the
superiority of the proposed work compared to the other stock prediction methods in
literature.

5 Conclusion

In this paper, we proposed the idea of online learning-driven KAF algorithms to

predict a stock’s mid-price. The performance was analyzed on Indian National Stock
Exchange (NIFTY-50) using ten different algorithms. The experiment was performed
on nine time windows. Different time windows were considered to show the predic-
tive capability and the superiority of the work for both interday and intraday trading.
712 S. Mishra et al.

Table 6 Result in terms of MSE and DS for mid price (open + close)/2
Time MSE Best algorithms out of ten DS Best algorithms out of ten
window discussed (According to MSE) discussed (According to DS)
1-Day 0.0149 KAPA 0.6247 KMCC
60 min 0.00256 QKLMS 0.6546 KNLMS
30 min 0.00224 QKLMS 0.6636 KNLMS
25 min 0.00184 QKLMS 0.6629 KNLMS
20 min 0.00143 QKLMS 0.6670 KNLMS
15 min 0.0010 QKLMS 0.6666 KNLMS
10 min 0.0005 QKLMS 0.6638 KNLMS
5 min 0.00012 QKLMS 0.6533 KNLMS
1 min 0.00011 KAPA 0.5965 QKLMS

In the literature, the intraday time windows are generally overlooked. We therefore
tried to fill the gap by proposing the online KAF method. Experimental results also
show that the KAF algorithms outperformed similar methods in terms of execution
time and error minimization.

References

1. Abu-Mostafa YS, Atiya AF (1996) Introduction to financial forecasting. Appl intell 6(3):205–
213
2. Clements MP, Franses PH, Swanson NR (2004) Forecasting economic and financial time-series
with non-linear models. Int J Forecast 20(2):169–183
3. Sagala TW, Saputri MS, Mahendra R, Budi I (2020) Stock price movement prediction using
technical analysis and sentiment analysis. In: Proceedings of the 2020 2nd Asia pacific infor-
mation technology conference, pp 123–127
4. Rafiul Hassan Md (2009) A combination of hidden Markov model and fuzzy model for stock
market forecasting. Neurocomputing 72(16–18):3439–3446
5. Ecer F, Ardabili S, Band SS, Mosavi A (2020) Training multilayer perceptron with genetic
algorithms and particle swarm optimization for modeling stock price index prediction. Entropy
22(11):1239
6. Moghaddam AH, Moghaddam MH, Esfandyari M (2016) Stock market index prediction using
artificial neural network. J Econom Fin Adm Sci 21(41):89–93
7. Pimenta A, Nametala CAL, Guimarães FG, Carran EG (2018) An automated investing method
for stock market based on multiobjective genetic programming. Comput Econ 52(1):125–144
8. Huang T-C, Zaeem RN, Barber KS (2019) It is an equal failing to trust everybody and to trust
nobody: stock price prediction using trust filters and enhanced user sentiment on twitter. ACM
Trans Int Technol (TOIT) 19(4):1–20
9. Shian-Chang H, Chei-Chang C, Jui-Te C, Wu C-F (2020) A novel intelligent option price fore-
casting and trading system by multiple kernel adaptive filters. J Comput Appl Math 369:112560
10. Liu W, Park I, Principe JC (2009) An information theoretic approach of designing sparse kernel
adaptive filters. IEEE Trans Neural Netw 20(12):1950–1961
Mid-Price Prediction Using Online Kernel Adaptive Filtering 713

11. Min H, Shuhui Z, Xu M, Tie Q, Ning W (2018) Multivariate chaotic time series online prediction
based on improved kernel recursive least squares algorithm. IEEE Trans Cybern 49(4):1160–
1172
12. Liu W, Pokharel PP, Principe JC (2008) The kernel least-mean-square algorithm. IEEE Trans
Sig Proc 56(2):543–554
13. Chen B, Zhao S, Zhu P, Príncipe CJ (2011) Quantized kernel least mean square algorithm.
IEEE Trans Neu Netw Learn Syst 23(1):22–32
14. Ningning Z, Aijing L, Pengjian S (2017) Multidimensional k-nearest neighbor model based
on EEMD for financial time series forecasting. Phys A: Stat Mech Appl 477:161–173
15. You-Shyang C, Ching-Hsue C, Wei-Lun T (2014) Modeling fitting-function-based fuzzy time
series patterns for evolving stock index forecasting. Appl Intell 41(2):327–347
16. Abraham A, Philip NS, Saratchandran P (2004) Modeling chaotic behavior of stock indices
using intelligent paradigms. arXiv preprint: cs/0405018
17. Ritika S, Shashi S (2017) Stock prediction using deep learning. Multimedia Tools Appl
76(18):18569–18584
18. Bezerra PCS, Albuquerque PHM (2017) Volatility forecasting via SVR-GARCH with mixture
of gaussian kernels. Comput Manage Sci 14(2):179–196
19. Kumar M, Thenmozhi M (2014) Forecasting stock index returns using ARIA-SVM, ARIMA-
ANN, and ARIMA-random forest hybrid models. Int J Banking Acc Finance 5(3):284–308
20. Ju-Jie W, Jian-Zhou W, Zhe-George Z, Shu-Po G (2012) Stock index forecasting based on a
hybrid model. Omega 40(6):758–766
21. Gurav U, Sidnal N (2018) Predict stock market behavior: role of machine learning algorithms.
In: Intelligent computing and information and communication. Springer, pp 383–394
22. Qiu Y, Yang H-Y, Lu S, Chen W (2020) A novel hybrid model based on recurrent neural
networks for stock market timing. Soft Comput 1–18
23. Juan P, Francisco Z-M, Botella-Rocamora P (2015) Online learning algorithm for time series
forecasting suitable for low cost wireless sensor networks nodes. Sensors 15(4):9277–9304
24. Vrablecová P, Ezzeddine AB, Rozinajová V, Šárik S, Sangaiah AK (2018) Smart grid load
forecasting using online support vector regression. Comput Electr Eng 65:102–117
25. Marco L, Matteo B, Paolo F (2013) Short-term traffic flow forecasting: an experimental compar-
ison of time-series analysis and supervised learning. IEEE Trans Intell Transp Syst 14(2):871–
882
26. Wen L, Lu Z, Lingxiao C (2019) Deep learning-based feature engineering for stock price
movement prediction. Knowl-Based Syst 164:163–173
27. Song Y, Lee JW, Lee J (2019) A study on novel filtering and relationship between input-
features and target-vectors in a deep learning model for stock price prediction. Appl Intell
49(3):897–911
28. Chenjie S, Massimo DP (2019) Improving trading technical analysis with TensorFlow long
short-term memory (LSTM) neural network. J Fin Data Sci 5(1):1–11
29. Anass N, Abdelouahid L (2018) Short-term stock price forecasting using kernel principal
component analysis and support vector machines: the case of casablanca stock exchange.
Proced Comput Sci 127:161–169
30. Liu Z, Loo CK, Pasupa K, Seera M (2020) Meta-cognitive recurrent kernel online sequential
extreme learning machine with kernel adaptive filter for concept drift handling. Eng Appl Artif
Intell 88:103327
31. Mojtaba N, Pooyan N, Hamed J, Amir M, Ely S et al (2020) Deep learning for stock market
prediction. Entropy 22(8):840
32. Penglei G, Rui Z, Xi Y (2020) The application of stock index price prediction with neural
network. Math Comput Appl 25(3):53
33. Adil M, Mhamed H (2020) Stock market prediction using LSTM recurrent neural network.
Proced Comput Sci 170:1168–1173
34. Ehsan H, Saman H (2019) CNNpred: CNN-based stock market prediction using a diverse set
of variables. Expert Syst Appl 129:273–285
714 S. Mishra et al.

35. Wu X, Haolei C, Jianjia W, Luigi T, Vincenzo L, Hamido F (2020) Adaptive stock trading
strategies with deep reinforcement learning methods. Inf Sci 538:142–158
36. Sergio G-V, Xiao-Jun Z, John K (2020) Stock returns prediction using kernel adaptive filtering
within a stock market interdependence approach. Expert Syst Appl 160:113668
37. Sergio G-V, Xiao-Jun Z, John K (2019) Learning from data streams using kernel least-mean-
square with multiple kernel-sizes and adaptive step-size. Neurocomputing 339:105–115
38. Van Vaerenbergh S, Santamaría I (2013) A comparative study of kernel adaptive filtering
algorithms. In: 2013 IEEE digital signal processing and signal processing education meeting
(DSP/SPE). IEEE, pp 181–186
39. Zhao S, Chen B, Zhu P, Príncipe JC (2013) Fixed budget quantized kernel least-mean-square
algorithm. Sig Proc 93(9):2759–2770
40. Zhao S, Chen B, Principe JC (2011) Kernel adaptive filtering with maximum correntropy
criterion. In: The 2011 international joint conference on neural networks. IEEE, 2011, pp
2012–2017
41. Masahiro Y (2012) Multikernel adaptive filtering. IEEE Trans Sig Proc 60(9):4672–4682
42. Fernandez-Bes J, Elvira V, Van Vaerenbergh S (2015) A probabilistic least-mean-squares filter.
In: 2015 IEEE international conference on acoustics, speech and signal processing (ICASSP).
IEEE, 2015, pp 2199–2203
43. Mahla N, Gholamreza M, Jamshid B (2019) Stock price prediction using deep learning algo-
rithm and its comparison with machine learning algorithms. Intell Syst Account Fin Manage
26(4):164–174
Blockchain-Based Model for Secure IoT
Communication in Smart Healthcare

Gawhar Hameed, Yashwant Singh, Shahidul Haq, and Bharti Rana

Abstract Over the last decade, a dramatic growth has been witnessed in health-
care monitoring systems and wireless healthcare monitoring systems have been
developed in many countries of the world. Smart healthcare is one of the key
areas where IoT infrastructures and solutions are widely used to facilitate the best
possible patient surveillance, accurate diagnosis, and timely operation of patients
with existing diseases. Smart healthcare systems on the other hand, face various
challenges including data transparency, traceability, immutability, and security. Also,
most of the existing healthcare systems are often controlled by a centralized moni-
toring unit that poses potential risks of single-point failures during natural calami-
ties. Being motivated by the aforementioned issues, this paper aims at leveraging
blockchain technology for smart healthcare to overcome the issues and challenges
that smart healthcare faces as a result of traditional security policies. We propose
a theoretical Autonomous Healthcare model, which divides the whole healthcare
scenario into hospital-patient collaborations. The proposed model uses multiple
blockchains, a global blockchain (GBC), which is used for collaboration among
hospitals. Thus, creating a global healthcare system in which a patient is not limited
to a certain set of region-based hospitals and doctors. Other blockchains (RBC) are
specific to the regions of the patients for problems that are minor and require less
interaction and consultation of doctors. Global blockchain (GBC) is a permission-less
blockchain while region blockchains (RBC) are permissioned ones. The proposed
model promises a decentralized framework that facilitates sharing and exchanging,
and integration of information across all the users and third parties.

Keywords Internet of t.hings (IoT) · Smart healthcare · Security attacks ·

Blockchain

G. Hameed · Y. Singh · S. Haq · B. Rana (B)

Department of Computer Science and Information Technology, Central University of Jammu,
Jammu and Kashmir, Jammu 181143, India
Y. Singh
e-mail: yashwant.csit@cujammu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 715
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_52
716 G. Hameed et al.

1 Introduction

The growth of the “Internet of things (IoT)” and related technologies has resulted
in the development of smart systems such as: smart cities, smart transport systems,
smart energy, and so on, where all objects (living and non-living) are interconnected
to form a physical network in which all processes such as sensing, processing, and
communication are automatically controlled and managed without human interven-
tion [1]. With a plethora of space for improvement, the healthcare industry has grown
tremendously by gaining some level of automation. The current healthcare industry
has advanced beyond hospitals by bringing together medical devices, healthcare
providers, and patients [2].
The Internet of things has brought about a significant transformation in the field
of healthcare by changing the conventional healthcare systems to the modern smart
healthcare system (SHS) [3]. SHS are developed to deal directly with the patient
health-related data [4]. SHS provides in-depth insights about disease symptoms and
reduces the need for routine health check-ups, which can benefit elderly people,
diabetes patients, and others with secure management.
Although SHS offers various benefits, there are also some issues that need to be
addressed. Security is the major concern in smart healthcare monitoring systems
because of the massive amount of sensitive patient data being transferred across
the Internet. The data is vulnerable to several security attacks [5]. Secondly, smart
healthcare systems are often controlled by a centralized monitoring unit. In case of
system failures, the stored data is inaccessible and there might be a chance of data
theft or corruption of data. The existing techniques are not sufficient in addressing
these security issues [6]. The purpose of the proposed model is to create a secure
blockchain-based global healthcare system in which a patient is not limited to a
certain set of region-based hospitals and doctors. Thus, provides global access to
patient information from any region at any time.
The acronyms used in the article are described in Table 1. The organization
of rest of the paper is as follows Sect. 2 discusses the literature work. Section 3
describes the security requirements and challenges in smart healthcare. Section 4
describes blockchain technology, its working, and various benefits in smart health-
care. Section 5 describes the proposed Autonomous Smart Healthcare model.
Section 6 presents the conclusion and future work.
The research contributions of the study are as follows:
• Security requirements and challenges in smart healthcare have been thoroughly
discussed.
• Blockchain utilization with various features in smart healthcare systems has been
explained.
• A theoretical Autonomous Smart Healthcare model using blockchain has been
proposed.
Blockchain-Based Model for Secure IoT Communication … 717

Table 1 Acronyms and their meaning

Acronym Meaning
SHS Smart healthcare system
IoT Internet of things
GBC Global blockchain
RBC Regional blockchain
CH Cluster head
TX Transaction
D Doctor
P Patient
PUK Public key
K Shared key
H1,1 Hospital 1 of cluster 1
PK Private key
EP Encrypted policy
P1,1,1 D1,1,1 Access details of patient 1 of hospital 1 of cluster 1 to doctor 1 of hospital 1 of
cluster 1
TXID Transaction identity
PHash Previous hash

2 Literature Survey

In El Zouka and Hosni [7], a secure authentication scheme for remote health moni-
toring of patients is presented. The presented system enables doctors to monitor
the vital signals of the patient. The authentication mechanism is categorized into
three steps namely, registration, login, and authentication. The fuzzy logic inference
method is used for the processing of sensor signals. The author proves the reliability
and effectiveness of the system.
In Hossain and Muhammad [8], the authors devise a remote health monitoring
system. Vital signals are collected by the smart objects from the body of a patient
and transmitted to a module used for analyzing the signals and taking appropriate
decisions. Several smart devices are utilized to collect the voice and other signals
of a patient. The captured signals are processed individually and serve as an input
for the emotion detection system. These signals are then combined to provide a final
value for making decisions regarding emotion.
In Chen et al. [9], a remote health monitoring system based on any cast routing is
presented. The system chooses the nearest receiver automatically as its destination
from a group of any cast to minimize the transmission delay and the control overhead.
The delay of the path recovery is also minimized by the new protocol using route
recovery from the intermediary routers belonging to the initial path. According to the
experimental results obtained, the authors believe that the scheme is fast and reliable.
718 G. Hameed et al.

In Shakeel et al. [10], the authors provide a secure framework for ensuring the
security and privacy of the transmitted health data over the Internet. A learning-
based system is developed for minimizing malware attacks on health data. Medical
data is examined across different layers based on the Q-learning approach to reduce
malware attacks. The performance of the suggested scheme is estimated based on
experimental outcomes.
In Elhoseny et al. [11], a security framework has been presented that provides
a way to securely transmit health data in healthcare systems. The working of this
model is based on four steps: first, the health data is encrypted by making use of a
hybrid encryption technique. This encrypted data is hidden behind a cover image to
generate a stego image. To recover the original data, embedded information is then
extracted and decrypted. This proposed system was implemented in the MATLAB
setup using simulation and the results were satisfactory.

3 Smart Healthcare: Security Requirements

and Challenges

The wireless sensors and wearables collect vital information about the patient for
predictive analysis. An unauthorized access to patient’s data can lead to a catastrophe,
not only compromising the privacy of patients but also puts their lives at risk. There-
fore the security challenges encountered in smart healthcare are discussed subse-
quently. Due to the emerging security vulnerabilities during data transmission, smart
healthcare necessitates to satisfy the various security requirements [12] as shown in
Table 2.
• Centralized storage
Healthcare systems are often controlled by a centralized monitoring unit. In
case of system failures, the stored data is inaccessible and there might be a chance
of data theft or corruption of data [14].
• Lack of synchronization
Maintaining synchronization among various entities of the Smart Healthcare
systems is a difficult task [10]. Since all the modules work independently from

Table 2 Security requirements for smart healthcare monitoring systems [13]

Requirement Description
Confidentiality Guarantees that patient health data is protected from unauthorized access by
third parties
Integrity Ensures that the patient data that has been collected is not modified
Availability Ensures healthcare data is available on-demand without delay
Anonymity Ensures the confidentiality of patient identity
Secure data transit Ensures the data is not modified while it is in transit
Blockchain-Based Model for Secure IoT Communication … 719

each other, and have their own standards. Therefore, combining the data is a
challenging task [15].
• Scalability
Smart healthcare systems expand or contracts depending upon the need of
patients. The existing system must not be compromised when the system is
modified [16].
• Availability
Historical records and current signals of the patient must be available to
the examiner at any time and location. Based on the current technology these
operations are likely unattainable.

4 Blockchain

A blockchain is a synchronized, shared ledger that includes linked blocks of trans-

actions. BC is a distributed and shared ledger that records all transactions without
involving a third party [17]. The first block in blockchain is called the genesis block.
Except the genesis block, every other block contains the previous block’s hash. Each
participating node in a blockchain network contains a copy of blockchain and when-
ever a new block is attached to the blockchain, changes are reflected in all other
blocks. The copies of blockchain get updated, preventing an intruder from modi-
fying the block [18]. Each node in the network possesses two keys: a public key used
to encrypt the messages sent to a node and a private key to decrypt the messages
and allow a node to read them. Therefore, blockchain uses public key encryption to
ensure non-repudiability and consistency, and irreversibility. The encrypted messages
can only be decrypted with an appropriate private key [19]. The whole structure of
blockchain is shown in Fig. 1.

4.1 Working of Blockchain

Blockchain is based on four key components including [20]: (a) a peer-to-peer

network: in which network nodes communicate with each other with the help of
a pair of keys (private/public). Transactions are signed using a private key, and the
public key is used as an address to be reachable on the network. (b) Distributed ledger:

Genesis Block Block i Block i+1

Block Header Block Header Block Header
Hash Hash Hash
Previous Hash Hash Previous Hash Previous Hash
Transacon Transacon Transacon

Fig. 1 Structure of blocks in a blockchain [19]

720 G. Hameed et al.

Block broadcasted
to the network

Start
PoW
Processing
Transaction Transaction Miner Miner 2 End
Miner n
request added to block

Miner n solves
PoW

PoW is broadcasted to peer

to peer network

Transaction
successful Block is added to the chain Verified

Fig. 2 Working of blockchain [23]

The ledger is open and public to everyone. (c) Consensus mechanism: Consensus is a
method to ensure that the network nodes verify the transactions and agree with their
order and existence on the ledger [21]. Proof of work (PoW) and proof of stake (PoS)
are the most common ones. (d) Public key cryptography: To chain blocks together,
the data in a block is subjected to a particular function called a ‘cryptographic hash.’
which creates a unique output for a specific input [22]. The various steps involved
in the blockchain is demonstrated in Fig. 2.
1. A user requests a transaction.
2. The transaction is added to the block and broadcasted to the network.
3. Miner nodes perform mining to add the transaction into the network. The miner,
who first solves the computation, broadcasts the newly created block into the
network.
4. The block is distributed to peer-to-peer network consisting of computers called
nodes.
5. The transaction is validated by network nodes with the support of algorithms.
After validation, the transaction is verified with the help of cryptocurrency.
6. Now the transaction is merged with another transaction to create a new block.
7. This new block is attached to the already existing blockchain without being
modified.
8. Transaction is successful.
Blockchain-Based Model for Secure IoT Communication … 721

Physical
Layer
S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn S1,S2,.Sn
V1 V1 V1 V1 V1 V1 V1 V1 V1

Layer
P1,2,1 P1,n,1

Patient
P1,1,1 P1,1,2 P1,2,2 P1,n,2 P1,1,3 P1,2,3 P1,n,3

Healthcare Blockchain Layer

H1,1 H1,2 H1,3

RBC2

RBC3
RBC1

GBC
H2,1 Hn,1 H2,2 Hn,2 Hn,3
H2,3

Doctor Layer
D1,1,1 D1,2,1 D1,n,1 D1,1,2 D1,2,2 D1,n,2 D1,1,3 D1,2,3 D1,n,3

A1 A1 A1 A1 A1 A1 A1 A1 A1

Fig. 3 Proposed autonomous model for smart healthcare using blockchain

4.2 Smart Healthcare: Blockchain Benefits

The smart healthcare sector is a perfect fit for blockchain technology because it has
the ability to solve the security challenges that smart healthcare systems face [24].
Blockchain technology has numerous benefits that could be used in smart healthcare
for secure communication as shown in Fig. 3.
• Decentralized Storage
Decentralized storage is the key advantage of blockchain technology. It is the
fundamental basis for improved security and authentication of data stored inside
the network. The content that is added to the blockchain is not controlled by any
central authority. Data storage is split between different servers, which allows
quicker access to medical data, better data quality, and device interoperability
[25].
• Security
From various perspectives, blockchain is more trustworthy and secure than
existing record-keeping systems. When a transaction is approved, it is encrypted
and attached to the preceding transaction. Furthermore, rather than storing infor-
mation on a single server, information is distributed throughout a network of
computers, preventing hackers from accessing transaction data [26].
• Persistency
722 G. Hameed et al.

Persistency is the key feature of blockchain. Because of the distributed ledger,

which is stored across multiple nodes, deletion of records after they have been
added onto the blockchain is almost impossible [22].
• Greater transparency
Blockchain transaction records are more transparent because they are available
to all users of the network. As a result, all blockchain users have equal access to
the network, allowing them to link, verify, and trace transaction activities [27].
Table 3 provides a comparative analysis of existing frameworks based on
blockchain technology for providing security solutions in Smart Healthcare systems
considering parameters such as the objectives, framework, merits, and limitations.
From the above table, we analyze that the existing smart healthcare monitoring
systems lack in terms of patient mobility [30] and privacy of the patient’s health
data [31–34]. Also most of the existing systems relay on the centralized architecture
which adds problems in case of natural calamities. Keeping these points in mind, we
must develop smart healthcare system in which the patient is not limited to a certain
set of region-based hospital but it should be a kind of global healthcare system [35–
39] so that patients can connect from anywhere and based on their need, treatment
is provided. In case they need to connect to a different doctor, the system must be
capable of providing them the desired service without compromising their health
data.

5 Proposed Autonomous Smart Healthcare Model Using

Blockchain

Taking into consideration the need of an autonomous healthcare model, our proposed
model divides the whole Healthcare scenario into hospital-patient collaborations as
shown in Fig. 3. We have used multiple blockchains in our architecture, one is a
global blockchain (GBC), which is used for collaboration of all the hospitals, thus
creating a global healthcare system in which a patient is not limited to a certain set
of region-based hospitals and doctors, other blockchains (RBC) are specific to the
regions of the patients for problems that are minor and require less interaction and
consultation of doctors. Global blockchain (GBC) is a permission-less blockchain
while as region blockchains (RBC) are permissioned ones.
The proposed model consists of four layers physical layer, patient layer, health-
care blockchain layer, and doctor’s layer. Physical layer consists of the sensors used
to monitor the health of patients. Patient layer consists of all the patients of all
the participating hospitals. Healthcare blockchain layer consists of all the partici-
pating hospitals in a blockchain and doctors layer consists of all the doctors of the
participating hospitals. Every participating hospital in a region consists of Regional
Blockchain (RBC). All the region-specific health clusters collaborate in a way to
form a global healthcare model with the creation of a global blockchain (GBC).
Blockchain-Based Model for Secure IoT Communication … 723

Table 3 Analysis of existing blockchain-based frameworks for smart healthcare [27]

Author Year Objective Framework/Model Merits Limitations
Mcghin 2020 To design and BHEEM The Implementation
et al. [25] analyze a smart functioning of of smart
health monitoring several contracts are not
system subsystems of defined formally
healthcare is
defined and
simulated with
well-defined
algorithms
Bigini et al. 2020 To propose Smart contract Improve the The cost linked
[21] various gathering, use, with utilizing
workflows for and sharing of smart contracts
efficient data patient data for smart
handling in the and data healthcare
healthcare sub-processors should be
environment calculated
using blockchain
technology
Hathaliya 2020 To design a eHealth The system is Lacks dynamic
et al. [23] blockchain-based architecture fault-tolerant storage selection
decentralized and reliable algorithm and
eHealth patients privacy
architecture
Kumar et al. 2020 To design, IHIoT model Various factors Lack of
[1] simulate and taken into implementation
analyze the consideration over different
performance of including blockchain
blockchain-based design, networks having
smart health care security, and different tools
system framework and techniques
Nasiri [28] 2019 Proposed a Key agreement Access time Patient mobility
framework that protocol and key has not been
integrates AI to generation considered
make healthcare time were
system smart and reduced
secure
Hathaliya 2019 A patient Machine learning Acceptable High
and Tanwar monitoring localization information
[29] system based on accuracy with retrieval delay
a wireless sensor low processing and redundant
network time data
(continued)
724 G. Hameed et al.

Table 3 (continued)
Author Year Objective Framework/Model Merits Limitations
Ray et al. 2019 A detailed S2SH framework The The integration
[26] framework for a application of of various
smart healthcare blockchain to subsystems must
system is the Indian be validated
developed healthcare before approval
system is
examined
in-depth
Al Karaki 2019 To develop a DASS-CARE Provides Validated is
et al. [6] framework based real-time missing
on blockchain access and
(DASS-CARE) preserving
integrity,
confidentiality,
and security of
patient
information
Velliangiri 2018 To develop an Ethereum protocol Overcomes a Patient privacy
and architecture that number of needs to be
Karthikeyan facilitates secure security improved
Karunya smart sensor vulnerabilities
[22] management in smart
healthcare
monitoring
Proposed 2021 To develop a Smart healthcare Decentralized –
model smart global model using model with
healthcare blockchain better security,
monitoring and
system transparency
of patient data

Let the collaborating hospitals in a region be identified as H = {H 1,1 , H 2,1 ,

….H n,1 }, where H 1,1 is identified as hospital 1 of cluster 1, H 2,1 is identified as
hospital 2 of cluster 1, and so on. Let the patients of a specific cluster be represented
by the set P = {P1,1,1 , P1,2,1 , ….P1,n,1 }, where P1,2,1 is identified as patient 1 of
hospital 2 of cluster 1. The set of doctors in a specific cluster be represented by set
D = {D1,1,1 , D1,2,1 , …D1,n,1 } whereas D1,2,1 represents the doctor 1 of hospital 2
of cluster 1. The set V 1 of patient 1 represents the variables monitored by various
wearable sensors such as ECG sensor, Oximeter sensor, Glucometer sensor, etc.,
and set A1 represents the actions prescribed by the doctor after analyzing the V 1 set.
These two sets can be described as V 1 = {sd1 , sd2 , …sdn } and A1 = {a1 , a2 , …an }.
We have described the three workflows of the model which include register patient,
check patient, and delegate patient. These workflows are discussed below as:
1. Register Patient
Blockchain-Based Model for Secure IoT Communication … 725

A patient registers with a particular hospital through hospital portal or manu-

ally by a hospital receptionist by providing details which typically include
patient’s name, address, disease or problem and various other identifiers such as
age, height, weight, etc. The transaction being encrypted by the hospital public
key is represented as EncryptHijPub (Pname , Paddress , Pdisease , Pgender , Page , Pheight ,
Pweight ). The particular hospital creates this transaction of the patient details and
broadcasts to the regional hospital cluster. The cluster node adds the transaction
to the blockchain through some consensus mechanism such as ‘proof of work’
or ‘proof of stake’. After the transaction is added to the regional blockchain
(RBC), RBC sends a notification to the particular doctor by a smart contract
policy. In the end the doctor sends a confirmation message to the patient. All the
subsequent transactions between the doctor and the patient are encrypted using
the shared secret key provided by the concerned hospital. The encrypted trans-
action can be represented as EncryptSK(Pi,j,k Di,j,k = (Pname , Pid , Ppuk , Dname ,
Did , Dpuk )). The proposed patient registration is illustrated using a workflow as
shown in Fig. 4.

2. Check Patient
After registration the patient send a hello message to the doctor by providing
a registration transaction Id. The doctor then issue a monitor vitals action on
the patient. The vitals are monitored using various wearable sensors that are
used by the patient. After a span of time, the patient’s wearable sensors sends
the vitals data to the concerned hospital. The participating clusters add the data
to the blockchain in the form of a transaction. After the data is added to the
blockchain, doctor is notified and gets an access to the patient’s vital data.
The doctor suggests further actions on the patient by issuing a transaction to
the blockchain. The wearable sensors are represented as V i = {s1 , s2 , s3 …}.
The transactions are represented as Encrpytsk (Pi,j,k Di,j,k V 1 = {sd1 ,sd2 ,…sdn },
A1 = {a1 , a2 , …an }), where sd1 represents as sensor data one, Pi,j,k patient

Fig. 4 Workflow of patient registration

726 G. Hameed et al.

Fig. 5 Workflow of patient check

I of hospital j of cluster k. V 1 and A1 represents first iteration of vital values

and actions. These two sets get evolved continuously until a stable set of V 1 is
obtained and the doctor is sure that patient does not need any more consultation.
The whole scenario is illustrated using a workflow as shown in Fig. 5.

3. Delegate Patient
If a doctor at some point feels to refer the patient to another expert doctor
in the field. The doctor can create a referral transaction to RBC server as
EncrpytHxPuk (Pgid , Did , R, Sk), where H xpuk refers to the public key of current
hospital, Pgid is the transaction id of the genesis block of patient. Did is the
doctors id, R is the reason for the referral, and Sk is the secret key of doctor
and patient. The concerned hospital is notified about the referral transaction and
delegate patient transaction is created in which the patient’s records are added in
a global blockchain (GBC). The concerned hospital cluster informs the referred
doctor. After addition of the block to the GBC, based on the smart contract
policy the concerned hospital gets notified, the hospital therefore assigns the
doctor to the patient by providing a new secret key and the genesis transaction
Id of the previous records of the patient as illustrated in Fig. 6.

Algorithms
Below discussed algorithms provide an idea how a transaction is validated and added
to the blockchain.
Blockchain-Based Model for Secure IoT Communication … 727

Fig. 6 Workflow of delegate patient

Algorithm 1: Validating a Transaction

ValTx
Begin
1. Patient P1 sends a request to Hij
2: Hij extracts Patients details and creates a transaction Tx as
3: Tx=( Pname, Ptype, PService, Ppuk, PHash),
4: Broadcast Tx to other Hospital Clusters Hcx
5: Set hash Difficulty=5
6: Tx.hash = Tx.compute_hash()
7: while not Tx.hash.Difficulty=5
8: Tx.hash = Tx.compute_hash()
9: if(addRBC(Tx)==1)
9: return True
End
728 G. Hameed et al.

Algorithm 2: Adding Transaction to Blockchain

addRBC(Tx)
Begin
1: Generate a block ‘Blk’ by adding Tx
2: PreviousHash = LastBlockHash // previous hash assigned to last
block hash
3: if PreviousHash!= Blk.PreviousHash
4: return False
5: else
6: RBC.append(blk)
7: return True
End

6 Conclusion

Internet of things (IoT) has emerged as a key technology for the development of SHS
over the past few years. SHS have been developed to provide real-time monitoring of
patients. However, SHS are exposed to several security risks including information
theft, security attacks and privacy of patient’s data. In this study, we discussed various
security issues that SHS faces as a result of traditional security policies. We also
discussed the features of blockchain that can be utilized in healthcare systems for
managing health data in a transparent, decentralized, and traceable manner. Finally,
we proposed a theoretical blockchain-based Autonomous smart healthcare model
which divides the whole Healthcare scenario into hospital-patient collaborations. The
proposed model uses a permission-less global blockchain (GBC) for collaboration
of all the hospitals, thus creating a global healthcare system in which a patient is not
limited to a certain set of region-based hospitals and doctors. Permissioned regional
blockchains (RBC) are used for specific regions of the patients. All the region-
specific health clusters collaborate in a way to form a global healthcare model with
the creation of a global blockchain (GBC).
In future, we will implement the proposed model using solidity framework. Also,
we have described the three workflows of the model which include register patient,
check patient, and delegate patient. In future, more workflows and functionalities
will be added.

References

1. Kumar A, Krishnamurthi R, Nayyar A, Sharma K, Grover V, Hossain E (2020) A novel smart

healthcare design, simulation, and implementation using healthcare 4.0 processes. IEEE Access
8:118433–118471. https://doi.org/10.1109/ACCESS.2020.3004790
Blockchain-Based Model for Secure IoT Communication … 729

2. Akkaş MA, Sokullu R, Ertürk Çetin H (2020) Healthcare and patient monitoring using IoT.
Internet of Things 11:100173. https://doi.org/10.1016/j.iot.2020.100173
3. Hathaliya JJ, Tanwar S, Tyagi S, Kumar N (2019) Securing electronics healthcare records in
healthcare 4.0: a biometric-based approach R. Comput Electr Eng 76:398–410. https://doi.org/
10.1016/j.compeleceng.2019.04.017
4. El Zouka HA, Hosni MM (2019) Secure IoT communications for smart healthcare monitoring
system. Internet of Things 100036. https://doi.org/10.1016/j.iot.2019.01.003
5. Aggarwal S, Chaudhary R, Singh G, Kumar N, Choo KR, Zomaya AY (2019) Blockchain
for smart communities: applications, challenges and opportunities. J Netw Comput Appl
144(July):13–48. https://doi.org/10.1016/j.jnca.2019.06.018
6. Al-Karaki N, Gawanmeh A, Ayache M, Mashaleh A (2019) DASS-CARE: a decentralized,
accessible, scalable, and secure healthcare framework using blockchain. In: 15th interna-
tional wireless communications and mobile computing conference IWCMC 2019, pp 330–335.
https://doi.org/10.1109/IWCMC.2019.8766714
7. El Zouka HA, Hosni MM (2019) Secure IoT communications for smart healthcare monitoring
system. Internet of Things xxxx:100036. https://doi.org/10.1016/j.iot.2019.01.003
8. Hossain MS, Muhammad G (2018) Emotion-aware connected healthcare big data towards 5G.
IEEE Internet Things J 5(4):2399–2406. https://doi.org/10.1109/JIOT.2017.2772959
9. Chen SK et al (2012) A reliable transmission protocol for zigbee-based wireless patient moni-
toring. IEEE Trans Inf Technol Biomed 16(1):6–16. https://doi.org/10.1109/TITB.2011.217
1704
10. Shakeel PM, Baskar S, Dhulipala VRS, Mishra S, Jaber MM (2018) Maintaining security and
privacy in health care system using learning based deep-Q-networks
11. Elhoseny M, Ramírez-González G, Abu-Elnasr OM, Shawkat SA, Arunkumar N, Farouk A
(2018) Secure medical data transmission model for IoT-based healthcare systems. IEEE Access
6(c):20596–20608. https://doi.org/10.1109/ACCESS.2018.2817615
12. Farahani B, Firouzi F, Chang V, Badaroglu M, Constant N, Mankodiya K (2018) Towards fog-
driven IoT eHealth: promises and challenges of IoT in medicine and healthcare. Futur Gener
Comput Syst 78:659–676. https://doi.org/10.1016/j.future.2017.04.036
13. Moosavi SR et al (2015) SEA: a secure and efficient authentication and authorization archi-
tecture for IoT-based healthcare using smart gateways. Procedia Comput Sci 52(1):452–459.
https://doi.org/10.1016/j.procs.2015.05.013
14. Tariq N, Qamar A, Asim M, Khan FA (2020) Blockchain and smart healthcare security: a survey.
Procedia Comput Sci 175(2019):615–620. https://doi.org/10.1016/j.procs.2020.07.089
15. Islam SMR, Kwak D, Kabir MH, Hossain M, Kwak KS (2015) The internet of things for
health care: a comprehensive survey. IEEE Access 3:678–708. https://doi.org/10.1109/ACC
ESS.2015.2437951
16. Lu Y, Sinnott RO (2019) Security and privacy solutions for smart healthcare systems. Elsevier
Inc.
17. Eklund JM (2019) Blockchain technology in healthcare: a systematic review. https://doi.org/
10.3390/healthcare7020056
18. Hussien HM, Yasin SM, Udzir SNI, Zaidan AA, Zaidan BB (2019) A systematic review for
enabling of develop a blockchain technology in healthcare application: taxonomy, substantially
analysis, motivations, challenges, recommendations and future direction
19. Griggs KN, Ossipova O, Kohlios CP, Baccarini AN, Howson EA, Hayajneh T (2021) Healthcare
blockchain system using smart contracts for secure automated remote patient monitoring.
2018:1–7
20. Id AP, Tapas N (2018) Blockchain and IoT integration: a systematic survey
21. Bigini G, Freschi V (2020) A review on blockchain for the internet of medical things: definitions,
challenges, applications, and vision. https://doi.org/10.3390/fi12120208
22. Velliangiri S, Karthikeyan Karunya P (2020) Blockchain technology: challenges and security
issues in consensus algorithm. In: International conference on computer communication and
informatics, ICCCI. https://doi.org/10.1109/ICCCI48352.2020.9104132
730 G. Hameed et al.

23. Hathaliya J, Sharma P, Tanwar S, Gupta R (2019) Blockchain-based remote patient monitoring
in healthcare 4.0. pp 87–91
24. Tripathi G, Abdul M, Paiva S (2019) Healthcare S2HS—a blockchain based approach for smart
healthcare system. Healthcare 100391. https://doi.org/10.1016/j.hjdsi.2019.100391
25. Mcghin T, Choo KR, Zhechao C, He D (2019) Blockchain in healthcare applications: research
challenges and opportunities. J Netw Comput Appl 135:62–75. https://doi.org/10.1016/j.jnca.
2019.02.027
26. Ray PP, Member S, Dash D, Salah K, Member S (2020) Blockchain for IoT-based healthcare:
background, consensus, platforms, and use cases. 1–10
27. Yaqoob I (2021) Blockchain for healthcare data management: opportunities, challenges, and
future recommendations. Neural Comput Appl 0123456789. https://doi.org/10.1007/s00521-
020-05519-w
28. Nasiri S (2019) Security requirements of Internet of Things-based healthcare system: a survey
study. 27(4):253–258. https://doi.org/10.5455/aim.2019.27.253-258
29. Hathaliya JJ, Tanwar S (2020) An exhaustive survey on security and privacy issues in health-
care 4.0. Comput Commun 153(February):311–335. https://doi.org/10.1016/j.comcom.2020.
02.018
30. Ambarkar SS, Shekokar N (2020) Toward smart and secure IoT based healthcare system, vol
266. Springer International Publishing
31. Yeh KH (2016) A secure IoT-based healthcare system with body sensor networks. IEEE Access
4:10288–10299. https://doi.org/10.1109/ACCESS.2016.2638038
32. Newaz AKMI, Sikder AK, Rahman MA (2020) A survey on security and privacy issues in
modern healthcare systems: attacks and defences. 1(1):1–40
33. Mawgoud AA, Karadawy AI, Tawfik BS A secure authentication technique in Internet of
medical things through machine learning
34. Reena JK, Parameswari R (2019) A smart health care monitor system in IoT based human
activities of daily living: a review. In: International conference on machine learning, big data,
cloud parallel computing, pp 446–448
35. Koutras D, Stergiopoulos G, Dasaklis T, Kotzanikolaou P, Glynos D, Douligeris C (2020)
Security in IoMT communications: a survey. pp 1–49. https://doi.org/10.3390/s20174828.
36. Redondi A, Chirico M, Borsani L, Cesana M, Tagliasacchi M (2013) An integrated system
based on wireless sensor networks for patient monitoring, localization and tracking. Ad Hoc
Netw 11(1):39–53. https://doi.org/10.1016/j.adhoc.2012.04.006
37. Luo E, Bhuiyan MZA, Wang G, Rahman MA, Wu J, Atiquzzaman M (2018) Privacyprotector:
privacy-protected patient data collection in IoT-based healthcare systems. IEEE Commun Mag
56(2):163–168. https://doi.org/10.1109/MCOM.2018.1700364
38. Kaur PD, Chana I (2014) Cloud based intelligent system for delivering health care as a
service. Comput Methods Programs Biomed 113(1):346–359. https://doi.org/10.1016/j.cmpb.
2013.09.013
39. Yang G et al (2018) IoT-based remote pain monitoring system: from device to cloud platform.
IEEE J Biomed Heal Informatics 22(6):1711–1719. https://doi.org/10.1109/JBHI.2017.277
6351
Internet eXchange Points: A Systematic
Case Study in Indian Context

Deepti Saraswat, Pronaya Bhattacharya, Samprat Bhavsar, Hardik Dave,

Karan Patel, Shivam Punjani, Ashwin Verma, and Sudeep Tanwar

Abstract At present, Internet eXchange points (IXPs) drive the communication

backbone of modern Internet, that allow the Internet to handle the high volume of
data traffic. IXPs support critical infrastructures like clouds, data centers, and hetero-
geneous networks, and thus, it is predicted that the IXPs crossing traffic volume would
reach 5016 exabytes by 2030. To cater to this high demand, researchers globally have
addressed novel solutions of IXPs peering management, and service orchestration in
clouds, data centers, and computational servers. IXP simplifies peering connections
of different autonomous system (AS) by reducing the portion of Internet service
providers (ISPs) traffic via upstream transit provider of latter, thereby minimizing
data rate consumption cost. Although IXPs have been the resilient backbone of the
modern Internet, research on the study of IXP is still in its early stages. In IXPs, the
data packets follow an optimum path from source to reach to the destination through

D. Saraswat · P. Bhattacharya (B) · S. Bhavsar · H. Dave · K. Patel · S. Punjani · A. Verma ·

S. Tanwar
Department of Computer Science and Engineering, Institute of Technology, Nirma University,
Gujarat, India
e-mail: pronoya.bhattacharya@nirmauni.ac.in
D. Saraswat
e-mail: deepti.saraswat@nirmauni.ac.in
S. Bhavsar
e-mail: 19mcec02@nirmauni.ac.in
H. Dave
e-mail: 19mcec05@nirmauni.ac.in
K. Patel
e-mail: 19mcec10@nirmauni.ac.in
S. Punjani
e-mail: 19mcec19@nirmauni.ac.in
A. Verma
e-mail: ashwin.verma@nirmauni.ac.in
S. Tanwar
e-mail: sudeep.tanwar@nirmauni.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 731
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_53
732 D. Saraswat et al.

multiple hops via different AS networks. Thus, motivated by the aforementioned

discussions, the article presents a systematic study of IXPs in the Indian context,
and have analyzed the IXPs performance on publicly available data sets. We have
presented comparative analysis in terms of connected AS average data rate, and user-
level context bifurcations. Thus, the analysis intends to be a starting point for building
smart solutions for the creation of robust IXPs that improves network services at a
global scale.

Keywords Internet · Autonomous systems · Exchange points · Peering links ·

Communication services

1 Introduction

The term ‘Internet’ is defined as a group of interconnecting networks that is

autonomous and provides a variety of resources to the end user. In this large pool of
available networks, the efficient interconnection and architecture among them is a
growing topic among researchers. Modern Internet has been spread across the globe
in three tiers acting as pillars of Internet traffic transmission. Tier 1 network utilizes
Internet protocol (IP) and provides capabilities to connect to any interconnect net-
work (i.e., Internet) via a settlement-free interconnection also known as peering [1].
This means that two Tier 1 networks can exchange data or traffic without any fees.
In the hierarchical structure of the Internet, tier 2 and tier 3 networks are responsible
to pay the IP transit fees for propagating their traffic to the rest of the Internet. The
backbone of internet constitutes a three-tier network architecture. Figure 1 shows the
architecture of IXPs.
• At Tier 1, Internet backbone transmits the data traffic through the core routers. Tier
1 networks maintains the principal data routes between large computer networks
[1]. These routers are maintained mainly by the government bodies, commercial
and IXPs.
• Tier 2 ISP provides an interface between Tier 1 and Tier 3 networks. Tier 2 ISPs
have regional or country reach.
• Tier 3 ISPs are closest to the end users and provides Internet connection by charging
nominal amount. Based on data traffic distribution and generation, Tier 3 ISPs pays
to Tier 2 ISPs.
IXPs are the fundamental part of the critical infrastructure and handles the peering
links between different networks. It provides the platform for interconnection to
ensure low-latency data transit via Internet backbone infrastructure. Nowadays, most
of the business and consumer Internet data traffic is handled by ISPs. These ISPs
form tier 2 or tier 3 level of Internet layer and have to pass through a large setup of
physical layer connecting the continents.
These networks face long round-trip time (RTT) values (higher latency) and result
in decreased performance of Internet connection. The main aspect here is the routing
A Systematic Study of Internet eXchange Points … 733

Fig. 1 Architecture of internet eXchange Points

capabilities of the network as well as the distance and most importantly, the number of
hops it has to pass. Using IXPs, the latency can be reduced by bringing participating
ISPs to the same platform. The routing performance can be further enhanced using
the concept of peering links, which can be implemented using IXPs. Border gateway
protocol (BGP) is an exterior routing protocol mainly used at the edge routers of
the AS. BGP is used for exchanging the routing information between the routers
of different AS. There are two variants of BGP protocol, viz., interior BGP (iBGP)
and exterior BGP (eBGP). iBGP is used to exchange routing information within the
AS, whereas eBGP or normal BGP is used for communication with external routers.
The BGP routing configuration mainly depends on the routing and peering policies
decided by the organization. The BGP is a path-vector routing protocol that utilizes
path information and network policies to make routing decisions. BGP significantly
maintains a graph that is revised based on the routing information transferred between
different internal or external BGP routers. Table 1 presents the list of abbreviations
used in the paper.

1.1 Motivation

The study of Internet backbone infrastructure and IXPs has been a topic of interest
for researchers globally. In the current scenario, a large amount of data traffic is
forwarded between different AS, and thus the overall analysis of IXP and its asso-
ciated peering links is of paramount importance. To date, researchers globally have
734 D. Saraswat et al.

Table 1 Abbreviations and their meanings

Abbreviation Meaning
AS Autonomous systems
BGP Border gateway protocol
CAIDA Center for applied internet data analysis
eBGP External BGP
iBGP Internal BGP
IXP Internet eXchange points
ISP Internet service providers
NIXI National internet exchange of India

proposed solutions that optimize the bandwidth and link load management, routing
control, and traffic distributions. We have focused on orchestration of different infras-
tructure, analysis, and their management as a coherent unit. To address the research,
the paper presents a systematic analysis of IXPs and analyzes the relation of IXP and
AS with iBGP and eBGP routing protocols. For the analysis part, we have considered
the Indian IXP regions and have presented the analysis. The study is directed toward
effective IXP management (load balancing of IXPs), and AS distribution with the
consideration of traffic points.

1.2 Research Contributions

The research contributions of the article are as follows.

1. We provide the comparative analysis for different parameters of Internet backbone
infrastructure, IXPs, and AS. The analysis helps to provide the current scenario
of IXP ecosystems and the complexity of the Internet structure.
2. We analyze the relationship between the IXPs and AS and integrate the routing
protocols such as iBGP and eBGP to study data flow through the Internet networks,
utilizing Tier 1 networks and IXP infrastructure. We study the various properties
of IXPs, AS, and routing in the Internet using BGP and provided a comparative
analysis for the above parametric components on a global scale.
3. Considering Internet networking and current growth of IXP and its efficacy, we
present a comparative analysis of the IXP data as well as IXP to AS relationship in
the context of Indian regions for the past few years. The current IXP establishments
at various cities in India based on IXP data and infrastructure based on database-
driven resource data are also presented.
A Systematic Study of Internet eXchange Points … 735

1.3 Article Structure

The article structure is as follows. Section 2 presents the studies carried out worldwide
on IXP and AS in Internet networking architecture. Section 3 presents an overview
of IXP and BGP systems. Section 4 presents the performance analysis of IXP and
AS in the context of global as well as Indian regions. Section 5 presents a case study
of analysis of IXP points in India located at heterogeneous locations, and finally,
Sect. 6 concludes the article.

2 Related Work

A variety of studies have been carried out on IXPs and autonomous systems. The
majority of them focus on the analysis of the current situation of IXPs and how the
latter can enhance the current Internet networking more efficiently. In [1], the overall
study of the topology of the Internet is carried out along with the economical analysis
of field and routing overlays. The topology evolution study was carried out by con-
structing AS-level maps of the Internet topology from the datasets, viz., CAIDA and
RouteViews. The routing performance was considered as the measurement parameter
for the study. Lv et al. [2] studied the geographical distributions of the IXPs along
with the creation time of the IXPs. The authors have analyzed the network traffic
type from the dataset PeeringDB which is a large database of information about the
IXPs and AS along with peer-to-peer network relationships formed by the IXPs. The
authors in [3] study the performance parameters for IXP, the path latency, and propose
a framework that compares the paths through IXPs with alternate paths available.
The study concludes that the alternate paths have superior performance compared to
another method, thus pointing to the need for improvement in the routing policies.
Gavazza et al. [4] presented the reference architecture of the modern Internet, and
presented software-defined networking (SDN) management planes to contribute to
the load balancing of IXPs.
Briain et al. [5] proposed a case study on the eastern African coast and present a
proof-of-concept system design for SDN-based IXP design that presents virtualized
container designs to automate the IXP functions. Authors in [6] proposed solutions for
IXP management through the identification of elephant flows. The elephant flows are
traditionally large traffic flows with persistent high data ingestion. The paper presents
a mechanism for the prediction of elephant flows through temporal patterns. For the
same, the authors have applied a local weighted regression model that predicts the
current flow behavior from the previous flows. The authors predict the volume and
duration on 32 samples and accurately predict 80% of successful elephant flow
prediction, with only 5% of false positives. Fanou et al. [7] have implemented a
system for profiling IXPs in the region of interest and monitoring their growth.
ARDA, a route-collector data analyzer tool provides insights into the status of the
IXP interconnection at local, national, and regional levels. The same was tested in the
736 D. Saraswat et al.

AfriNIC region. Subramani et al. [8] have developed IXmon an open-source system
for detection of adversaries such as distributed reflective denial of service (DRDoS)
for high-data IXPs and peering hubs.
In [9], the analysis of the IXP ecosystem in Brazil was carried out using BGP
datasets obtained from looking glass (GS) server. The insights of the peering rela-
tionships between the autonomous systems per IXP were provided by inspecting the
IPv4 and IPv6 prefix distribution. In [10], the authors study the inter-domain Internet
topology and provide a country-specific analysis. The network science parameters
were measured in a country-specific manner using routing information base (RIB)
dataset of RouteViews project and routing information service (RIS).
The paper brings out analysis of IXPs conducted in terms of AS distribution based
on network traffic (at global level and at Indian context), daily average data traffic at
AS (at global level and at Indian context), utility of AS, and number of established
peering links. The analysis also depicts the data rate distribution in terms of traffic
type existing at IXP level that brings out advantages of future enhancement of IXPs
in various regions. IXP boasts benefits to customers such as reduced cost, latency,
and efficient local Internet connectivity. However, IXP still faces challenges such
as lack of expertise and the setup costs, unavailability of reliable environment for
local data exchange, back-haul connectivity, and high time-to-revenue. However, the
challenges can be addressed though usage of efficient network platforms, increased
automation, and scalable infrastructure.

3 Key Terminologies in Internet Infrastructure

The design of IXP involves an ISP router and connects to an Ethernet network pro-
vided at the IXP. The ISP peers with other participants at the IXP using BGP. Each
participant in BGP communicates with all peers in a complete fashion, or either a
subset of the peers via an AD provided in the IXP core. BGP makes traffic exchange
routing decisions regarding network rules, hop counts, and other parameters con-
trolled by network administrators.

3.1 Internet eXchange Points

The traditional layers of the Internet hierarchy consists of different transit providers
or upstream providers that charges transit fees to propagate the data traffic to the rest
of the Internet, either from the ISPs, or different AS. Because of this reason, ISPs/AS
were brought to the same platform to improve the latency, RTT, routing performance,
and save cost. IXP enhances the efficiency of transmission of data flowing through the
Internet. The objective of establishing an IXP is to provide a platform for participating
networks to be able to connect with external networks through a common switching
mechanism rather than third-party networks for a transit connection. For mutual
A Systematic Study of Internet eXchange Points … 737

benefit, the transmission of data traffic is not billed by the participating AS, ISPs, or
content delivery networks (CDNs).
IXPs consists of a collection of numerous network switches for different par-
ticipating ISPs or AS. The former is normally established in the same city as the
participating AS to reduce the latency. IXPs utilizes fiber-optic inter-repeater links
(FOIRL), and fiber-distributed data interface (FDDI) rings when switches were not
employed. The routing functionality of the data traffic between two participating
ASs is provided using the BGP, an exterior routing protocol. Since the participating
AS has established a peering relationship, the BGP routers announce the routes via
the already established peering relationship. The other participant chooses whether
to accept the routes or not. The routers at the edge of AS are called edge routers
or eBGP peers and employ BGP to maintain information and facilitate information
exchange with other participating systems’ border or edge routers.

3.2 Border Gateway Protocol

Border gateway protocol (BGP) is an exterior routing protocol that facilitates the
exchange of routing information between the edge routers of different ASs. It works
internally as iBGP within AS and eBGP with external systems. Each participating
network or AS is called a BGP speaker or peer; exchanges the routing information
with the neighboring peers or systems, via network prefix announcements. Each
peer maintains a table that contains information about all the routes to systems and
exchanges this information to the neighboring peers. In this way, the BGP ensures that
the participating AS collects all the routing information coming from the neighboring
peers and propagate the information further. For reaching a particular destination,
BGP can take n different available paths. The most suitable path is based on the
organization’s policies that depend on factors like reduced cost, increase reliabil-
ity, and speed. BGP is vulnerable to security attacks; however, it can be secured
using a combination of internet key exchange (IKE) and Internet protocol security
(IPSec) [11].

4 Comparative Analysis of IXPs Performance

In this section, we present the comparative analysis for IXPs, ISPs as well as other
existing AS. The analysis is carried out at two levels, viz., at the global level and Indian
IXP ecosystem context. We also present the improvement measures for specific
properties of IXP and AS in subsequent sections.
738 D. Saraswat et al.

Fig. 2 Distribution of AS-based on network traffic type [12]

4.1 Analysis of IXPs and as at Global-Level

A variety of network traffic types exist based on the direction of traffic that moves
between different AS, viz., mostly inbound, heavy inbound, balanced, mostly out-
bound, and heavy outbound. Here, inbound traffic indicates the data that is coming
into the system or network, whereas outbound refers to the data traffic which is going
outside from the system or network.
Fig. 2 shows the ratio comparison of AS with different network traffic types. The
data is collected for autonomous systems for the year 2010 and 2016 from PeeringDB
[12]. It can be inferred from the figure that majority of the traffic is either balanced
or mostly inbound.
Figure 3 shows the comparison of data traffic rate in the AS from the data of the
year 2010 to 2016. It can be seen that the majority of the AS face traffic in two data
rates, viz., 100–1000 Mbps and 1–5 Gbps. From the above data, it is possible to
derive that the capacity of autonomous systems in larger data traffic ranges has been
increased. The impact of increased AS connection to IXPs enables more peering
and connectivity and customer AS (network) directly exchanges traffic with one or
several other networks.
Tables 2 and 3 showcases the distribution of data rate traffic information at AS
connected to various IXPs in the year 2010 and 2016 according to their varying
speed.
Table 4 shows the differentiation of AS in three different categories based on
collected data. Three categories are defined, viz., transit/access AS, enterprise AS,
and content AS. Transit/access AS are the ones that provide the functionality to route
the data traffic from one AS to another. Based on the classification, it is clear that
the majority of AS in the global scenario falls under transit/access or enterprise AS.
The categorization of AS based on data collected from CAIDA [13]. It also depicts
the comparison of the total number of AS having different types of links that provide
access to the other network.
A Systematic Study of Internet eXchange Points … 739

1200
2010
1000 2016

800
AS Count

600

400

200

0
s

s
s

s
s

s
s
bp

bp
bp

bp
bp

bp
bp
M

G
G

G
M

G
0G
0
10

0+

00
20

00

-5

30
5-

10

10
0-

10

20

0-

0-
0-

20

50
10

Data traffic rate

Fig. 3 Comparison of number of AS connected to IXPs and daily average data traffic rate [12]

Table 2 Total number of AS according to data traffic rate in the year 2010
Traffic at AS - 2010
Traffic information Total AS
Not disclosed 1057
0–20 Mbps 199
20–100 Mbps 436
100–1000 Mbps 324
1–5 Gbps 99
5–10 Gbps 80
10–20 Gbps 99
20–50 Gbps 80
50–100 Gbps 60
100+ Gbps 41
100–200 Gbps 17
200–300 Gbps 9
300–500 Gbps 8
500–1000 Gbps 14
1 Tbps+ 12
740 D. Saraswat et al.

Table 3 Total number of AS according to data traffic rate in the year 2016
Traffic at AS - 2016
Traffic information Total AS
Not disclosed 4160
0–20 Mbps 129
20–100 Mbps 319
100–1000 Mbps 1182
1–5 Gbps 1059
5–10 Gbps 484
10–20 Gbps 292
20–50 Gbps 275
50–100 Gbps 156
100+ Gbps 101
100–200 Gbps 36
200–300 Gbps 44
300–500 Gbps 32
500–1000 Gbps 54
1 Tbps+ 66

Table 4 Number of AS of different types

Type of AS Number of AS
Transmit/access 21,721
Enterprise 27,456
Content 2341

Table 5 Comparison of total number of P2P and P2C links

Year P2P Link P2C Link
1998 852 4921
2020 227,470 132,565

There are two types of links, viz., transit links and peering links. The transit links
are also called provider-to-customer (P2C) where it charges fee from the individual
ISP and AS for providing transit to rest of the Internet. The peering links are also
called peer-to-peer (P2P) links where the participating peers (AS) share mutual traf-
fic without any cost. This is also called settlement-free peering [2]. Based on the
comparison of data and present insight, there is a massive exponential growth of P2P
links which shows that there is a huge demand for settlement-free peering to save
cost, reduce latency, and increase bandwidth.
A Systematic Study of Internet eXchange Points … 741

Fig. 4 Comparison of total

AS connected to IXP [13]

Table 5 shows that there is an increase in peering links that exploit the benefits of
peering. For the classification of links, 4395 links were classified using [12], 46,329
links were classified by CAIDA, and 783 were classified manually. The numbers of
P2P links and P2C links from the year 1998 to 2020 have increased nearly 267 times
and nearly 27 times, respectively.

5 Case Study: Analysis of IXP Eco-System in India

In our research work, there is a crucial emphasis on the analysis of IXPs in India,
particularly due to the lack of any significant work that analyzes the potential and
capacity of currently developing IXP infrastructure in India. As per statistics provided
by VNI, IDC, and Ovum, the demand for Internet video in India has projected a
growth of 640.1% during the year 2010–2015 which is second largest compared to
any country in that duration. Australia tops the list with a growth of 1508.8%.
The analysis projects an increase in the majority of IXPs in India to achieve the
benefit of peering. This is advantageous in providing larger bandwidths and increased
efficiency to Internet users in remote areas. It was found that there are 22 active and
working IXPs in India established in different cities and the majority of these IXPs
have IPv6 deployment. Figure 4 shows the comparison of AS mapping to Indian
IXPs IDs for two consecutive years (2018 & 2019) from data collected using [13] as
mentioned in previous section. IXP IDs 515, 276, and 917 show a drastic increase in
the number of AS. The trend shows a significant increase in the autonomous systems
connected to particular IXP (with IXP IDs 515, 276 located at Mumbai and IXP
ID 917 located at Chennai), leading to increased demand for peering for efficient
Internet infrastructure establishment in Indian cities.
742 D. Saraswat et al.

Fig. 5 Daily average data

rate of IXPs in different
cities of India [14]

Fig. 6 Users ratio of Indian

fixed-line operators as on
2019 [15]

Figure 5 shows the data rate and daily traffic exchange values averaged over
multiple ISPs deployed at particular cities. The daily traffic exchange values are also
averaged depending on load balance conditions of IXP; the exchange value varies
from around 50,000 INR to 6,00,000 INR when port capacity is varied from 100 Mbps
to 10 Gbps. This data is obtained from various ISPs registered with National Internet
Exchange of India (NIXI) [14]. It is worth mentioning that the data represented is
not static and varies as per every connection deal as the latter may vary on the basis
of profit or loss incurred.
In India, the leading ISPs provide narrowband and broadband services based on
user-oriented requirements like power consumption, transmitter power range, and
data rate. The leading fixed-line operators provide services to users based on varied
factors like reliability, cost, scalability, QoS, and communication throughput. Figure
6 lists the leading Indian fixed-line operators for the year 2019 that provides vari-
ous services like Internet, landline connectivity, cellular, etc. Among these fixed-line
operators, BSNL tops the list with 60.48%. According to Telecommunication Regu-
A Systematic Study of Internet eXchange Points … 743

Fig. 7 Subscribers ratio of major ISPs in India as on 2019 [15]

latory Authority of India (TRAI), there were nearly 2.1 crores of wireline subscribers
and 341 ISPs in India as of 2019 [15]. Figure 7 indicates ratio-wise subscribers of
leading ISPs in India. These ISP provides both narrowband and broadband services
as mentioned above.

6 Conclusion and Future Scope

The paper presents a comparative analysis of IXP, AS, and ISP that allows the explo-
ration of IXPs for overall global and Indian scenarios. The ASs were bifurcated
in buckets to observe and compare P2P and P2C network traffic links and average
daily Internet data traffic. A case study is presented to analyze the IXP ecosystem
concerning the Indian context. The IXPs from different cities were considered and
compared the AS and average data rates connected to each IXP. The performance
indicates that IXP can give much better results in terms of latency and bandwidth of
individual AS when established at the correct location. It also encourages peering
links which is beneficial mutually and provides better efficiency and performance
for overall Internet infrastructure when implemented on large scale. Establishment
of IXP reduces transit costs as the exchange of the data traffic for the participant
744 D. Saraswat et al.

is free. Due to efficient transmission of data and direct interconnection, bandwidth

can be improved significantly between adjacent systems. As mentioned above, it
reduces latency significantly as the participating systems connect directly without
the intervention of third-party networks.
In the future, the authors would investigate the impact of network-coded CDN
on IXPs that provide users with specified content type. For the same, SDN-enabled
switch configurations would be studied that selectively bifurcates the flows based on
content resource links.

References

1. Ahmad MZ, Guha R (2011) Internet exchange points and internet routing. In: 2011 19th
IEEE international conference on network protocols, pp 292–294. https://doi.org/10.1109/
ICNP.2011.6089065
2. Lv T, Qin D, Ge L (2018) Research and analysis of statistical characteristics of internet exchange
points. In: 2018 sixth international symposium on computing and networking workshops (CAN-
DARW), pp 558–560. https://doi.org/10.1109/CANDARW.2018.00110
3. Ahmad MZ, Guha R (2012) A tale of nine internet exchange points: studying path latencies
through major regional ixps. In: 37th annual IEEE conference on local computer networks, pp
618–625. https://doi.org/10.1109/LCN.2012.6423683
4. Gavazza JAT, Melo JC, da Silva TB, Alberti AM, Rosa PF, de Oliveira Silva F, Verdi FL,
Suruagy JA (2020) Future internet exchange point (fixp): enabling future internet architectures
interconnection. In: Barolli L, Amato F, Moscato F, Enokido T, Takizawa M (eds) Advanced
information networking and applications. Springer International Publishing, Cham, pp 703–
714
5. Briain DÓ, Denieffe D, Okello D, Kavanagh Y (2020) Enabling models of internet exchange
points for developing contexts. Dev Eng 5:100057. https://doi.org/10.1016/j.deveng.2020.
100057
6. da Silva MVB, Jacobs AS, Pfitscher RJ, Granville LZ (2020) Predicting elephant flows in
internet exchange point programmable networks. In: Barolli L, Takizawa M, Xhafa F, Enokido
T (eds) Advanced information networking and applications. Springer International Publishing,
Cham, pp 485–497
7. Fanou R, Sanchez-Agüero V, Valera F, Mwangi M, Coffin J (2019) A system for profiling
the ixps in a region and monitoring their growth: spotlight at the internet frontier. Int J Netw
Manage 29(2):e2056
8. Subramani K, Perdisci R, Konte M (2020) IXmon: detecting and analyzing DRDoS attacks at
internet exchange points
9. Bucke Brito SH, Silva Santos MA, dos Reis Fontes R, Lachos Perez DA, Lourenço da Silva H,
Esteve Rothenberg CR (2016) An analysis of the largest national ecosystem of public internet
exchange points: the case of brazil. J Commun Inf Syst 31(1) (2016). https://doi.org/10.14209/
jcis.2016.23. https://jcis.sbrt.org.br/jcis/article/view/371
10. Çakmak G, Aydin MN (2017) A country-specific analysis on internet interconnection ecosys-
tems. In: 2017 9th international Congress on ultra modern telecommunications and control sys-
tems and workshops (ICUMT), pp 232–237. https://doi.org/10.1109/ICUMT.2017.8255151
11. Bakkali S, Benaboud H, Ben Mamoun M (2013) Security problems in BGP: an overview. In:
2013 national security days (JNS3), pp 1–5. https://doi.org/10.1109/JNS3.2013.6595458
12. Peeringdb, ”the interconnection database” (2020). https://www.peeringdb.com/. Accessed on
15 Apr 2021
13. Caida, ”caida data server” (2020). http://data.caida.org/datasets. Accessed on 15 Apr 2021
A Systematic Study of Internet eXchange Points … 745

14. National internet exchange of India (nixi), ”routing and tariff policy” (2019). https://nixi.in/
en/routing-and-tarrif-policy. Accessed on 15 Apr 2021
15. Telecom subscription data (2019). https://www.trai.gov.in/release-publication/reports/
telecom-subscriptions-reports. Accessed on 15 Apr 2021
Review on Security of Internet of Things:
Security Requirements, Threats,
and Proposed Solutions

Rayeesa Muzafar, Yashwant Singh, Pooja Anand, and Zakir Ahmad Sheikh

Abstract The Internet of Things (IoT) is a new technology that is quickly gaining
popularity as a major research subject. The Internet of Things has begun to trans-
form and restyle our lifestyle due to its rapid growth. As many things are linked to
the Internet and these things correspond openly with one another with no human
intercession, so the risk of cyberattacks is very high. Billions of connected things
communicate with each other and can exchange sensitive information that may be
leaked. Hence, strengthening IoT’s security and preserving users’ privacy are a major
challenge. This paper seeks to provide a thorough examination of IoT security. IoT
architecture with the taxonomy of security requirements depending on the attacks’
aims is proposed after analyzing several IoT security threats. Furthermore, recent
security solutions are explained and classified according to the application domains
in which they are used.

Keywords IoT · Security · Privacy · IoT threats · Security requirements · IoT

vulnerabilities · IoT attacks

1 Introduction

The Internet of Things (IoT) is an arrangement of gadgets that can communicate

with each other and detect interactions to improve our lifestyle and focus on three
key functions data transmission, data reception, and data processing [1]. The term
IoT was given by Kevin Asthon in 1999 [2]. There are currently nine billion things
linked to the Internet. The Internet of Things (IoT) envisions connecting objects all
around the world. It is estimated that 75% of connected devices is exposed to cyber-
assaults [3]. According to estimates, 25% of attacks on the industrial sector will be

R. Muzafar (B) · Y. Singh · P. Anand · Z. A. Sheikh

Department of Computer Science and Information Technology, Central University of Jammu,
Jammu, Jammu and Kashmir 181143, India
Y. Singh
e-mail: yashwant.csit@cujammu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 747
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_54
748 R. Muzafar et al.

35
30
25
20
15 IoT devices
10 Non IoT devices
5
0

Fig. 1 Estimated devices for IoT and non-IoT [5]

ascribed to corrupt IoT systems by the end of 2020 [4]. Figure 1 shows an estimated
graph of the predicted spike of IoT devices soon, based on a Statista report [5]. The
complexity of controlling IoT is growing at an exponential rate. Advancement in
a smart environment, smart homes, smart traffic, smart industries, smart farming,
smart grids, etc., had made life very easier are some examples where IoT is strongly
recommended.
As a transformational technique for delivering a plethora of services, the Internet
of Things introduces a slew of security problems. The inbuilt vulnerabilities in IoT
devices are the source of these issues. The market is filled with these insecure smart
gadgets, which are simple to manipulate to gain access to an IoT system remotely.
The exponential rise of IoT and its integration with other technologies has created
a larger attack surface [6]. End-users and developers are unaware of the security
threats that come with today’s sophisticated smart applications. Cyberattacks like
Mirai take use of these flaws in IoT devices [7].
Following are the contributions of the paper. To begin, it is necessary to have
prior knowledge of the infrastructure with which we are dealing to comprehend IoT
security; hence, we have covered IoT architecture; we have examined various threats
and security requirements in IoT; we have presented various IoT security solutions.
The rest of the paper is organized as follows. Section 2 provides a brief literature
review of recent developments in IoT, and the IoT architecture is discussed in Sect. 3.
Section 4 presents a brief about IoT threats and requirements for security. Proposed
security solutions are presented in Sect. 5. Finally, Sect. 6 concludes the paper.

2 Related Work

A study of protocols, access control models, and frameworks in IoT was under-
taken by Ouaddah et al. [8]. The survey explored the objectives of scalability, flex-
ibility, context awareness, interoperability, and granularity for preserving security
and privacy against various existing methods of access control.
Review on Security of Internet of Things… 749

The authors described the perception, network, and application levels as well as
some IoT security attacks. Then, they addressed security vulnerabilities and attacks
at various tiers, as well as potential solutions. The basic shortcoming of this study
is that various crucial security issues including access control, integrity, and data
privacy are not included [9].
The security challenges in IoT were defined by Oracevic et al. Confidentiality,
integrity, and authentication were the three key security needs they evaluated. After
presenting some current IoT security solutions, they explored prospective routes for
eventual study in the realm of IoT security. The survey’s fundamental flaw is that
the authors mainly focused on fundamental security needs, overlooking other critical
security issues including access control, privacy, availability, and trust [10].
The authors addressed IoT security threats and vulnerabilities, as well as known
ways to address these issues. The authors presented an analysis of the Internet of
Things (IoT) by providing the distinction between the security challenges of IoT
and that of a typical security network. The study discussed various layers of IoT
(perception, network, and application), and various security threats related to each
layer have been highlighted. The author then discussed various remedies for IoT
threats and vulnerabilities. This study examines the security vulnerabilities posed by
IoT devices in-depth. The IoT security categorization is still unclear, which is one
of the work’s limitations [11].
In Abdur et al. [12], IoT applications such as smart home, healthcare, and industry
have been discussed. The fundamental security requirements in IoT including
access control, confidentiality, authentication, and privacy have also been explored.
The authors have oversimplified the security needs of IoT applications without
considering any existing solutions which is a security flaw of this study.
In Mendez Mena et al. [13], the authors discussed current standards and security
goals for IoT devices. The study specified various security criteria for IoT gadgets
and data protection. The vulnerabilities of several technologies, such as WSN and
RFID flaws, were determined, and several remedies were offered. They also discussed
security issues and potential solutions. They did not, however, go into great detail
about the weaknesses in the enabling technology.
The survey was conducted by Neshenko et al. that exposed a unique classifica-
tion of IoT vulnerabilities, including layer-by-layer vulnerabilities, their remediation
strategies, impact, and capabilities for situational awareness. The authors also offered
a data-driven technique for evaluating IoT maliciousness empirically [14].
We have critically examined certain papers related to security solutions to IoT,
dealing with issues at different layers of the protocol stack. None of the existing
solutions are perfectly matching all the requirements, hence lagging with one or the
other necessity. Some IoT security solutions and their challenges are mentioned in
Table 1.
From the above table, we analyzed that IoT security needs to be taken into consid-
eration. As IoT include more devices day by day, cyber-threat and cyberattack are a
question. The privacy and security of the IoT system should be a major priority, and
there should be no exposure to cyberattacks.
750 R. Muzafar et al.

Table 1 Comparison of the proposed survey

Author Year Description Challenges
Ouaddah et al. [8] 2017 The survey explored the Internet protocols were
objectives of scalability, affected
flexibility, context awareness, To create an access control
interoperability, and model framework based on
granularity for preserving blockchain
security and privacy against
various existing methods of
access control
Chahid et al. [9] 2017 A survey was done on IoT Confidentiality,
security for five layers authentication
Oracevic et al. [10] 2017 A survey for IoT security for Authentication, integrity,
physical and virtual things confidentiality
was presented
Alaba et al. [11] 2017 A survey for IoT security for Trust, privacy, authorization,
three layers was presented authentication
Abdur et al. [12] 2017 A study for the security of Confidentiality, privacy,
IoT was presented access control, authentication
Mendez Mena et al. [13] 2018 A survey was done on IoT Privacy, availability, integrity,
security and privacy for 3 and confidentiality
layers
Neshenko et al. [14] 2019 Classification of IoT To create ways for detecting
vulnerabilities was presented, compromised and susceptible
and also, a data-driven IoT devices, vulnerabilities in
method for providing IoT software are
IoT-specific malicious automatically updated
signatures was presented

3 IoT Architecture

The authors briefly describe the ITU-Telecommunication Standardization Sector

Y. 2002s generic IoT design [15]. In this architecture, there are three layers, i.e.,
perception layer, network layer, and application layer.
Perception Layer: It is also called as device layer of IoT and gives a physical
meaning to each thing. Data sensors are incorporated in many forms such as RFID
tags, infrared sensors, or any other sensor devices that can sense the motion, temper-
ature, humidity, and speed among other things [16]. This layer collects important
information about things from the sensors which are connected to them and trans-
lates it into digital signals that are then transmitted for further action to the network
layer. Threats common to this layer are relay attack, node capture, jamming attack,
eavesdropping.
Network Layer: The function of this layer is to accept relevant information from
the device layer in the form of digital signals and supports various communication
networks such as Wi-Fi, GPRS network, Bluetooth, Zigbee, and GSM [17]. To limit
Review on Security of Internet of Things… 751

the risks of threats and attacks, proper security tools such as any form of encryption
or NIDS can be used. Threats common to this layer are routing attack, Sybil attack,
the man-in-the-middle attack, and DDoS attack.
Application Layer: Depend on the analyzed data, this layer develops IoT appli-
cations for various types of industries. The application layer is particularly useful
in the broadscale development of IoT systems because applications stimulate the
development of IoT [18]. Smart homes, smart health care, smart transportation, smart
grids, are some related applications for IoT. phishing attacks, DoS, session hijacking,
malicious code injection are threats to this layer.

4 IoT Security Threats

The potentials of the cyberattack are a question, as more devices are included in the
IoT every day [19]. Weak passwords, insecure firmware updates, improper mecha-
nisms for authentication, ports for open debugging, and unsafe interfaces make the
IoT system prone to cyberattacks [20]. This section analyzes different IoT security
threats and their purposes. In Table 2, the analysis is summarized.

Table 2 IoT security attacks

Attack Method Purposes
Node tampering Substitute sensor node physically Service availability is affected,
and intruders get access to
sensitive information
Replay attack Information is retransmitted by Trust and confidence of network
the malicious node are obtained
Wormhole attack Intruder receives packets in the Additional attacks are launched;
network at one point, tunnel them data confidentiality is breached
to a different part of the network,
and then replay them into the
network from there
Man-in-the-middle attack Information between two nodes The system is taken down, alter,
is secretly intercepted update, eavesdrop the
communication
Denial of service Send data packets to IoT network Data acquisition is compromised;
the network is disabled; service
provider resources are exhausted
Sinkhole attack Traffic is diverted such that the Additional attacks are launched;
base station does not receive data confidentiality is breached
complete data from nodes
Spoofing The attacker pretends to be a Mac address is modified; intruder
legitimate user pretends to be someone else they
are not
752 R. Muzafar et al.

(a) Spoofing: Spoofing is a user’s capacity to repudiate knowledge and seems to

be something they are not.
(b) Wormhole: During a wormhole attack, the intruder receives packets in the
network at one point, tunnel them to a different part of the network, and then
replay them into the network from there. It is also called a tunneling attack.
(c) Denial of services (DoS): It occurs when an attacker overloads a server with
different types of packets, causing it to crash or begin rejecting legitimate
users from the service. When a botnet strikes, it is known as a distributed
denial-of-service attack (DDoS).
(d) Node Capturing: Attackers may attempt to capture or replace the IoT system’s
node with a malicious node. The attacker controls the new node, which seems
to be part of the system.
(e) Sinkhole attack: This attack diverts traffic so that the base station does not
receive complete data from nodes. As a result, the data provided by devices
are jeopardized.
(f) Replay attack: In this type of attack, valid data transmission is repeated or
delayed intentionally or fraudulently.
(g) Man-in-the-middle attack: Intruder secretly interrupts data transmission
between two or more devices and can interrupt, edit, update, or replace the
data without the knowledge of the victim.

4.1 Security Requirements

Security is a great test for the livelihood of IoT. For IoT, security requirements are
both difficult and compulsory as noted by Alqassem [21]. Security requirements were
described by three properties as coded by CIA Triad, i.e., confidentiality, integrity,
and availability but was insufficient for the context of security. A list of requirements
for security is given in Table 3.

Table 3 Requirements for security

Security requirement Definition
Integrity [22] To guarantee precision, wholeness, as well as the absence of
unauthorized manipulation of data
Confidentiality [22] To guarantee that data will be confidential by granting access to an
authorized client
Privacy [23] To guarantee that only the user has charged to disclose his personal
information
Availability [24] To ensure reliable access to all services in the system
Trustworthiness [25] The ability to verify the identity and build trust.
Review on Security of Internet of Things… 753

5 IoT Security Solutions

This section describes recent solutions proposed for the security of IoT. The authors
presented an intelligent security framework for IoT devices which assures data confi-
dentiality and authentication [26]. It uses symmetric and asymmetric key encryption,
with the key pair created using Regev’s learning with errors (LWEs) mechanism [27].
In Song et al. [28], the authors introduce chaos-based privacy preservation (CPP), a
lightweight communication protocol for safeguarding smart home systems. Agents
and a server are part of the planned smart home concept. The agents communicate
data to the central controller regularly, and the central controller responds with orders
and responses. An asymmetric key, created via a chaos-based technique, is used to
encrypt the data. For each data transmission, the symmetric key is updated. A secure
smart shopping system was given by authors using ultra-high frequency RFID [29].
All objects in this system are RFID tagged, and RFID readers are installed in smart
shelves and smart carts. The authors used ECC-based symmetric and asymmetric
encryption/decryption to secure connection with the server. For healthcare system,
author focused on distributed access control for the security of patients informa-
tion [30]. A lightweight authentication scheme has been presented. Authors in [31]
emphasized mutual authentication in Smart City applications. The paper describes
a four-phase system setup, key generation, encryption, and decryption methodology
for lightweight public-key cryptography. Sigfox provides security support with a
variety of solutions for security, including public-key infrastructure, a robust fire-
wall, hardware security module, and on-the-go security dispatching security solu-
tion; all of which are advantageous in the dynamic IoT environment setup [32]. We
can employ honeypots to figure out what techniques and attack vectors attackers use
to carry out harmful operations [33]. According to a review of per-day traffic, the IoT
honeypot HIoTPOT discovers that the majority of attackers are interested in locating
susceptible devices [34].

6 Conclusion

The Internet of Things (IoT) is a rapidly growing technology field that offers end
customers a variety of cost-effective, efficient, and simple-to-use applications and
services. However, security is one of the most important concerns as IoT systems
become more widely used. We identified and studied the current state of security
requirements in the IoT in this article. Common types of IoT attacks occurring in
general and attacks occurring at different layers of IoT architecture in particular
have been discussed in great detail. Furthermore, we have identified the necessity of
intelligent object design for greater autonomy in terms of recognizing and recovering
from threats.
754 R. Muzafar et al.

Table 4 Security solutions

System Technologies Objectives Domain
Intelligent security Symmetric and To address privacy and Smart environment
framework asymmetric encryption confidentiality in the
using lattice-based IoT
cryptography
CPP Symmetric encryption To achieve Smart home
using chaos-based privacy-preserving in
cryptography, message communication
authentication code protocol
Secure smart shopping Symmetric and To propose a secure Smart environment
system asymmetric encryption smart shopping system
based on ECC,
message authentication
code
LDAC-KS Lightweight To preserve medical Health care
encryption/decryption data privacy and
using pairing-based facilitate secured data
cryptography retrieval
The lightweight mutual Lightweight To design a secure Smart city
authentic cation public-key encryption mutual authentication
protocol protocol for
resource-constrained
devices
Sigfox Public-key To provide security IoT environment
infrastructure, a robust support with a variety
firewall, hardware of solutions for
security module, and security
on-the-go security
dispatching security
solution
Honeypots Hardware/software It traps the adversaries IoT environment
security module by imitating real IoT
assets but having no
value for them, by
calculatedly creating
security vulnerabilities

References

1. Ud Din I et al (2019) The internet of things: a review of enabled technologies and future
challenges. IEEE Access 7:7606–7640. https://doi.org/10.1109/ACCESS.2018.2886601
2. Dudhe PV, Kadam NV, Hushangabade RM, Deshmukh MS (2017) Applications. In: Inter-
national conference on energy, communication data analytics and soft computing, pp
2650–2653
3. Naik S, Maral V (2017) Cyber security—IoT. In: RTEICT 2017—2nd IEEE interna-
tional conference on recent trends electronics, information and communication technology
proceedings, vol 2018, pp 764–767. https://doi.org/10.1109/RTEICT.2017.8256700
Review on Security of Internet of Things… 755

4. att-cybersecurity-insights-vol-5-the-ceos-guide-to-data-security @ www.business.att.com.
[Online]. Available: https://www.business.att.com/learn/research-reports/att-cybersecurity-ins
ights-vol-5-the-ceos-guide-to-data-security.html
5. Bhatt RS, Ragiri PR (2021) Security trends in Internet of Things: a survey. SN Appl Sci
3(1):1–14. https://doi.org/10.1007/s42452-021-04156-9
6. Anand P, Singh Y, Selwal A, Alazab M, Tanwar S, Kumar N (2020) IoT vulnerability assess-
ment for sustainable computing: threats, current solutions, and open challenges. IEEE Access
8:168825–168853. https://doi.org/10.1109/ACCESS.2020.3022842
7. Kolias C, Kambourakis G, Stavrou A, Voas J (2017) DDoS in the IoT: Mirai and other botnets.
Computer (Long Beach Calif) 50(7):80–84. https://doi.org/10.1109/MC.2017.201
8. Ouaddah A, Mousannif H, Abou Elkalam A, Ait Ouahman A (2017) Access control in IoT:
survey and state of the art. In: International conference on multimedia computing and systems—
proceedings, vol 0, pp 272–277. https://doi.org/10.1109/ICMCS.2016.7905662
9. Chahid Y, Benabdellah M, Azizi A (2017) 07934655.Pdf
10. Oracevic A, Dilek S, Ozdemir S (2017) Security in internet of things: a survey. In: International
symposium on networks, computers and communications ISNCC 2017. https://doi.org/10.
1109/ISNCC.2017.8072001
11. Alaba FA, Othman M, Hashem IAT, Alotaibi F (2017) Internet of things security: a survey. J
Netw Comput Appl 88:10–28. https://doi.org/10.1016/j.jnca.2017.04.002
12. Abdur M, Habib S, Ali M, Ullah S (2017) Security issues in the internet of things (IoT): a
comprehensive study. Int J Adv Comput Sci Appl 8(6). https://doi.org/10.14569/ijacsa.2017.
080650
13. Mendez Mena D, Papapanagiotou I, Yang B (2018) Internet of things: survey on security. Inf
Secur J 27(3):162–182. https://doi.org/10.1080/19393555.2018.1458258
14. Neshenko N, Bou-Harb E, Crichigno J, Kaddoum G, Ghani N (2019) Demystifying IoT secu-
rity: an exhaustive survey on IoT vulnerabilities and a first empirical look on internet-scale
IoT exploitations. IEEE Commun Surv Tutorials 21(3):2702–2733. https://doi.org/10.1109/
COMST.2019.2910750
15. Zhang N et al (2017) Understanding IoT security through the data crystal ball: where we are
now and where we are going to be. [Online]. Available: http://arxiv.org/abs/1703.09809
16. Bandyopadhyay D, Sen J (2011) Internet of things: applications and challenges in technology
and standardization. Wirel Pers Commun 58(1):49–69. https://doi.org/10.1007/s11277-011-
0288-5
17. dac @ onlinelibrary.wiley.com. [Online]. Available: https://onlinelibrary.wiley.com/doi/epdf/
https://doi.org/10.1002/dac.3981
18. Fang WD, He W, Chen W, Shan LH, Ma FY (2016) Research on the application-driven archi-
tecture in internet of things. Front Artif Intell Appl 293:458–465. https://doi.org/10.3233/978-
1-61499-722-1-458
19. Thakur K, Qiu M, Gai K, Ali ML (2016) An investigation on cyber security threats and
security models. In: Proceedings—2nd IEEE international conference on cyber security cloud
computing CSCloud 2015—IEEE international symposium smart cloud, IEEE SSC 2015, pp
307–311. https://doi.org/10.1109/CSCloud.2015.71
20. Internet_Of_Things_Iot_Security_Best_Practices_IEEE_Community_led_White_Paper @
www.academia.edu. [Online]. Available: https://www.academia.edu/32053241/Internet_Of_
Things_Iot_Security_Best_Practices_IEEE_Community_led_White_Paper
21. 2591062 @ dl.acm.org. [Online]. Available: https://dl.acm.org/doi/abs/https://doi.org/10.1145/
2591062.2591201
22. Patel VH, Patel S (2021) A review on IoT security: challenges and solution using lightweight
cryptography and security service mechanisms offloading at fog. SSRN Electron J 640–648.
https://doi.org/10.2139/ssrn.3769832
23. Misra S, Maheswaran M, Hashmi S (2017) Security challenges and approaches in internet of
things. Springer, Cham
24. Roman R, Zhou J, Lopez J (2013) On the features and challenges of security and privacy in
distributed internet of things. Comput Networks 57(10):2266–2279. https://doi.org/10.1016/j.
comnet.2012.12.018
756 R. Muzafar et al.

25. Abomhara M, Køien GM (2015) Cyber security and the internet of things: vulnerabilities,
threats, intruders and attacks. J Cyber Secur Mobil 4(1):65–88. https://doi.org/10.13052/jcs
m2245-1439.414
26. Sridhar S, Smys S (2017) Intelligent security framework for IoT devices. In: International
conference on inventive systems and control intelligence, pp 1–5
27. Regev O (2009) On lattices, learning with errors, random linear codes, and cryptography. J
ACM 56(6):1–37. https://doi.org/10.1145/1568318.1568324
28. Song T, Li R, Mei B, Yu J, Xing X, Cheng X (2017) A privacy preserving communication
protocol for IoT applications in smart homes. IEEE Internet Things J 4(6):1844–1852. https://
doi.org/10.1109/JIOT.2017.2707489
29. Li R, Song T, Capurso N, Yu J, Couture J, Cheng X (2017) IoT applications on secure smart
shopping system. IEEE Internet Things J 4(6):1945–1954. https://doi.org/10.1109/JIOT.2017.
2706698
30. Yang Y, Zheng X, Tang C (2017) Lightweight distributed secure data management system for
health internet of things. J Netw Comput Appl 89:26–37. https://doi.org/10.1016/j.jnca.2016.
11.017
31. Li N, Liu D, Nepal S (2017) Lightweight mutual authentication for IoT and its applications.
IEEE Trans Sustain Comput 2(4):359–370. https://doi.org/10.1109/TSUSC.2017.2716953
32. Coman FL, Malarski KM, Petersen MN, Ruepp S (2019) Security issues in internet of things:
vulnerability analysis of LoRaWAN, sigfox and NB-IoT. In: Global IoT summit, GIoTS 2019—
proceedings. https://doi.org/10.1109/GIOTS.2019.8766430
33. Farivar F, Haghighi MS, Jolfaei A, Alazab M (2020) Artificial intelligence for detection, estima-
tion, and compensation of malicious attacks in nonlinear cyber-physical systems and industrial
IoT. IEEE Trans Ind Informatics 16(4):2716–2725. https://doi.org/10.1109/TII.2019.2956474
34. Gandhi UD, Kumar PM, Varatharajan R, Manogaran G, Sundarasekar R, Kadu S (2018)
HIoTPOT: surveillance on IoT devices against recent threats. Wirel Pers Commun
103(2):1179–1194. https://doi.org/10.1007/s11277-018-5307-3
Emerging Security Issues in IoT

Umaira Ahad, Yashwant Singh, and Pooja Anand

Abstract The Internet of Things (IoT) is an innovative model that combines the
physical world with the Internet and turns out to be one of the major hi-tech break-
throughs of computer technology. The low openness and absence of interoperability
of a significant number of these gadgets in an enormous region of magnitude will
make it extremely hard to plan explicit wellbeing techniques and actualize a partic-
ular security strategy for sustainable IoT. Moreover, IoT networks presented are
helpless against assaults planned to disturb the organization. Thus, special measures
are required to combat the upward security apprehensions in IoT. In this article,
these security concerns with respect to different layers of basic architecture of IoT
are studied in detail; with other emerging challenges, IoT is facing today.

Keywords Internet of Things · Intrusion detection system · Security in IoT

1 Introduction

The Internet of Things (IoT) is a rapidly emerging networking model that aims to link
a wide range of devices to the Internet, collecting sensor-generated data, controlling
appliances and devices remotely, monitoring environments, vehicles, and buildings,
etc. [1]. IoT’s vision is to use smart technology to link and control things at anytime,
anywhere. IoT was first introduced in 1998, and Kevin Ashton invented the word in
1999 [2]. Moreover, the rise in the number of Internet-linked artifacts has prepared
the IoT a progressively rising subject in recent years and is estimated to increase
exponentially in the near-term years. According to some estimations, the numbers of
associated gadgets before the end of 2025 is anticipated to hit 75 billion. The growth
in IoT devices by 2040 is presented in Fig. 1. Additionally, in the coming years,

U. Ahad (B) · Y. Singh · P. Anand

Department of Computer Science and Information Technology, Central University of Jammu,
Jammu, Jammu and Kashmir 181143, India
Y. Singh
e-mail: yashwant.csit@cujammu.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 757
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_55
758 U. Ahad et al.

Growth of IoT Devices

10
120

9
IOT DEVICES (IN BILLIONS)

100

POPULATION (IN BILLION)

7 8
80

6
60

4 5
40

2 3
20

1
0

0
2003 2010 2010 2015 2020 2025 2030 2040
Iot Devices
population

Fig. 1 Growth of IoT devices with population [6]

assaults that threaten resource-constrained IoT devices have multiplied manifold.

Along with this, the inherent vulnerabilities have become the part and parcel of the
IoT-based systems used in both industrial sectors and home environments [3].
Data security as per the trinity, i.e., data confidentiality, accessibility, and integrity
are even a prey. To guarantee data protection, organizations must classify data
according to its vulnerability and then allow access to people based on their autho-
rization to access each class of data. In addition, data protection must not be disre-
garded in the current IoT period in such a way that personal details can be revealed
publicly or to an individual not allowed to access it [4]. Based on the study of
risk categories and IoT protection mechanisms, detection systems should be further
configured. The intrusion detection technique evaluates assaults using data collected
from various computer network collecting points. Integration of an active security
approach capable of intercepting and responding to intrusions before they reach the
network is required [5].
As the Internet is the heart and core of IoT, practically all security concerns that
exist on the internet also affect IoT. In comparison to other traditional networks,
the IoT’s vital nodes are allocated to locations without administrative monitoring,
with constrained capacity and resources, leaving the IoT’s security concerns partic-
ularly challenging. However, the apparent benefits offered by the IoT paradigm are
accompanied by major loopholes. The majority of data acquired by IoT devices is
confidential and requires protection. These sensitive IoT data might be an open invi-
tation for hackers to seize them and use them in a variety of ways [7]. These sensitive
security issues are being stressed upon in this article. The paper is structured in the
Emerging Security Issues in IoT 759

following sections: Sect. 2 presents the background of IoT in terms of the basic
architecture of IoT followed by the applications of IoT; Sect. 3 delves deeper into
security issues in IoT. Section 4 presents various challenges in IoT, and finally, we
conclude in Sect. 5.

2 Background and Preliminaries

2.1 Architecture of IoT

As there is no unique architecture for the IoT, various architectural styles are
suggested by different researchers. Researchers typically offer an IoT-layered archi-
tecture [8]. Three-layer architecture: This is the basic architecture of IoT, consisting
of three layers: perception layer, network layer, and application layer as shown in
Fig. 2.
(a) Perception Layer: Sensors are used in the perception layer to feel and collect
knowledge about the world. The protocols for communication between devices
in an IoT network are included in this layer. The number of homogeneous and
heterogeneous devices linked to the Internet via communication protocols is
increasingly growing. In terms of battery capacity, computing capabilities, and
storage capital, IoT devices are limited. Wi-Fi, NFC, Zigbee, Sigfox, and long-
range wide area networks (LoRaWAN) are IoT awareness layer technologies
that link the physical world to the network [9].

(b) Network Layer: The network layer’s most powerful role is to link the device
and sensing layers. Fixed and mobile telephone networks, cable networks, the
Internet, electrical communication networks, and private networks are exam-
ples of transmission networks. Fiber networking, cellular Internet, Ethernet
access, and satellite connectivity are examples of network layer communication
approaches [11].

Fig. 2 Architecture of IoT

[10] Application Applications Specified Service to User like
Layer intelligent transportation, smart home etc.

Network Interconnection Communication Protocols like

Layer
ZigBee, Bluetooth, Wi-Fi etc.

Perception
Layer Physical Devices like sensors, RFID, Cameras
etc.
760 U. Ahad et al.

(c) Application layer: Data formatting and presentation are handled by the appli-
cation layer. On the Internet, HTTP is widely used as the application layer.
Due to its verbose nature, which results in a large decoding overhead, HTTP
is unsuitable for resource-constrained environments. Several alternative proto-
cols, such as message queue telemetry transport (MQTT), and constrained
application protocol (CoAP), have been proposed for IoT settings [12].

2.2 Applications of IoT

The IoT applications are widespread and expanded in all aspects of people’s everyday
lives that generally cover culture, sectors, and the environment. Some IoT applica-
tions are shown in Fig. 3. In the broad sphere of science and technology, IoT finds
its application. It is used in computing, big data, smart city apps, smarter energy
management systems, mobile charging apps, smart retail, automated transport, envi-
ronmental monitoring, delivery of water, urban protection, smart surveillance, smart
supply chain, etc. IoT devices allow hospitals to monitor their patients’ health at
home, reducing hospital stays while also providing real-time data that might save lives
[13]. A smart home system has the potential to make our lives much easier. Starting
from energy conservation, which includes the power controls feature of AC appli-
ances and the thermostat, much of which is controlled to reduce the amount of energy
used. IoT technologies have the potential to turn reactive care processes into construc-
tive wellbeing systems. Current medical science relies on services that are deficient
in vital real-world data. For medical review, it often uses leftover records, managed
environments, and volunteers. Via research, real-time field data, and monitoring, the
IoT opens doors to a sea of useful data [14].

Fig. 3 Applications of IoT

Smart Wearables
City

Smart
Healthcare IoT homes
Applications

Agriculture Traffic
Monitoring
Emerging Security Issues in IoT 761

3 Security Issues in IoT

Different security risks in IoT implementations for the four layers are discussed in
this section. The attacks that could be rendered on these four layers are depicted in
Fig. 4.
(a) Security issues at sensing layer
• Node Capturing: A few low-power nodes, such as sensors and actuators,
are used in IoT applications. These nodes are insecure against a variety
of attackers. In the IoT framework, the assailants may effort to replicate
or replace the hub with a malignant hub. The new hub may give off an
impression of being essential for the framework is however constrained by
the assailant. This may force the IoT system’s cumulative protection to be
negotiated [15].
• Malicious Code Injection Attack: The assault includes the aggressor
infusing around noxious code into the memory of the hub. By and large,
the firmware or programming of IoT hubs are overhauled broadcasting
live, and this stretches a passage to the aggressors to infuse noxious code.
Utilizing such malignant code, the assailants may compel the hubs to play
out some unintentional capacities or may even attempt to get to the total
IoT framework [16].
• False Data Injection Attack: When the aggressor seizes the hub, he or she
can use it and inject false data into the system of IoT. This might cause

Sensing Networking Middle Application

Layer Layer Layer Layer
Node Phishing Data Theft
Capture Site Flooding Attack
Attacks Attack in
Access Cloud Access
Malicious Attack Control
Code Attack
DOS Cloud
Injection Malware
Attack Attack Sniffing
Injection Attacks
Data
False Data Transit
Injection Signature Malicious
Attack
Attack Wrapping Code
Routing Attack Injection
Side-channel Attack Attack
attack(SCA)
SQL
Eavesdropping injection Rreprogramming
and Attack Attack
Interference Men in
the
Booting middle
Vulnerabilities Attack

Fig. 4 Attacks in IoT [15]

762 U. Ahad et al.

erroneous consequences and the IoT application breaking down. A DDoS

attack might also be launched using this method [17].
• Side-Channel Attacks (SCA): A variety of side-channel attacks can result in
confidential data being leaked. Processor microarchitectures, electromag-
netic emanation, and resource usage both expose confidential data to adver-
saries. Power use, laser-based attacks, pacing attacks, and electromagnetic
attacks are all examples of side-channel attacks [18]. When integrating cryp-
tographic modules, modern chips take different defense systems to avoid
side-channel assaults.
• Eavesdropping and Interference: IoT systems are frequently made up of
a variety of nodes that are distributed in open spaces. As a consequence,
eavesdroppers can control those IoT applications. All through various stages
along with transmitting data or authentication, attackers may listen in and
collect information.
• Booting Attacks: During the boot operation, edge devices are susceptible
to a variety of attacks. This is because the built-in security processes aren’t
allowed at that time. As a result of this weakness, when node devices are
rebooted, attackers might try to target devices. Since gadgets drive through
sleep–wake cycles and are usually low-powered, securing the boot phase in
these devices is critical [19].
(b) Security issues at network layer
• Phishing Site Attack: These assaults regularly allude to attacks where a
few IoT gadgets can be focused on by an insignificant exertion put by the
assailant. The aggressors expect that at any rate not many of the gadgets
will turn into a casualty of the assault. There is a chance of experiencing
phishing destinations throughout clients visiting pages on the Web. When
the client’s record and secret phrase are undermined, the entire IoT climate
being utilized by the client gets defenseless against cyberattacks. In IoT, the
network layer is completely defenseless to phishing site attacks [20].
• Access Attack: In this type of assault, an unknown entity or adversary gains
access to the IoT network. The assailant can keep on remaining in the
organization undetected for the long term. Rather than causing network
disruption, this form of attack aims to steal sensitive data. IoT frameworks
transmit and receive critical data regularly; as a result, they are extremely
vulnerable to such assaults [21].
• DDoS/DoS Attack: In distributed denial-of-service (DDoS) attack, when a
Web site is “brought down by hackers,” it is almost always the result of a
DDoS attack. In plain words, this means that hackers wanted to make a Web
site or device inaccessible by overwhelming or crippling it with too much
traffic [22].
• Data Transit Attacks: In IoT applications, a large volume of data is shared
and processed. Since data are important, hackers and other foes continue
to pursue it. Cyberattacks can compromise data held in the cloud or on
local computers, but data that are traveling from one place to another are
Emerging Security Issues in IoT 763

even more so. In IoT applications, there is a lot of data transfer between
sensors, actuators, the cloud, and other devices. As a consequence of the
various connection technologies being used in transforming, IoT apps are
vulnerable to data attacks [23].
• Routing Attacks: During data transfer, malicious nodes in IoT applications
might redirect routing routes. The sinkhole is another form of routing attack
in which a fictitious routing path and recruits nodes to use it to route traffic
by an adversary. A wormhole attack is a type of attack that, when combined
with other types of attacks including sinkhole attacks, can pose a serious
security risk. Another attack is a wormhole attack when this attack combines
with others such as sinkhole attacks leads to a serious security issue [24].
(c) Security issues at middleware layer
• Man-in-the-Middle Attack: The MQTT protocol uses a publish–subscribe
model for communication between clients and subscribers, with the
MQTT broker serving as a middleman. By separating the subscribing and
publishing clients from one another, messages can be transmitted without
the recipient’s awareness. If an attacker acquires control of the broker and
act as a man-in-the-middle, the attacker will have complete control over all
communication, even if the clients do not approve [25].
• SQL Injection Attack: A malicious SQL statement can be inserted into a
program by an attacker. The attackers can then access the personal infor-
mation of the user and even change database records. SQL injection attacks
are caused by illegal access to private data such as credit card information,
passwords, or personal user credentials [26].
• Signature Wrapping Attack: In the middleware of Web services, XML signa-
tures are used. Through leveraging simple object access protocol (SOAP)
vulnerabilities, the attacker may crack the algorithm’s signature and accom-
plish operations or modify the message in a signature wrapping assault
[27].
• Cloud Malware Injection: The adversary could gain access, insert a virtual
machine, or malware using cloud malicious injection into the cloud. The
attacker impersonates a genuine service by attempting to construct a mali-
cious service module. Throughout this manner, the assailant can also get
access to the victim’s service requests, collect private records, and alter the
data as needed.
• Flooding Attack in Cloud: Flooding attack is alike to a DoS attack in the
cloud, and it affects service quality. The aggressors constantly send requests
to service to deplete cloud resources. These assaults can hugely affect cloud
frameworks by expanding the load on the cloud servers [28].
(d) Security issues at the application layer
• Data Thefts: IoT systems work with a lot of sensitive and personal infor-
mation. Information on the move is much more vulnerable to attacks than
764 U. Ahad et al.

information that is still, and there is a lot of information creation in IoT appli-
cations. When IoT apps are susceptible to data intrusion attacks, customers
may be unwilling to register their sensitive information. Any of the methods
and protocols being used to safeguard IoT applications against cyberat-
tacks include encryption technology, device isolation, client and network
authentication, privacy security, and so on [29].
• Access Control Attacks: A mechanism that restricts access to information or
accounts to only approved users or processes is access control. In IoT appli-
cations, an access control attack is critical since granting access exposes the
entire IoT architecture to attackers.
• Malicious Code Injection Attacks: Intruders frequently take the quickest or
fastest route to gain access to a device or network. If the device is vulnerable
to malicious scripts and misdirection as a consequence of insufficient code
testing, it will be an attacker’s first point of entry. Cross-site scripting (XSS)
is a method used by attackers to inject malicious code into a trusted Web
site. If an XSS attack is successful, an IoT account can be hijacked, and the
IoT device can be rendered useless [30].
• Sniffing Attacks: Sniffer software can be used by foes to track IoT appli-
cations for network traffic. If there are not adequate defense mechanisms
ready to prevent this attack, the foe will be capable to access sensitive data
of the user [31].
• Reprogram Attacks on the off chance that the programming mechanism
isn’t ensured; the assaulters can modify IoT protests distantly. This may
prompt the hacking of IoT organizations.

4 Emerging Challenges in IoT

Developing an IoT network alone would be a challenge, and this is a challenge.

Developing IoT security algorithms allows us to gain a better knowledge of the
entire challenges. IoT challenges can be summed up as and are depicted in Figure 5
but are not bound to these only [32].
• Data mining challenges: Data mining applications are becoming more relevant
as more data becomes usable for processing and interpretation. Data mining soft-
ware may be used to trigger disciplinary procedures to address urgent operating

Fig. 5 Challenges in IoT

IoT
Challeges

Data
Privacy Security Standard
Mining
Challenge Challenge Challenge
Challenge
Emerging Security Issues in IoT 765

challenges or to alert managers about revelations about competitors’ competitive

moves and customer demand shifts that could affect their short- and long-term
market operations. Using an algorithm and mathematical models, data must be
tamed and interpreted. Unstructured photographs and video files do not lend them-
selves to traditional data processing techniques. There is a lack of qualified data
scientists, which is compounded by the need for specialized data processing tech-
niques to pit data streams from sensor networks and multimedia content data
[33].
• Privacy challenge: In comparison to traditional networks, IoT privacy concerns are
becoming increasingly prevalent. Because data involve confidential info, protec-
tion of privacy is a major security concern in the IoT. Security of IoT must shield
wider management items and stages over traditional data security in line with
the integration of things, networks, and services. The prevailing security style is
based on human interactions and possibly will not be suited for usage in the IoT
[34].
• Security challenge: The possible security vulnerability increases as the sheer
variety of embedded devices incorporated into IoT systems grows. While the IoT
increases business efficiency and improves human livelihoods, it also expands the
attack surface for hackers and other data breaches. Each system had an average
of 25 holes or risks of breaching the home network. Data encryption methods are
seldom used by IoT devices. This tendency arises as a result of the difficulty posed
by the widespread use of identical IoT devices. Apart from the capacity of certain
gadgets to mechanically connect to other devices, this means that IoT users and
developers alike must verify that they are not harming other users or the Internet
itself [35].
• Standard Challenge: In the growth of IoT, standards are quite significant. A stan-
dard is required to ensure that all entities may use and access information in the
same way. Professional design of IoT applications, infrastructures, devices, and
services will be supported by standardization and cooperation. Overall, collabo-
rating parties’ standards, as well as the information models and protocols included
in the standards, must be open. All members must be allowed to participate in the
standard-setting process, and the resulting standards must be freely and publicly
available [36].
There are numerous security concerns in the IoT, including confidentiality,
integrity, availability, and authentication. We examined different methods that might
be implemented to the IoT environment to address these concerns. The proposed
framework is presented in Fig. 6 to mitigate these issues. Lightweight-based cryp-
tographic solutions can overcome these issues; some learning-based solutions (ML,
DL, transfer learning, federate learning) and blockchain-based solutions can also be
implemented to maintain data privacy and integrity.
766 U. Ahad et al.

Fig. 6 Proposed solution framework for IoT

5 Conclusion

The IoT has a vital role in the advancement of high tech progress in the direction
of automation and application design using machine learning and artificial intelli-
gence. In both of these applications, data play a critical role. Data from devices or
environments that are connected through networking interfaces must be transferred
in a secure manner. The presence of the IoT worldview over the most recent couple
of years has been releasing countless menaces and possible assaults against the secu-
rity and protection of IoT articles and people. These dangers lead to hampering
the acknowledgment of this worldview on the off chance that they have been left
without appropriate countermeasures threatening its long-term viability. This article
covered the background and preliminaries to IoT. Then, in next section, the security
issues at different layers of IoT were critically analyzed and detailed. Additionally,
the emerging challenges like privacy, security, data mining, and business challenges
were highlighted. In the future, we aspire to build a secure blockchain-based IoT
protocol that will safeguard IoT systems from the majority of integrity assaults.

References

1. Meneghello F, Calore M, Zucchetto D, Polese M, Zanella A (2019) IoT : internet of threats

? A survey of practical security vulnerabilities in real IoT devices. IEEE Internet Things J 1.
https://doi.org/10.1109/JIOT.2019.2935189
Emerging Security Issues in IoT 767

2. Bhuvaneswari V, Porkodi R (2014) The internet of things (IOT) applications and commu-
nication enabling technology standards: an overview. In: Proceedings—2014 International
conference on intelligent computing applications ICICA, pp 324–329. https://doi.org/10.1109/
ICICA.2014.73
3. Stellios I, Kotzanikolaou P, Psarakis M, Alcaraz C, Lopez J (2018) A survey of IoT-enabled
cyberattacks : assessing attack paths to critical infrastructures and services. IEEE Commun
Surv Tutorials 1. https://doi.org/10.1109/COMST.2018.2855563
4. Makhdoom I, Abolhasan M, Lipman J, Liu RP, Ni W (2018) Anatomy of threats to the internet
of things. IEEE Commun Surv Tutorials 1. https://doi.org/10.1109/COMST.2018.2874978
5. Liu L, Xu B, Zhang X, Wu X (2018) An intrusion detection method for internet of things
based on suppressed fuzzy clustering. Eurasip J Wirel Commun Netw 1:2018. https://doi.org/
10.1186/s13638-018-1128-z
6. Anand P, Singh Y, Selwal A, Singh PK, Felseghi RA, Raboaca MS (2020) IoVT: internet of
vulnerable things? threat architecture, attack surfaces, and vulnerabilities in internet of things
and its applications towards smart grids. Energies 13(18):1–23. https://doi.org/10.3390/en1318
4813
7. Hameed A (2019) Security issues in IoT : a survey. In: International conference on innovation
and intelligence informatics, computing, and technologies, pp 1–5
8. Mrabet H, Belguith S, Alhomoud A, Jemai A (2020) A survey of IoT security based on a
layered architecture of sensing and data analysis. Sensors (Switzerland) 20(13):1–20. https://
doi.org/10.3390/s20133625
9. Sarkar C et al (2014) DIAT : a scalable distributed architecture for IoT. X(X):1–10. https://doi.
org/10.1109/JIOT.2014.2387155
10. Li D, Deng L, Lee M, Wang H (2019) IoT data feature extraction and intrusion detection system
for smart cities based on deep migration learning. Int J Inf Manage 49(April):533–545. https://
doi.org/10.1016/j.ijinfomgt.2019.04.006
11. Dawoud A, Shahristani S, Raun C (2018) Deep learning and software-defined networks:
towards secure IoT architecture. Internet of Things 3–4:82–89. https://doi.org/10.1016/j.iot.
2018.09.003
12. Lv W, Meng F, Zhang C, Lv Y, Cao N, Jiang J (2017) A general architecture of IoT system.
659–664. https://doi.org/10.1109/CSE-EUC.2017.124
13. Mathew PS, Pillai AS, Palade V (2018) Applications of IoT in healthcare
14. Chaudhary S (2019) CRAIoT : concept, review and application (s) of IoT. In: 2019 4th
International conference on internet things smart innovation and usages, pp 1–4
15. Deogirikar J, Vidhate A (2017) Security attacks in IoT: a survey. In: Proceedings of the inter-
national conference on IoT social mobile, analytics and cloud, I-SMAC, pp 32–37. https://doi.
org/10.1109/I-SMAC.2017.8058363
16. Anand P, Singh Y, Selwal A, Alazab M, Member S, Tanwar S (2020) IoT vulnerability
assessment for sustainable computing : threats, current solutions, and open challenges.
168825–168853. https://doi.org/10.1109/ACCESS.2020.3022842
17. Hassija V, Chamola V, Saxena V, Jain D, Goyal P, Sikdar B (2019) A survey on IoT security:
application areas, security threats, and solution architectures. IEEE Access 7:82721–82743.
https://doi.org/10.1109/ACCESS.2019.2924045
18. Tawalbeh L, Muheidat F, Tawalbeh M, Quwaider M (2020) IoT privacy and security : challenges
and solutions. Appl Sci 1–17
19. Andrea I, Chrysostomou C, Hadjichristofi G (2015) Internet of things : security vulnerabilities
and challenges. pp 180–187
20. Frustaci M, Pace P, Aloi G, Fortino G (2017) Evaluating critical security issues of the IoT
world : present and future challenges. 4662. https://doi.org/10.1109/JIOT.2017.2767291
21. Mukherjee BA (2015) Physical-layer security in the internet of things : sensing and
communication confidentiality under resource constraints. 103(10)
22. Galeano-Brajones J, Carmona-Murillo J, Valenzuela-Valdés JF, Luna-Valero F (2020) Detec-
tion and mitigation of DoS and DDoS attacks in iot-based stateful SDN: an experimental
approach. Sensors (Switzerland) 20(3). https://doi.org/10.3390/s20030816
768 U. Ahad et al.

23. Sharma V, You I, Andersson K, Palmieri F, Rehmani MH, Lim J (2020) Security, privacy and
trust for smart mobile-internet of things (M-IoT): a survey. https://doi.org/10.1109/ACCESS.
2020.3022661
24. Yugha R, Chithra S (2020) A survey on technologies and security protocols: reference for
future generation IoT. J Netw Comput Appl 169:102763. https://doi.org/10.1016/j.jnca.2020.
102763
25. Tournier J, Lesueur F, Le Mouël F, Guyon L, Ben-Hassine H (2020) A survey of IoT protocols
and their security issues through the lens of a generic IoT stack. Internet of Things 100264.
https://doi.org/10.1016/j.iot.2020.100264
26. Ngu AH, Gutierrez M, Metsis V, Nepal S, Sheng QZ (2017) IoT middleware: a survey on
issues and enabling technologies. IEEE Internet Things J 4(1):1–20. https://doi.org/10.1109/
JIOT.2016.2615180
27. Ibitoye O, Shafiq O, Matrawy A (2019) Analyzing adversarial attacks against deep learning
for intrusion detection in IoT networks. arXiv
28. Pal S, Hitchens M, Rabehaja T, Mukhopadhyay S (2020) Security requirements for the internet
of things: a systematic approach. Sensors (Switzerland) 20(20):1–34. https://doi.org/10.3390/
s20205897
29. Nebbione G, Calzarossa MC (2020) Security of IoT application layer protocols : challenges
and findings. 1–20. https://doi.org/10.3390/fi12030055
30. Obaidat MA, Obeidat S, Holst J, Al Hayajneh A, Brown J (2020) A comprehensive and system-
atic survey on the internet of things: security and privacy challenges, security frameworks,
enabling technologies, threats, vulnerabilities and countermeasures. Computers 9(2). https://
doi.org/10.3390/computers9020044
31. Benkhelifa E, Welsh T, Hamouda W (2018) A critical review of practices and challenges in
intrusion detection systems for IoT: toward universal and resilient systems. IEEE Commun
Surv Tutorials 20(4):3496–3509. https://doi.org/10.1109/COMST.2018.2844742
32. Chen S et al (2020) A vision of IoT : applications, challenges, and opportunities with China
perspective. 1(4):349–359
33. Challenges S (2021) Internet of things: evolution, concerns and security challenges
34. Farhan L, Alissa AE (2017) A survey on the challenges and opportunities of the internet of
things (IoT)
35. Sengupta J, Ruj S, Das Bit S (2019) A comprehensive survey on attacks, security issues and
blockchain solutions for IoT and IIoT. J Netw Comput Appl 102481. https://doi.org/10.1016/
j.jnca.2019.102481
36. Panchiwala S, Shah M (2020) A comprehensive study on critical security issues and challenges
of the IoT world
Robust and Imperceptible Multiple
Watermarking Using Transform Domain
Algorithm for Digital Media

Namita Agarwal, Amit Kumar, and Pradeep Kumar Singh

Abstract This paper discussed a novel robust multiple watermarking method with
three transform domain techniques, discrete wavelet transforms (DWT), discrete
cosine transforms (DCT) and singular value decomposition (SVD) for color images.
For security reasons, secret media is inserted into same multimedia items to offers
an additional level of security and to achieve two important performance metrices of
watermarking. Firstly, the original media (image) is disintegrated into first level DWT
and generate DWT coefficients then select the lower frequency sub-band for applying
DCT then SVD. The secret image is also altered by DCT decomposition then SVD
transformation is applied on DCT coefficients in embedding process. Then after, a
watermarked image is produced by applying converse of all transform domain SVD
then DCT and DWT. The secret media can be extracted with recovery algorithm.
This procedure has been generously tried and evaluated against various attacks of
watermarking and it is found that it achieves better robustness and imperceptibility.

Keywords Digital image watermarking · Discrete wavelet transforms (DWTs) ·

Discrete cosine transforms (DCT) · Singular value decomposition (SVD) ·
Robustness

1 Introduction

The Internet becomes an attractive fortune of information to many billion consumers

in the era of twenty-first century. The instant growth of media information technolo-
gies has shined the way for transmission and data circulation worldwide. This digital
disruption has nurtured the escalation of prohibited actions like copyright protec-
tion, illegal distribution, unauthorized copying and rewriting of multimedia infor-
mation. In such a scenario security of senders, receivers, and content of electronic

N. Agarwal (B) · A. Kumar

Jaypee University of Information Technology, Solan, Himachal Pradesh, India
P. K. Singh
KIET Group of Institutions, Ghaziabad, Uttar Pradesh, India

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 769
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_56
770 N. Agarwal et al.

media is of serious significance [1]. Cryptography, steganography and watermarking

several data secure techniques have been designed to protect multimedia data. Digital
watermarking guarantees to shield the digital media from malicious attacks, piracy,
interfering and distribution of information. Digital watermarking is the process of
embedding the sensitive data into the accepting image without humiliating image
visual quality. After recovery process, this sensitive information or the watermark is
useful for content identification and copyright protection. Some believable applica-
tions of watermarking are electronic voting system, chip and hardware protection,
smart healthcare and numerous additional are shown in Fig. 1 [2–9].
Robustness, imperceptibility, embedding capacity and security are various perfor-
mance parameters for any watermarking algorithm [10]. Robustness can be described
as ability of an algorithm to recovered and authenticate a watermark after being
exposed to different attacks. Imperceptibility is defined as keeping the visual quality
of original image after it is exposed to many distortions in the presence of water-
marking algorithm. The amount of secret data that is feasible to embed into an original
image is called embedding capacity. These watermarking performance parameters
are independent to one another. Security is described by perseverance of water-
marking algorithm against intended or unintended attacks to recover or alter secret
information.
Based on the working are of watermarking, it is classified into two parts spatial
and transform domain [11]. Spatial domain methods found easy to embed, infor-
mation is directly embedded in pixel strengths of the original image. However,
this domain offers low embedding capacity and powerless to watermarking attacks.
Spatial domain techniques are least significant bit, patch work and local binary pattern

Fig. 1 Applications of watermarking [7]

Robust and Imperceptible Multiple Watermarking Using Transform … 771

and many more. In contrast to spatial domain, transform domain technique offers
better robustness, imperceptibility, higher embedding capacity. Transform domain
algorithms are DCT, DWT, SVD, redundant DWT (RDWT), lifting wavelet transform
(LWT) and divisive normalization transform (DNT) and many more.

2 Related Work

Some watermarking methods have been discussed for security of data in past decades.
Some motivating watermarking procedures for color images are discussed.
The author of [12] designed a robust and imperceptible watermarking based
on DWT and SVD. Two distinct watermarks were inserted in color image matrix
applying continuous, segmented and combined watermarking techniques. The author
of [13] proposed a Quaternion discrete cosine transform (QDCT), SVD with
computer-generated hologram-based watermarking method for color images. The
discussed technique founds secure and robust against many attacks. According
to Mahajan and Patil [14] they discussed an efficient hybrid watermarking tech-
nique with SVD and DWT. Firstly, the first secret information is fixed by second
secret mark then joint watermark is inserted into the host media. In Kumar et al.
[15] author proposed a watermarking algorithm based on RDWT, non-subsampled
contourlet transform (NSCT), SVD and set partitioning in hierarchical trees (SPIHT)
is described to achieve a robustness and imperceptibility. Arnold transform is also
discussed for more security. Designed algorithm gained robustness, imperceptibility
as well as high capacity. Wavelet-based watermarking is very much responsible for
highest robustness and are presented in [16]. The maximum functioning of wavelet
constructed watermarking scheme significantly lies on embedding and recovery
method. Some hybrid watermarking algorithm that uplifts the performance metrices
are presented by researchers [17–19].
A robust hybrid multiple watermarking schemes based on DWT, DCT and SVD
are discussed in this paper. This algorithm offers the satisfactory performance
measures such as PSNR and NC.

3 Theoretical Background

In this paper, a multiple image watermarking technique is proposed, grounded on

DWT, DCT and SVD. Firstly, an original image is decomposed by DWT and generate
a DWT coefficient then approximation band is selected for DCT then SVD. The secret
media (image) will be transformed by DCT. The reason behind DCT decomposition
of image watermark as it contains low frequency information and as long as no loss of
data while extracting the secret image from watermarked image. A concise narration
of these techniques is discussed in given below sections.
772 N. Agarwal et al.

Fig. 2 DWT illustration

[22]

4 Discrete Wavelet Transform (DWT)

The main concepts behind DWT it includes decaying of a media (image) in frequency
conduits of continual bandwidth on logarithmic gage. The 2-D signal like image
is decaying by DWT into a fixed number of four sub-bands (LL, LH, HL, HH)
that are non-overlapping multi resolution, approximation band (LL), horizontal
sub-band (LH), vertical sub-bands (HL) and diagonal sub-bands (HH) [20]. This
procedure can be repetitive to obtain a level of disintegration as shown in Fig. 2.
Approximation/lower frequency sub-bands are more delicate to human eyes [21].

5 Discrete Cosine Transform (DCT)

This transform divides an image matrix into three different parts of dissimilar regu-
larities low, high and middle frequency constants. In middle frequency, it is easy to
enclose the data that offers an extra challenge to the lossy compression techniques.
Due to energy compaction property, it makes DCT very familiar to use in image
watermarking [23]. This transform is used in both spatial and transform domain
watermarking system. Hence, 2-D transform can be depicted by following Eq. 1.

N −1 N −1
2  (2 p + 1)xπ (2q + 1)yπ
F(x, y) = c(x)c(y) f ( p, q) cos cos (1)
N p=0 q=0 2N 2N

where x, y, p, q = 0, 1, 2…, N − 1, and


√1for ∈= 0
c(x) = 2 (2)
1 Otherwise
Robust and Imperceptible Multiple Watermarking Using Transform … 773

The evaluation formula for inverse DCT (IDCT) is given by following Eq. 3:

N −1 N −1
2  (2 p + 1)xπ (2q + 1)yπ
F( p, q) = c(x)c(y)F(x, y) cos cos (3)
N x=0 y=0 2N 2N

6 Singular Value Decomposition (SVD)

The Eq. 4 shows the SVD transformation of an image in rectangular matrix, denoted
by P is as follows:

SVD(P) = U SV T (4)

where U and V are orthogonal matrices and S represents the diagonal matrix of same
size. SVD is usually working on the concept of finding singular value decomposition
of an image and then update the singular values to embed the watermark [8]. Nowa-
days, SVD can be merged with other techniques like DCT and DWT and getting
very trendy in watermarking applications.

7 Performance Procedures

Robustness and imperceptibility are two performance parameters for watermarking

algorithm [24]. A higher PSNR value denotes that watermarked media is more
analogous to original media. It also means that watermark is more imperceptible.

(Imax )2
PSNR = 10 log10 (5)
MSE
wherever I max is maximum pixel rate of an image and mean square error (MSE) is
defined in Eq. 6


M−1 N −1  2
1
MSE = P(m, n) − P̂(m, n) (6)
M × N m=0 n=0

where P is cover image matrix with size M × N and P̂ is watermarked media

(image) matrix with size of M × N. Good imperceptibility is received if cover and
watermarked image looks similar.
Normalized correlation is another performance parameter is used to calculate the
robustness, by estimating the resemblances between cover and secret media [8].
774 N. Agarwal et al.
 
i j W original(i, j)W recovered(i, j)
NCC =  
j W original (i, j)
2
i

Here, W (original) and W (recovered) depicts the cover and extracted watermarks.

8 Proposed Algorithm

The recommended procedure is the grouping of DWT, DCT and SVD, which
increases the robustness without substantial deprivation of the multimedia object
(image) quality in contradiction of the signal processing attacks. The discussed
method is the combination of two different process one is embedding and second is
extraction. Figures 3 and 4 shows the block diagram for embedding and extraction
procedures.

9 Experimental Results and Analysis

The size of cover and the secret image is 512 × 512 for testing the proposed method.
In Barbara image the embedding algorithm is grounded on DWT, DCT and SVD. In
this experiment PSNR and NC values are calculated at different gain factors (Fig. 5).
In Table 1 algorithm tested the PSNR and NC values at varying scale factor for
Barbara image. The proposed algorithm attains maximum PSNR is 37.69 dB at gain
factor 0.01 and maximum NC is 0.9998 at gain factor 0.1 without any attack. This
table shows that increase in gain value, imperceptibility of images is degraded while
improvement in robustness.
The Table 2 shows the outcome of proposed method for four different cover images
with different secret image at same gain value 0.05. The maximum and minimum
PSNR value achieved from image Lena is 37.65 and 35.98 from image Mandrill. The
maximum NC value attained is 0.9948 from image Lena and lowest NC is 0.9889
from Mandrill image.
The Table 3 shows the robustness achieved by proposed algorithm for different
attacks at different noise densities. The uppermost NC value 0.9995 is achieved at
JPEG-compression (QF 90). The lowest NC is 0.6998 obtained by Salt and Pepper
attack at noise density 0.05.

10 Conclusion

In this paper a novel robust multiple watermarking using transform domain tech-
niques such as DWT, DCT and SVD instead of utilizing DWT, DCT and SVD
Robust and Imperceptible Multiple Watermarking Using Transform … 775

Fig. 3 Embedding
algorithm
776 N. Agarwal et al.

Fig. 4 Extraction algorithm

Fig. 5 (a–d) Lena, Barbara,

Mandrill and Peppers images
used as cover and secret
media

Table 1 Performance
Image Gain factor PSNR NC
evaluation of proposed
method at varying gain value Barbara 0.01 37.69 0.9936
Barbara 0.05 37.58 0.9958
Barbara 0.1 36.12 0.9998
Robust and Imperceptible Multiple Watermarking Using Transform … 777

Table 2 Performance evaluation for different images at same gain value

Cover image Watermark image Gain factor PSNR NC
Lena Lena 0.05 37.65 0.9948
Peppers Peppers 0.05 36.27 0.9938
Mandrill Peppers 0.05 35.98 0.9889
Barbara Lena 0.05 37.55 0.9942

Table 3 Performance
S. No. Attacks Noise density NC value
evaluation of watermarking
algorithm against attacks 1 JPEG-compression QF = 10 0.9613
from image Barbara 2 JPEG-compression QF = 50 0.9728
3 JPEG-compression QF = 90 0.9995
4 Salt and pepper 0.01 0.7989
5 Salt and pepper 0.05 0.6998
6 Salt and pepper 0.001 0.9978
7 Gaussian noise 0.01 0.7883
8 Filtering 1×1 0.9992
9 Scaling 1.1 0.7651

in combination is presented for color images. For security reason a secret image is
embedded into the cover media to produce a watermarked image, it achieves an appro-
priate performance measure in terms of robustness and imperceptibility. Designed
algorithm may discover possible application in healthcare systems, electronic voting
systems, video transmission and many more.

References

1. Ray A, Roy S (2020) Recent trends in image watermarking techniques for copyright protection:
a survey. Int J Multimedia Inf Retrieval 1–22
2. Sharma A, Singh AK, Kumar P (2018) Combining Haar wavelet and Karhunen Loeve
transforms for robust and imperceptible data hiding using digital images. J Intell Syst
27(1):91–103
3. Mousavi SM, Naghsh A, Abu-Bakr SAR (2014) Watermarking techniques used in medical
images: a survey. J Digit Imaging 27(6):714–729
4. Horng SJ, Rosiyadi D, Fan P, Wang X, Khan MK (2013) An adaptive watermarking scheme
for e-government document images. Multimedia Tools Appl 72(3):3085–3103
5. Mohanty SP, Sengupta A, Guturu P, Kougianos E (2017) Everything you want to know about
watermarking: from papers marks to hardware protection. IEEE Consum Electron Mag 6(3):83–
91
6. Singh AK, Dave M, Mohan A (2014) Wavelet based image watermarking: futuristic concepts
in information security. Proc Natl Acad Sci, India, Sect A 84(3):345–359
778 N. Agarwal et al.

7. Agbaje MO, Awodele O, Ogbonna AC (2015) Applications of digital watermarking to

cyber security (cyber watermarking). In: Proceedings of informing science and IT education
conference (InSITE), pp 1–11
8. Singh D, Singh SK (2017) DWT-SVD and DCT based robust and blind watermarking scheme
for copyright protection. Multimedia Tools Appl 76(11):13001–13024
9. Singh AK, Kumar B, Dave M, Mohan A (2015) Robust and imperceptible dual watermarking
for telemedicine applications. Wireless Pers Commun 80(4):1415–1433
10. Bhatnagar G (2012) A new facet in robust digital watermarking framework. AEU—Int J
Electron Commun 66(4):275–285
11. Begum M, Uddin MS (2020) Digital image watermarking techniques: a review. Information
11(2):110
12. Mohananthini N, Yamuna G (2016) Comparison of multiple watermarking techniques using
genetic algorithms. J Electr Syst Inf Technol 3(1):68–80
13. Li J, Lin Q, Yu C, Ren X, Li P (2018) AQDCT and SVD based color image watermarking scheme
using an optimized encrypted binary computer-generated hologram. Soft Comput 22(1):47–65
14. Mahajan LH, Patil SA (2013) Image watermarking scheme using SVD. Int J Adv Res Sci Eng
2(6):69–77
15. Kumar C, Singh AK, Kumar P, Singh R, Singh S (2020) SPIHT—based multiple image
watermarking in NSCT domain. Concurrency Comput Pract Experience 32(1)
16. Lai CC, Tsai CC (2010) Digital image watermarking using discrete wavelet transform and
singular value decomposition. IEEE Trans Instrum Meas 59(11):3060–3063
17. Lin WH, Horng SJ, Kao TW, Fan P, Lee LC, Pan Y (2008) An efficient watermarking method
based on significant difference of wavelet coefficients quantization. IEEE Trans Multimedia
10(5):746–757
18. Awasthi M, Lodhi H (2013) Robust image watermarking based on discrete wavelet trans-
form, discrete cosine transforms and singular value decomposition. Adv Electr Electron Eng
3(8):971–976
19. Golshan F, Mohammadi K (2013) A hybrid intelligent SVD-based perceptual shaping of digital
image watermark in DCT and DWT domain. Imaging Sci J 61(1):35–46
20. Langekaar GC, Setyawan I, Lagendijk RL (2000) Watermarking digital image and video data.
A state-of-the-art overview. IEEE Signal Process Mag 17(5):20–46
21. Ali M, Ahn CW (2018) An optimal image watermarking approach through cuckoo search
algorithm in wavelet domain. Int J Syst Assur Eng Manage 9(3):602–611
22. Daphal SD, Aher PK, Galande SB, Jagtap SK (2018) Comparison of key transform domain
watermarking methods based on performance measures. In: International conference on
communication, information and computing technology (ICCICT). IEEE, pp 1–6
23. Abdulrahman AK, Ozturk S (2019) A novel hybrid DCT and DWT based robust watermarking
algorithm for color images. Multimedia Tools Appl 78(12):17027–17049
24. Khan M, Masood F (2019) A novel chaotic image encryption technique based on multiple
discrete dynamical maps. Multimed Tools Appl 78(18):26203–26222
Blockchain-Driven and IoT-Assisted
Chemical Supply-Chain Management

Pronaya Bhattacharya, Ashwin Verma, and Gulshan Sharma

Abstract Blockchain (BC) technologies have gained prominence in the scientific

and industrial sectors, as they have the potential to benefit a wide range of sectors. The
shift is attributed owing to the key benefits of immutability, chronology, and owner-
ship at every point in the supply chain. Thus, through BC, and smart contracts (SCs),
every point in the supply chain can capture and exchange transactional data, and thus,
it enables efficient and automated chain operations. Thus, BC adoption introduces
traceability and transparency and lowers costs. Owing to the potential benefits of BC,
promising solutions can be adopted in the chemical industry sector, where BC-based
solutions can be potentially used to manage inventory management, to logistics, and
ensure that fake chemicals are not sold by malicious entities, without the require-
ment for third-party validation or authorization. Moreover, chemical products can be
assigned low-powered Internet-of-Things (IoT) digital tags that simplify the chem-
ical inventory operations. Thus, in this paper, we propose a BC-based IoT-assisted
reference architecture for chemical industries and present the entire supply-chain
operation. To support the claim, a practical case study, ChemBlocks is presented, that
presents the potential benefits of integration of IoT tagging on chemical products,
with ledger state information managed via BC. At every supply-chain point, SCs are
executed that automated the transfer of funds, and asset transfer is achieved. The
proposed case study eliminates cost and effort duplication and increases the return
of investment (ROI) of chemical industries.

Keywords Blockchain · Chemical industries · Internet-of-Things ·

Manufacturing supply chain · Smart contract · Smart tagged objects

P. Bhattacharya (B) · A. Verma

Institute of Technology, Nirma University, Ahmedabad, Gujarat, India
e-mail: pronoya.bhattacharya@nirmauni.ac.in
A. Verma
e-mail: ashwin.verma@nirmauni.ac.in
G. Sharma
Durban University of Technology, Durban, South Africa

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 779
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_57
780 P. Bhattacharya et al.

1 Introduction

In smart manufacturing ecosystems, industries have shifted toward smart and auto-
mated supply-chain builds, where the industrial processes are automated through
cyber-physical processes (CPS) that connect industrial stakeholders over heteroge-
neous networked applications. In such supply-chain builds, trusted supply relation-
ships have gained prominence owing to the increased challenges of large distances,
product genuineness, and quality of the product. In the chemical industries, chemi-
cals are produced in mass quantities, where the raw chemicals are provided by basic
chemical suppliers and are then processed and produced in chemical plants. These
raw chemicals can be forged, owing to the different quantities of processed chemi-
cals, and sold to manufacturers at a higher cost. Thus, fake chemical generation and
detection is a timely process, and this affects the entire supply-cycle [7].
Once the chemicals are packed, they are assigned low-powered Internet-of-Things
(IoT)-based tokens, that are barcodes, quick-response (QR), or near-field commu-
nication (NFC) codes assigned on packaged chemicals, and sent to manufacturer
plant [5]. Figure 1 represents the generic chemical supply-chain cycle and highlights
the main challenges in the ecosystem. From there, the final deliverable chemicals
are produced, and products are packed, assigned IoT digital tokens, and sent to the
warehouse. Once the chemicals reach the warehouse, they are sold to small-scale

Fig. 1 Challenges in distribution in the generic supply chains in chemical industries

Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 781

industries from open markets. Here, illegal hoarding and black-marketing of chemi-
cals are possible, by reducing the supply of chemicals to open markets, the prices are
hiked by third-party vendors. Once the hiked chemicals are bought by small-sized
industries by such third-part vendors, they are not assured of the manufacturer, and
ownership is not clear, in case of inferior quality chemicals. The prices are further
hiked by small-sized vendors, to cover their logistic cost, and are sold at very high
prices at supermarkets to general customers [8]. Thus, the global supply-chain pro-
cess has to assure trust, transparency, and ownership through the entire chain, so that
the entire process is visible to end customers.
Blockchain (BC) technology provides a means of building trust between suppli-
ers and customers along the chemicals supply chain. BC allows verified provenance
of the source of raw materials, tracking and tracing of processed chemicals from
suppliers, through assures the quality control and inventory details to be added in
the immutable transaction ledgers [10]. Thus, it allows that the desired product qual-
ity is maintained over the entire supply chain points, and counterfeit chemicals are
detected, and the tracing and ownership to the malicious stakeholders are also iden-
tified [1]. In the chemical supply-chain ecosystems, we can document the entire
process, and transactional payments between any two supply points can be done via
smart contracts (SCs). SCs include the supplier, manufacturer, and packaging origin
and create a single truth point across the entire ledger [2].
At each point of the supply chain, we consider that on every chemical product, IoT
tags are attached. IoT tagging allows contactless reading and writing of information.
Data that is written onto an IoT tag can be encrypted and published in BC, and
stakeholders can determine the proper access control on the data sharing, despite the
information being stored at distributed and heterogeneous networked locations [14].

1.1 Motivation

In chemical industries, drug composition to prepare the chemical in the manufactur-

ing plant is a critical process. Currently, chemical industries are suffering from the
issue of the genuineness of raw materials, high costs due to illegal hoarding and dis-
tribution. Owing to these limitations, BC is a preferred choice, and the integration of
BC in chemical industries is still in its infancy. Thus, the paper discusses a reference
architecture that integrates BC at different supply points in the distribution ecosystem
and assures trust and transparency in the entire supply chain. At any supply-chain
point, SCs assure that the required operating conditions of chemical storage, distri-
bution, and collection are preserved. The paper also proposes a real-time case study
that presents an industry use-case deployment of BC in the chemical industry. The
potential benefits and issues and challenges are discussed for practical realization of
BC-driven chemical industry supply-chain mechanisms.
782 P. Bhattacharya et al.

1.2 Research Contributions

The following are the research contributions of the paper.

• A BC-based and IoT tag-driven reference architecture is proposed. Through IoT tag
objects, we assure a low-cost solution of digitizing the chemical packaging records,
and information of the source and destination endpoints. Once the chemical records
(CR) are fetched, the meta-information is stored as a transactional ledger in BC.
• A case study of BC-driven chemical industry, named as ChemBlocks is discussed
that presents the practical deployment conditions and the SCs that are proposed at
different points of the supply chain.

1.3 Article Structure

The chapter is divided into five sections. Section 2 presents the proposed reference
architecture that integrates BC and IoT to assist low-powered traceable supply-chain
operations. Section 3 presents the open issues and potential challenges of the deploy-
ment. Section 4 presents the discussion of a proposed case study, ChemBlocks for
chemical supply-chain ecosystems. Section 5 concludes the chapter.

2 BC-Driven and IoT-Assisted Chemical Supply-Chain

Management: The Reference Architecture

In this section, we present the reference architecture of BC-driven and IoT-assisted

chemical distribution and supply-chain management. The details are presented as
follows.

2.1 The Entity Model

In the propose ecosystem, we consider the following entities.

• Chemical supplier: The chemical supplier, denoted by E CS , distributes the raw
chemicals that are mixed in appropriate proportions in the chemical manufacturing
plant.
• Chemical manufacturer: The chemical manufacturer, denoted as E CM , collects
the raw chemicals from E CS , processes them in desired proportion, and pre-
pares the final raw chemical, denoted by FC . FC is then packaged into chemi-
cal containers, and a RFID token TFC is attached. TFC contains the attributes like
{Fid , Fcom , FDOM , FDOE , Fcost }, where Fid denotes the product ID, Fcom denotes
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 783

the chemical composition information, FDOM denotes the date of packaging, FDOE
denotes the expiry date, and Fcost denotes the chemical cost.
• Logistics head: Once E CM prepares TFC , the packages are collected and trans-
ferred for shipments through the land, water, and air. The entire shipment process
is handled by the logistics head, denoted by E LH . E LH is also responsible to handle
cross-border financial settlements, duty and toll-taxes, and payment settlement to
intermediaries who assist in the shipment process. E LH delivers FC to the manu-
facturer, who is responsible for preparing the chemical product.
• Suppliers and buyers: Once the chemical product is ready from FC , denoted as
CP, SCs are initiated depending on the product orders. Seller denotes E CM , from
which buyers (market buyers), purchase the product. The buyers, denoted as E B ,
purchases the final product that includes the total cost C as T (E CS + C(E LH +
C(MCP ) + T , where C(E CS ) denotes the raw chemical packaging cost, C(CLH )
denotes the logistics cost, and C(MCP ) denotes the manufacturing cost of C P,
and T denotes the incurred taxes and other operational costs. In BC, this cost C is
transparent between E CS , E LH , E CM , E B .
• End-customers: End-customers, denoted by E CU , purchases CP from open mar-
kets, or supermarkets. The final cost, FC can be computed as FC = C + E LH +
C(S) + T , where C(S) denotes the cost of product maintenance at supermarkets.
As C is transparent to E CU via BC, and the cost of E LH , T , and C(S) is entered
by respective stakeholders in BC, the FC is transparent.

2.2 The Layered Reference Architecture

In this subsection, we discuss the reference architecture as a layered model between

the entities. In the reference architecture, we consider three layers, namely the IoT
layer, the supply-chain layer, and the contract layer. Figure 2 presents the architecture
details.
IoT Layer—At this layer, we consider two supply-chain points. One point is between
E CS and E CM , and the second one is between E CM and E B . At both supply points,
we consider the raw chemical is packaged and processed into containers and are
assigned a digital token identity, denoted as TFC . The digital identity can be an RFID
barcode, QR, or NFC code, that can be processed through a low-powered reader
device Rd . The containers are shipped through E LH to destination points. To assist
low-powered networking, we consider an IoT protocol stack S that consists of a
communication module (ZigBee, WiFi, or 6LowPAN-assisted addresses). We con-
sider IPv6 addressing module that communicates through low-powered personal area
networks like ZigBee, WiFi, and 6LowPAN addresses, with information of source
IPv6 and destination IPv6, for packet routing and delivery. At the transport layer,
we consider real-time barcode bytes streaming assisted through either TCP or UDP,
depending on the connection approach. The packet payload information is encrypted
through lightweight symmetric ciphers, like elliptic curve cryptography (ECC), or
advanced encryption standard (AES), or implementors might prefer trusted execu-
784 P. Bhattacharya et al.

Fig. 2 Reference
architecture of IoT-assisted
BC-driven chemical
supply-chain ecosystem

tion environments (TEE). For service discovery, we consider the service discovery
protocol (SDP), or multicast DNS (mDNS) approach [6].
Supply-Chain Layer—At this layer, we consider a microservice functional environ-
ment, for specific functionalities like asset ownership functionality, chemical meta-
information {Fid , Fcom , FDOM , FDOE , Fcost }, logistics cost functionality that includes
shipments via land, water, and air cargo facilities, over international land, air, and
water boundaries. Here, the microservice handles the internal payment information
through digital cryptocurrencies. This allows the payments to be traceable, and in
the case of multi-currency scenarios, an exchange server ES is formed that converts
one form of cryptocurrency to other, based on the current conversion rates. Another
microservice involves the chemical stock inventory service. It includes information
{OS , TS , FS }, where OS denotes the ordered quantity of stock, TS denotes the deliv-
ered quantity, and FS denotes the bad-stock quantity. Overall, we consider that k
micro-services are operational, denoted as MS = {MS1 , MS2 , . . . , MSk }.
Each microservice has an associated local database DMSk that maintains the state
and ledger information [4]. MS communicates with each other through event-driven
application programming interface (API) gateways, which involves two event han-
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 785

Fig. 3 SCs interface among the communicating entities

dlers, one for entity service handler that allows chemical stakeholders to loosely
interact with MS via the web gateways. The other event handler is the management
service handler, which allows event segregation, and web-event data handling, for-
matted exchange, and information caches. The implementations might adopt the rep-
resentational state transfer (REST) web-interface communication points that allow
stateless client–server HTTP request–reply paradigm. The session state information
is cached at the client, which allows faster query response for frequently selected
objects [12]. REST allows the functionality of transferable executable codes to the
server end. The data exchange is normally preferred in JavaScript object notation
(JSON) files. In simple object access protocol (SOAP), the web-communication
interface follows the transactional states (ACID properties) and presents security
interface endpoints through web-service security (WS-security), and web services
description language (WDSL) [3].
The Smart Contract Layer—At this layer, we consider the transactional ledgers
Tl are maintained between the chemical stakeholders. We maintain the ledger state
information Il = {Aid , Od , Tid , Q, P}, where I denotes the ledger information, Aid
denotes the asset-identifier, Od denotes the ownership records, Tid denotes the asset
transfer information that contains the information of final ownership, Q denotes the
final quantity of FC , or CP transferred between different supply-chain points, and P
786 P. Bhattacharya et al.

denotes the price of the asset. The information is managed through a permissioned
contract-chain information between two entities E 1 and E 2 , as depicted in Fig. 3. As
depicted in the figure, we consider the contract between E 1 and E 2 , depicted as buyers
and sellers in the contract C. We consider the transfer of chemical product CP as asset
information. The contract conditions are finalized between E 1 and E 2 , and payment
channel mode is setup. We consider a permissioned SC transfer mode of assets,
and information is maintained through ledger states L s . The contract deployment
contains three functionalities, assetContract interface, CreateChemicalAsset, and
transferAsset.
The information is stored in world-state as a response, and read–write buffers are
updated once the contact conditions are fulfilled. A channel setup C is established
for secure communication between E 1 and E 2 , and it maintains the privacy of asset
transfer. The appropriate gas fee for contract execution is deducted, and the chemical
asset ownership is transferred. The wallets W E1 and W E2 are, respectively, debited
and credited once the contract is finalized. The contract information is published in a
distributed ledger, known as interplanetary file systems (IPFS), and the information
is visible to all authorized stakeholders, whose entries are considered in the access
control lists [11]. The meta-information of the recorded ledger in IPFS is then stored
as an unconfirmed transaction in Mempool [13]. Miners E m proposes the mining
phase, to initiate a block proposal B to the added to BC. To assess a record stored
in IPFS, the hash of the meta-information is referenced from BC, and the IPFS key
is fetched. Any authorized stakeholder then assesses the IPFS through two sets of
keys, one is the IPFS key and the other is the users private key.

3 Open Issues and Challenges

In this section, we discuss the potential issues of BC adoption in chemical supply-

chain ecosystems.
1. Node scalability—In chemical supply-chain ecosystems, owing to huge amount
of data transfer among different stakeholders, the BC-governance model has
to assure lightweight resource requirements for node validations. Recently,
lightweight consensus approaches like IOTA, Merkle-DAG, and others are pro-
posed, but there is no uniformity of the consensus protocols with the data gov-
ernance model of heterogeneous applications.
2. Privacy requirements—In BC, the data is transparent and auditable. Thus, autho-
rized stakeholders, based on access-list permissions, can view the ledger data.
However, the access control mechanisms in BC are not universal, and most SC
implementations do not have access control privilege. Thus, the privacy of data
is not controllable in most implementations. Moreover, user anonymity is not
guaranteed, as in public chains, the transaction ledger is visible through a public
wallet address. Recently, industry practitioners have shifted toward controlled
lock mechanisms in BC that proposes lock mechanisms on data transfer, access,
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 787

and ownership. However, the open-source implementation is still in its infancy,

and most solutions are proprietary.
3. Operational cost—The next issue is the overall operational cost (OPEX) of the
BC-based supply-chain ecosystems. In the chemical industries, the supply chain
might consist of a lot of intermediate supply-chain points, and many stakeholders
are involved in the entire process. The distribution of chemicals transparently,
with assured chronology among all supply points, would involve high transac-
tional costs, due to block mining and validations. Moreover, the block mining
time is proportional to the mining difficulty, or the miners with a high amount
of stakes. Miners with high computational resources might collude with other
stakeholders in the chain, to monopolize the mining process, thereby increasing
the price of the chemicals.
4. Security attacks—BC is considered secure, but it is vulnerable to security attacks
like mining attacks, power attacks (51% attack), validator-based attacks, and
collusion attacks. Moreover, SCs are programmable codes that are vulnerable
to flaws in code logic that might induce vulnerabilities like gas-attacks, con-
tract flow-attack, injection-attack, dependency-injection, name-space pollution
attacks, re-entracy, and many more. Thus, formal verification of SCs against the
possible attacks should be considered before the final deployments. However,
the identification of attack vectors is complex in BC ecosystems.
5. Data management—In the chemical supply chain, a chemical supplier (sender)
may transact in Bitcoin wallets, and the receiver wallet might be accepting
Ethereum. Thus, the data management among the transactional ledgers would
require effective exchange server integration in the payment system to perform
the interconversion in real-time. At present, the conversion exchanges are limited
and are not supported in all supply chain ecosystems.
6. Consensus approach—With the emergence of assisted technologies like IoT,
artificial intelligence, and big data ecosystems, the design of generic consensus
approaches that operate with the collected data from these platforms is in the
early stages. In such ecosystems, BC leverages low transactional throughput
and mostly is not suitable for public-chain systems. Alternative designs involve
the design of permissioned consensus, but it reduces the BC-openness and trans-
parency flow among all supply-chain stakeholders. Thus, the design of optimized
public consensus schemes, that manages the node throughput is an open issue.
7. Cross-boundary legislation—Due to international regulations and government
legislation, cryptocurrency is not considered as a monetary medium of transfers
in many countries. To fully realize the potential of global supply-chain opera-
tions, logistics would require cross-boundary uniformity in legislative rules that
involves payments via cryptocurrency tokens.
788 P. Bhattacharya et al.

Fig. 4 ChemBlocks: the operational flow of the propose case study

4 ChemBlocks: A Case Study on RFID-Assisted

Smart-Tagging in BC-Leveraged Chemical Supply
Chains

In this section, we present an industry-ready practical deployment case study scenario

of BC-based chemical supply-chain distribution and management operations [9]. The
deployment, named as Makerchain, is designed to propose a cyber-driven payment
model lifecycle, with SCs auditable trail history. We have modified the proposed
model and have included the smart tagging based objects, and have proposed a
model, named as ChemBlocks, that integrates RFID-assisted smart tagging at every
supply-chain point, with the transactional meta-information stored in BC-ledgers.
Figure 4 presents the detailed flow of the proposed case study. The components of
the model are discussed in detail as follows.

4.1 Ordering of Raw Material

As discussed in Sect. 2, we use the same entity notations in the case study for better
readability. We consider the raw chemical composition as Com(R), decided by E CS ,
and Q units of R are ordered by E CM . At this point, we maintain a SC1 transac-
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 789

tional asset information as <asset:R, quantity:Q >. The SC ledger of E CS includes

<Sup_data, Order_ID>, which is orderID, and supplier information, and E CM veri-
fies the SC ledger, and appends < M_id> state to the world-ledger. R is then pack-
aged into containers P(R), and is tagged via RFID-tokens, denoted as TFC , and is
sent to outbound process C. The SC information is recorded as Tx.State in Ethereum
virtual machine (EVM) ledger as byte-code, and meta-inf is stored in IPFS, and
block is proposed.

4.2 Logistics: Outbound and Inbound

Outbound C state is maintained as asset < TFC , Q >, and P(R) is attached with
< C_ID, Tag_ID, Price>, and is sent to E LH that ships and manages the cross-border
settlements through cryptocurrency exchange wallets. At inbound I , the shipment
verification is done through a barcode reader, and SC2 transactional ledger state is
maintained and recorded in BC between E CM and E LH .

4.3 Final Chemical Packaging and Outbound to Open

Markets

E CM prepares the final chemical product CP from FC in T quantities, which is RFID-

tagged, and is outbound to E B , with logistic intermediary E LH . The cost of the CP,
plus the cost of logistics, is included and tagged into the packaged product, with
information < C_ID, Tag_ID, Price>. The supplier E CM data Sup_data is included
in SC3 asset information, and recorded in BC.

4.4 Inbound at Open Markets and Delivery to End-Customers

E LH delivers the final packaged product in W quantities, with the constraint W < T ,
as we assume T − W items are damaged in transit. An asset < FP, W > is created
between E CM and E B , and E B updates its inventory stock information as <Sup_data,
S_ID, E_B>, and stores the chemical products at markets. The packaged product
is then bought by the end-customer, E CU . As all SC Tx.States is recorded as block
proposals in public chain, E CU has a view right of the entire supply-chain chronology,
and thus, it ensures complete fairness and eliminates any sort of illegal marketing,
middlemen, and hoarding of chemical products that inflate the prices of chemical
product.
790 P. Bhattacharya et al.

5 Conclusion

In chemical supply-chain ecosystems, BC is gaining widespread prominence to

improve the transparency of supply-chain endpoints. Coupled with smart IoT-assisted
tagging, it streamlines and automates the entire record and inventory maintenance,
with ease in data payment mechanisms, owing to SCs programmable logic. This
improves the efficacy of supply chains and lowers production costs. Motivated by
these facts, in the paper, we presented a reference architecture of possible BC deploy-
ment in chemical supply chains and discussed the key points in the ecosystem. The
open issues and challenges are further discussed, which presents the readers with
emerging gaps to leverage research in a similar direction. Finally, we presented a
supporting case study, ChemBlocks, that discusses the integration of BC and RFID-
smart tagging in the ecosystem.
As part of future work, we would propose security models to support the data pro-
tection against possible attack boundaries, with the inclusion of anti-counterfeiting
and copyright protection models. It would allow an end-to-end holistic vision toward
the design of formal BC-based chemical supply chains in the chemical sector.

References

1. Al-Jaroodi J, Mohamed N (2019) Blockchain in industries: a survey. IEEE Access 7:36500–

36515
2. Bodkhe U, Mehta D, Tanwar S, Bhattacharya P, Singh PK, Hong WC (2020) A survey on
decentralized consensus mechanisms for cyber physical systems. IEEE Access 8:54371–54401.
https://doi.org/10.1109/ACCESS.2020.2981415
3. Brown R (2021) Connecting energy management systems to enterprise business systems using
soap and the xml web services architecture. In: Web based energy information and control
systems. River Publishers, pp 447–458
4. Cerny T, Donahoo MJ, Trnka M (2018) Contextual understanding of microservice architecture:
current and future directions. SIGAPP Appl Comput Rev 17(4):29–45. https://doi.org/10.1145/
3183628.3183631
5. Farsi M, Latsou C, Erkoyuncu JA, Morris G (2020) RFID application in a multi-agent
cyber physical manufacturing system. J Manuf Mater Process 4(4). https://doi.org/10.3390/
jmmp4040103, https://www.mdpi.com/2504-4494/4/4/103
6. Khalid M, Mujahid U, Najam-ul Islam M, Choi H, Alam I, Sarwar S (2021) Ultralightweight
resilient mutual authentication protocol for IoT based edge networks. J Ambient Intell Human-
ized Comput 1–12
7. Khan SAR, Zkik K, Belhadi A, Kamble SS (2021) Evaluating barriers and solutions for social
sustainability adoption in multi-tier supply chains. Int J Prod Res 59(11):3378–3397. https://
doi.org/10.1080/00207543.2021.1876271
8. Kleinekorte J, Fleitmann L, Bachmann M, Kätelhön A, Barbosa-Póvoa A, von der Assen N,
Bardow A (2020) Life cycle assessment for the design of chemical processes, products, and
supply chains. Annu Rev Chem Biomol Eng 11(1):203–233. https://doi.org/10.1146/annurev-
chembioeng-011520-075844, pMID: 32216728
9. Leng J, Jiang P, Xu K, Liu Q, Zhao JL, Bian Y, Shi R (2019) Makerchain: a blockchain
with chemical signature for self-organizing process in social manufacturing. J Cleaner Prod
234:767–778. https://doi.org/10.1016/j.jclepro.2019.06.265, https://www.sciencedirect.com/
science/article/pii/S0959652619322309
Blockchain-Driven and IoT-Assisted Chemical Supply-Chain … 791

10. Patel NS, Bhattacharya P, Patel SB, Tanwar S, Kumar N, Song H (2021) Blockchain-envisioned
trusted random oracles for IoT-enabled probabilistic smart contracts. IEEE Internet Things J
8(19):14797–14809. https://doi.org/10.1109/JIOT.2021.3072293
11. Patel SB, Bhattacharya P, Tanwar S, Kumar N (2021) Kirti: a blockchain-based credit recom-
mender system for financial institutions. IEEE Trans Network Sci Eng 8(2):1044–1054. https://
doi.org/10.1109/TNSE.2020.3005678
12. Riaz U, Hussain S, Patel H (2021) A comparative study of rest with soap. In: Fu W, Xu Y,
Wang SH, Zhang Y (eds) Multimedia technology and enhanced learning. Springer International
Publishing, Cham, pp 485–491
13. Shukla A, Bhattacharya P, Tanwar S, Kumar N, Guizani M (2020) Dwara: a deep learning-based
dynamic toll pricing scheme for intelligent transportation systems. IEEE Trans Veh Technol
69(11):12510–12520. https://doi.org/10.1109/TVT.2020.3022168
14. Yuvaraj S, Sangeetha M (2016) Smart supply chain management using internet of things (IoT)
and low power wireless communication systems. In: 2016 International conference on wireless
communications, signal processing and networking (WiSPNET), pp 555–558. https://doi.org/
10.1109/WiSPNET.2016.7566196
Correction to: Impact of Green
Communication and Technology System

Sumit Dhariwal and Avinash Raipuria

Correction to:
Chapter “Impact of Green Communication and Technology
System” in: P. K. Singh et al. (eds.), Emerging Technologies
for Computing, Communication and Smart Cities, Lecture
Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_44

In the original version of the book, the author Sumit Dhariwal’s affiliation has been
changed from “Manipal University, Jaipur, India” to “Manipal University Jaipur,
Rajasthan, India” in the Chapter “Impact of Green Communication and Technology
System”.
The chapter and book have been updated with the changes.

The updated original version of this chapter can be found at

https://doi.org/10.1007/978-981-19-0284-0_44

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 C1
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0_58
Author Index

A Baljeet Kaur, 649

Aalif Hamid, 151 Balusu Srinivasa Rao, 375
Abhijit Sinha, 633 Baseem Khan, 233
Abouhawwash, Mohammad, 405 Bharti Bisht, 547
Abubakar Ahmad Musa, 563 Bharti Rana, 715
Adarsh Kumar, 389 Bhaskar Dubey, 591
Adeniyi, Abidemi Emmanuel, 505 Bhaumik Thakkar, 405
Adesola, Falade, 183, 193 Bhavani, R., 683
Ahad, Umaira, 757
Ahmed, Kharimah Bimbola, 203
Alfa, Abraham Ayeba, 203 D
Amit Kumar, 769 Damasevicius, Robertas, 203
Amit Pandey, 533 Deepa, D., 81
Amit Sehgal, 259 Deep Upadhyaya, 329
Anand Nayyar, 389 Deepti Saraswat, 519, 731
Anand Shivam, 165 Devam Kumar Sinha, 671
Anil Kr. Tiwari, 533 Dilip Kumar Sharma, 123
Ankur Choudhary, 563 Divyansh Bansal, 661
Anshika Goyal, 95 Divyanshu Singh, 419
Anukriti Singh, 671 Dixit Patel, 329
Arpana Rawal, 107
Arpita Tripathy, 95
F
Arun Solanki, 217
Farnazbanu Patel, 405
Aryabhatta Ganguly, 137
Ashwin Verma, 233, 519, 731, 779
Attah, Blessing Iganya, 203 G
Atufaali Saiyed, 617 Garima Jain, 275
Avinash Raipuria, 605 Gawhar Hameed, 715
Azeta, Ambrose, 183, 193 Geethanjali, D., 683
Gulshan Sharma, 779
Gunjan Chhabra, 31
B
Babita Jajodia, 419
Babita Yadav, 3 H
Bachwani Dhiraj, 329 Hardik Dave, 731
© The Editor(s) (if applicable) and The Author(s), under exclusive license 793
to Springer Nature Singapore Pte Ltd. 2022
P. K. Singh et al. (eds.), Emerging Technologies for Computing, Communication
and Smart Cities, Lecture Notes in Electrical Engineering 875,
https://doi.org/10.1007/978-981-19-0284-0
794 Author Index

Hari Mohan Rai, 491 Namrata Dhanda, 671

Harish Kumar, 41 Neelu Jyothi Ahuja, 389
Harmandeep Kaur, 259 Neerav Pawar, 577
Himanshu Sharma, 491 Netik Bhardwaj, 491
Hiresh Gupta, 291 Nidhi Sinha, 633
Husen Saifibhai Nalawala, 481

O
I Ogundokun, Roseline Oluwaseun, 505
Inaam Zia, 533 Oluranti, Jonathan, 203, 505
Iram Warsi, 291 Omolola, Ademola, 183, 193
Oni, Aderonke, 183, 193

J
Jai Prakash Verma, 303 P
Jaymin Shah, 481 Pallavi Arora, 649
Jitali Patel, 591 Parita Oza, 481
Joshiram, T., 347 Parul Gandhi, 547
Junaid Maqbool, 55 Pathak, V. K., 41
Jyotsna Sengupta, 467 Pawan Singh, 533
Pooja Anand, 317, 747, 757
Pooja Gohil, 165
K Poonam Tanwar, 95, 577, 661
Karan Patel, 731 Pradeep Kumar Singh, 769
Keerthi Samhitha, B., 81 Preeti Aggarwal, 15, 55
Khushi Patel, 617 Priya Mishra, 247
Komal Shah, 233 Priya, R., 683
Kuldeep Limbachiya, 165 Pronaya Bhattacharya, 233, 405, 519, 731,
779

L
Laskhmi Kumari, S. V. R., 347 R
Lokesh Kumar, 491 Raghav Jagya, 661
Lokesh Nagar, 69 Rashi Agarwal, 41
Luxmi Sapra, 31 Rashmi Mishra, 275
Ravin Ahuja, 183, 193, 203, 505
Ravreet Kaur, 15, 55
M Rayeesa Muzafar, 747
Madhabi Ganguly, 137 Richa Dhagat, 107
Madhulika Bhatia, 247 Ritendra R. Sawale, 69
Maher Thakkar, 303 Rohit Tanwar, 31
Manish Chauhan, 491
Meenatchi Sundram, 633
Misra, Sanjay, 183, 193, 203, 505 S
Mohammed Shakil Malek, 165, 329 Sachin Gupta, 3
Mohit Sharma, 247 Sachin Kumar, 451
Monika, 291 Sahitya Diwakar, 661
Monika Thapliyal, 389 Samprat Bhavsar, 731
Ms Shweta Sharma, 577 Santoshi Raju, V., 361
Mustansir Nizam, 439 Satish Saini, 259
Saumya Tripathi, 123
Shafqat Nabi Mughal, 439
N Shahidul Haq, 715
Namita Agarwal, 769 Shambhavi Mishra, 701
Author Index 795

Sharnil Pandya, 165 Thangella Aravind, 375

Shashwat Singh Mann, 247
Shivam Punjani, 731
Shrutika Chawla, 15 U
Shubhangi Verma, 451 Umesh Bodkhe, 519
Shubhankar Rahi, 577 Urvashi Tyagi, 95
Shweta Sharma, 661
Siddhant Patel, 303
Simran Jakhodia, 419 V
Simrat Walia, 467 Vaibhav Kalra, 577
Singh, O. P., 451 Vaibhav K. Khatavkar, 69
Sivasangari, A., 81 Varsha Choudhary, 317
Smita Agrawal, 481 Varun Sapra, 31
Subhajit Kar, 137 Venkatesh, P., 361
Sudeep Tanwar, 233, 405, 519, 731 Vidhi Pandya, 617
Suja C. Mana, 81 Vignesh, R., 81
Sukairaj Hafiz Imam, 563 Vipul Mishra, 701
Sumita Mishra, 451 Vishaka Ralegankar, 405
Sumit Dhariwal, 605
Sunil Kumar, 41
Sunita Soni, 107 Y
Swati Jindal, 591 Yashi Bansal, 31
Yashwant Singh, 317, 715, 747, 757
Yogesh Patel, 233
T
Tanveer Ahmed, 701
Tarana Singh, 217 Z
Teixeira, Marcio Andrey, 649 Zakir Ahmad Sheikh, 747