0% found this document useful (0 votes)
13 views81 pages

Driven-Circuits-And-Systems-Select-Proceedings-Of-Ised-2021-Lecture-Notes-In-Electrical - Engineering-811-37243994

The document provides information about various eBooks available for download on ebooknice.com, including titles related to artificial intelligence, business intelligence, and other engineering topics. It highlights the 'Artificial Intelligence Driven Circuits and Systems' book, which is a collection of proceedings from the ISED 2021 conference, edited by Biswajit Mishra, Jimson Mathew, and Priyadarsan Patra. The document also outlines the scope of the Lecture Notes in Electrical Engineering series, covering a wide range of electrical engineering topics.

Uploaded by

kleckaragula
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
13 views81 pages

Driven-Circuits-And-Systems-Select-Proceedings-Of-Ised-2021-Lecture-Notes-In-Electrical - Engineering-811-37243994

The document provides information about various eBooks available for download on ebooknice.com, including titles related to artificial intelligence, business intelligence, and other engineering topics. It highlights the 'Artificial Intelligence Driven Circuits and Systems' book, which is a collection of proceedings from the ISED 2021 conference, edited by Biswajit Mishra, Jimson Mathew, and Priyadarsan Patra. The document also outlines the scope of the Lecture Notes in Electrical Engineering series, covering a wide range of electrical engineering topics.

Uploaded by

kleckaragula
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 81

Download the Full Ebook and Access More Features - ebooknice.

com

(Ebook) Artificial Intelligence Driven Circuits


and Systems: Select Proceedings of ISED 2021
(Lecture Notes in Electrical Engineering, 811) by
Biswajit Mishra (editor), Jimson Mathew (editor),
Priyadarsan Patra (editor) ISBN 9789811669392,
https://ebooknice.com/product/artificial-intelligence-
9811669392
driven-circuits-and-systems-select-proceedings-of-
ised-2021-lecture-notes-in-electrical-
engineering-811-37243994

OR CLICK HERE

DOWLOAD EBOOK

Download more ebook instantly today at https://ebooknice.com


Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason;


Viles, James ISBN 9781459699816, 9781743365571,
9781925268492, 1459699815, 1743365578, 1925268497
https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Vagabond, Vol. 29 (29) by Inoue, Takehiko ISBN


9781421531489, 1421531488

https://ebooknice.com/product/vagabond-vol-29-29-37511002

ebooknice.com

(Ebook) Business Intelligence: 6th International


Conference, CBI 2021, Beni Mellal, Morocco, May 27–29,
2021, Proceedings by Mohamed Fakir, Mohamed Baslam, Rachid
El Ayachi ISBN 9783030765071, 3030765075
https://ebooknice.com/product/business-intelligence-6th-international-
conference-cbi-2021-beni-mellal-morocco-
may-2729-2021-proceedings-34612268
ebooknice.com

(Ebook) Boeing B-29 Superfortress ISBN 9780764302725,


0764302728

https://ebooknice.com/product/boeing-b-29-superfortress-1573658

ebooknice.com
(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena
Alfredsson, Hans Heikne, Sanna Bodemyr ISBN 9789127456600,
9127456609
https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

ebooknice.com

(Ebook) BSRIA Guide BG 29/2021: Pre-Commission Cleaning of


Pipework Systems by Martin Ronceray, Chris Parsloe ISBN
9780860227847, 0860227847
https://ebooknice.com/product/bsria-guide-bg-29-2021-pre-commission-
cleaning-of-pipework-systems-58260956

ebooknice.com

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT


II Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-
math-1c-and-2c-2002-peterson-s-sat-ii-success-1722018

ebooknice.com

(Ebook) 29, Single and Nigerian by Naijasinglegirl ISBN


9781310004216, 1310004218

https://ebooknice.com/product/29-single-and-nigerian-53599780

ebooknice.com

(Ebook) Jahrbuch für Geschichte: Band 29 ISBN


9783112622223, 3112622227

https://ebooknice.com/product/jahrbuch-fur-geschichte-band-29-50958290

ebooknice.com
Lecture Notes in Electrical Engineering 811

Biswajit Mishra
Jimson Mathew
Priyadarsan Patra Editors

Artificial
Intelligence
Driven Circuits
and Systems
Select Proceedings of ISED 2021
Lecture Notes in Electrical Engineering

Volume 811

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


Biswajit Mishra · Jimson Mathew ·
Priyadarsan Patra
Editors

Artificial Intelligence Driven


Circuits and Systems
Select Proceedings of ISED 2021
Editors
Biswajit Mishra Jimson Mathew
ULP-IC Lab Department of Computer Science
DAIICT and Engineering
Gandhinagar, Gujarat, India Indian Institute of Technology Patna
Patna, Bihar, India
Priyadarsan Patra
Vice-Chancellor’s Office and School
of Computing
DIT University
Dehradun, Uttarakhand, India

ISSN 1876-1100 ISSN 1876-1119 (electronic)


Lecture Notes in Electrical Engineering
ISBN 978-981-16-6939-2 ISBN 978-981-16-6940-8 (eBook)
https://doi.org/10.1007/978-981-16-6940-8

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 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
Contents

An Efficient and Affordable R-Pi Based Cardiac Disease Detection


System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Neha Arora and Biswajit Mishra
Performance Evaluation of IoT Enabled Pedometer for Estrus
Detection in Dairy Cows in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Yasha Mehta and Biswajit Mishra
Toward an Adversarial Model for Keystroke Authentication
in Embedded Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Niranjan Hegde and Sriram Sankaran
A Multi-phase LC-Ring-Based Voltage Controlled Oscillator . . . . . . . . . . 41
Sounak Das and Subhajit Sen
Efficient Quantum Implementation of Majority-Based Full Adder
Circuit Using Clifford+T-Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Laxmidhar Biswal, Bappaditya Mondal, Anindita Chakraborty,
and Hafizur Rahaman
Energy Efficient SRAM Design Using FinFETs and Potential
Alteration Topology Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Samarth Agarwal and Rajeevan Chandel
A Varactor-Based Impedance Tuner Circuit Operating at ISM
Frequency Band Centered at 2.45 GHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Chithra Liz Palson, Deepti Das Krishna, and Babita Roslind Jose
Deadlock Avoidance in Torus NoC Applying Controlled Move
via Wraparound Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Surajit Das and Chandan Karfa
Formal Modeling and Verification of Starvation Freedom in NoCs . . . . . 101
Surajit Das and Chandan Karfa

v
vi Contents

Reduced Graphene Oxide Soil Moisture Sensor with Improved


Stability and Testing on Vadose Zone Soils . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Kamlesh S. Patle, Salman siddiqui, Hemen K. Kalita,
and Vinay S. Palaparthy
NeuralDoc-Automating Code Translation Using Machine Learning . . . . 125
Sai Sree Harsha, Aditya Chandrashekhar Sohoni, and K. Chandrasekaran
Transfer Fault Detection in Finite State Machines Using Deep
Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Habibur Rahaman, Santanu Chattopadhyay, and Indranil Sengupta
Detecting Anomalies in Power Consumption of an Internet
of Things Network Using Statistical Techniques . . . . . . . . . . . . . . . . . . . . . . 153
Edwin Jose, Ajai John Chemmanam, Bijoy A. Jose, and Asif Mooppan

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165


About the Editors

Biswajit Mishra is a Professor and Head of ULP-IC Lab at Dhirubhai Ambani


Institute of Information and Communication Technology (DA-IICT), India. He
received his B.E. in Electronics and Communication Engineering from the National
Institute of Technology, Jaipur in 1996. Dr. Mishra received his M.S. and Ph.D.
degrees in Electronic Engineering from the University of Southampton, U.K., in
2003 and 2007, respectively. In 2007, he joined the University of Southampton as a
Research Fellow and from 2010 to 2013 he was a Senior Scientist at ESPLAB-EPFL-
Switzerland, working on ultra-low power and subthreshold design methodology
and energy harvesting electronic circuits. His research interests include battery-less
electronics, ultra-low-power circuits, sub-threshold design methodology, and system
implementation for BSN and WSN.

Jimson Mathew is an Associate Professor and Head of the Department in the


Computer Science and Engineering, Indian Institute of Technology Patna, India.
He did his Masters in computer engineering from Nanyang Technological Univer-
sity, Singapore, and a Ph.D. degree in computer engineering from the University of
Bristol, U.K. He has held positions with the Centre for Wireless Communications, the
National University of Singapore, Bell Laboratories Research Lucent Technologies
North Ryde, Australia, Royal Institute of Technology KTH, Stockholm, Sweden, and
Department of Computer Science, University of Bristol, UK. Dr. Mathew is a Senior
Member of IEEE and has also served as a Guest Editor for ACM TECS. His research
interests include fault-tolerant computing, computer vision, machine learning, and
IoT Systems.

Prof. (Dr) Priyadarsan Patra is the Pro Vice-Chancellor and Distinguished


Professor of DIT University, India.He earlier served as a Chief Architect and Prin-
cipal Scientist of the R&D divisions of Intel Corp (USA) and as the Dean of Computer
Sciences at UPES University and Dean of Research at Xavier University. Over the
years, he created and led world-class research and development for Servers, Systems-
on-Chip, and smart devices. His research interests include Computer and Systems
Architecture, Applications of Machine Learning, Big Data & Cloud Computing,

vii
viii About the Editors

and the Design of Low-power and Intelligent systems. Prof. Patra had developed an
architecture for building Single Flux Quantum Circuits and adiabatic circuits for his
dissertation.
Elected Senior Member of both the IEEE and the ACM and an IE Fellow, Prof.
Patra is the Founding Chair of the global IEEE System Validation and Debug Tech-
nology Committee and serves on several international Boards of Advisors. Dr. Patra
holds a Ph.D. degree in Computer Sciences from the University of Texas at Austin,
an M.S. from the University of Massachusetts at Amherst, and a B.Engg from the
Indian Institute of Science, Bangalore.
An Efficient and Affordable R-Pi Based
Cardiac Disease Detection System

Neha Arora and Biswajit Mishra

Abstract This paper proposes a Raspberry Pi (R-Pi) based system to automatically


detect and classify most of the atrial and ventricular cardiac diseases. The system
provides a necessary solution for resource constrained regions to timely detect fatal
cardiac conditions. The R-Pi receives the user specific information from the ECG
application specifically developed for the smart mobile phones and 2-Lead ECG
data from the ECG sensors connected with limb electrodes and the Arduino Nano
board. The ECG data along with the user specific information is further processed
for lead separation, ECG feature extraction, and disease classification. The obtained
results are sent to the doctor via email. Along with the system development, this work
proposes various algorithms for feature points detection and disease classifications.

Keywords ECG · R-Pi · Disease classification · Feature extraction

1 Introduction

ECG recordings are frequently used to detect Arrhythmias that are relatively quiet
in the early stages. However, they may provide valuable knowledge about an indi-
vidual’s fitness and aid in the detection of underlying heart anomalies and the timely
detection can prove to be life saving. While not all of arrhythmias are permanent
or necessitate medical treatment, they can signal the onset of serious heart diseases.
An ECG depicts the electrical activity of the heart and provides a large amount of
information on the functionality of the heart required for the proper diagnosis of

Supported by DST/SERB- CRG, Govt. of India, Research fund Ref: CRG/2019/004747.

N. Arora · B. Mishra (B)


Dhirubhai Ambani Institute of Information and Communication Technology (DA-IICT),
Gandhinagar, Gujarat, India
e-mail: biswajit_mishra@daiict.ac.in
N. Arora
e-mail: 201721007@daiict.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 1
B. Mishra et al. (eds.), Artificial Intelligence Driven Circuits and Systems,
Lecture Notes in Electrical Engineering 811,
https://doi.org/10.1007/978-981-16-6940-8_1
2 N. Arora and B. Mishra

Fig. 1 Typical features of an ECG signal

various diseases. As illustrated in Fig. 1, a typical ECG comprises segments, waves,


points, and QRS complex. These include P wave, Q point, R peak, S point, and T
wave.
Cardiovascular diseases are one of the major causes of death globally [1]. In India,
the average death rate due to cardiovascular diseases (CVDs) is significantly higher
than the global average [2]. Due to the lack of primary health care in India, it is
quite challenging to perform regular checkups in the hospital environments. It is
even more troublesome for rural areas, where the availability of clinics and medical
experts is comparatively low [3]. Hence, there is a significant need for automated,
low-cost physiological testing systems that are simple to use, reliable, and can be
used at home settings.
The development of a portable ECG monitoring system has been a topic of inves-
tigation over the past few years. In [4], a quasi real-time method was developed
for separating out ventricular ectopic beats based on QRS template matching and
R-R intervals assessment from both supraventricular and paced beats of the ECG.
However, in order to operate, they required human assistance. In [5], ECG signal
acquisition and classification of signals into three categories namely normal, Atrial
Fibrillation, and Myocardial Infarction were performed on an ARM processor using
wavelet analysis for feature extraction and support vector machine for the classifica-
tion. A device that measured an ECG signal using an e-health sensor shield attached
to an R-Pi was demonstrated in [6]. Further, MATLAB was used to analyze the heart
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 3

rate variability based on ECG data in order to obtain the values. In [7], ECG sensors
gathered patient physiological data, which was then analyzed by an Arduino micro-
controller. LabVIEW software was installed on the doctor’s phone, which accessed
the patient’s health details via email.
Our proposed system overcomes the dependency on offline processing by deploy-
ing the processing and disease detection algorithms on R-Pi. Equal to the size of a
credit card, R-Pi operates as if it were a regular machine at a lower cost without
any significant trade-off. Our system provides the results for most of the atrial and
ventricular cardiac diseases such as Arrhythmias, Atrial Fibrillation, Hypocalcemia,
Hypercalcemia, AV Blocks, and Myocardial infarction. Due to the Internet enabled
and compact size of the system, it is easily deployed even at remote locations.

2 Proposed System Architecture

The proposed system architecture is shown in Fig. 2. To make the system more flexible
to home environments, four limb reusable ECG clamp electrodes are used. The four
electrodes, namely, RA (right arm), LA (left arm), LL (left leg), RL (right leg) are
paired into two different leads with a pairing of RL, RA, and LA in Lead I and RL,
RA, and LL in Lead II. These two leads are fed into two different AD8232 sensors
through a 3.5 mm jack present on the board. The analog output from the sensors is
fed into Arduino nano as it has the inbuilt ADC to convert the Analog ECG data to
discrete sampled one and it is transmitted to R-Pi at a baud rate of 19200. Our system
operates at a sampling frequency 184 Hz.
Additionally, our system considers specific criteria to assess the individual’s pre-
liminary health status. These factors include the individual’s age, gender, and smok-
ing status through an Android Mobile Application.

2.1 Android Mobile Application for User Specific


Information

To develop an android mobile application, we have used the Message Queuing


Telemetry Transport (MQTT) protocol. It is a lightweight, publish-subscribe net-
work protocol that transports messages between different devices. The advantage of
using this protocol is bandwidth efficiency and power.
The central communication point in the MQTT is the broker (R-Pi), which is
in charge of dispatching all messages between senders (patients) and the rightful
receivers (doctors). To obtain the information, the first step is to run a python script
on the R-Pi to assign the status of R-Pi as a broker. To connect the android phone with
R-Pi, we need to enter the IP address of the R-Pi on the mobile phone. After forming
the connection between the two, the user specific details are filled and submitted. The
4 N. Arora and B. Mishra

Fig. 2 Proposed system architecture

details filled will be stored in R-Pi in the text format. Figure 3 shows the screenshots
of the developed android application and the proposed system architecture.
After receiving the user specific information the R-Pi starts storing the Lead I
and Lead II ECG signals. Lead I and Lead II signals are utilized to obtain remaining
frontal leads that are processed to classify various atrial and ventricular diseases.
The obtained results include the Heart Rate Variability analysis of ECG and the
summarized report on Arrhythmias, Atrial Fibrillation, PR anomalies, QT anomalies,
and ST-T deviations.

3 ECG Signal Processing Algorithm

After determining all the frontal leads, the leads are processed for R peaks, QRS
Complex, J point, baseline, P, and T waves of the ECG signal.
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 5

Fig. 3 Home pages of developed ECG application

3.1 QRS Complex, J Point and Baseline Detection

For detecting the R peaks, the widely accepted Pan Tompkins algorithm [8, 9] has
been used. The QRS complex has been detected by the difference operation method
stated in [10]. Detection of J point is done by creating a search space between t R +
20 ms to t R + 100 ms duration, where t R represents the location of the corresponding
R peak. Three consecutive points with a slope less than or equal to 2.5 µV/sec are
obtained and if the above-stated condition is satisfied, the midpoint is considered
as the J point. Otherwise, 60 ms is added to the S point to be the J point [11]. The
detection of the baseline of the signal is considered within a window spanning from
t R − 100 ms to t R − 40 ms. The window is further searched for 20 ms with an average
minimum slope value. The selected duration’s average amplitude is considered as
the baseline level of the signal. Search space for J point and Baseline is shown in
Fig. 4. Baseline detection plays an important role in MI detection.

3.2 Proposed T Wave and P Wave Detection Algorithms

The T wave may be biphasic or monophasic in nature in one or the other leads and
are necessary for detecting the cardiac anomalies such as MI, hypocalcemia, and
hypercalcemia. To obtain T waves that may be biphasic or monophasic in nature,
6 N. Arora and B. Mishra

Fig. 4 J point and baseline detection of an ECG signal

we propose an algorithm that makes use of QRS complex, J point, and baseline of
the signal as inputs with the ECG signal. This information is vital to create accurate
search spaces for the probable T waves. The algorithm begins with the local maxima
in the probable search spaces. During the course of experimentation, we observed that
the signal is mainly affected by the baseline wandering and subtraction of adaptive
baseline levels from the signal leads to improvement in results.
The ECG signal that is preprocessed with 0.5–10 Hz bandpass filter optimizes the
T wave energy. Search space of 200 ms duration is created starting with the Star tT
point. t J represents the J point location and RR(i) represents the corresponding RR
interval. To make the search space adaptive of baseline changes and other drifts
present in the signal, the baseline level has been subtracted from the search space.
Further, we have determined maximum and minimum values in the adaptive search
space with the respective locations. The proposed algorithm for T wave detection
is shown in Fig. 5. Following this, the absolute maximum and minimum values are
compared to obtain the biphasic or monophasic T wave location.
The detection algorithm of the P wave consists of preprocessing band pass filter
to optimize the P wave frequencies and a stage to fix the QRS amplitude levels. It is
done so that the QRS signal does not interfere in the detection of the P wave. After
the preprocessing stage, we categorized the RR segments based on the Premature
Ventricular Contractions (PVC) conditions. If the signal belongs to the PVC class
then, the P wave detection for that particular case is not possible. Otherwise, a
temporal search space for P wave as mentioned in [12] is considered. However, the
mentioned literature utilized the phasor transform of the ECG signal, whereas we
are processing the signal in the time domain.
The maximum’s location in the search space is further compared with the previous
cycle’s T wave location. If the T wave location (tT ) and the P wave location (t P ) is
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 7

Fig. 5 Proposed T wave detection algorithm

same for more than 40% (value is chosen based on our observation during the exper-
iments) times of the total length of T waves, it is assumed that the probable location
for P waves is coinciding with the T wave locations. Therefore, the search space for P
wave is shifted and the maximum amplitude in the selected duration is assumed to be
the probable P wave location. Further, P wave amplitude ( Amplitude P ) is compared
with 1% value of R peak amplitude ( Amplitude R ); if the condition satisfies, the P
wave peak is considered as the valid P peak. Finally, t P is searched for the local
maximum in the 40 ms region for obtaining the final P wave location. The proposed
algorithm for P wave detection is shown in Fig. 6.

3.3 Disease Classification

After determining all the feature points such as P, Q, R, S, and T waves, various
features such as R-R interval, ST-T segment, QT interval, PR interval, P wave, and
R-R features for cardiac diseases have been detected. Based on the R-R intervals of the
Lead II signal, various arrhythmic conditions, namely Tachycardia, Bradycardia, R on
8 N. Arora and B. Mishra

Fig. 6 Proposed P wave detection algorithm

T Condition, Asystole, Bigeminy, Trigeminy, Ventricular condition, and Interpolated


premature ventricular conditions can be detected [13].
Similarly, for detecting and localizing four types of Myocardial Infarctions, viz.,
Anterior, Inferior, Inferolateral, and lateral cases an algorithm based on 6 Lead ECG
signal (Frontal Leads) are discussed in [14].
QT interval is the time starting from the QRS complex to the end of the T wave.
It represents the total time for activation of the ventricles and recovery to the normal
resting state [15]. To determine the To f f point, a search space of 80 ms duration
has been created that starts from T peak + 40 ms. For the given interval, we have
determined the 16 ms window with the minimum slope value by using a moving
window method and the starting point of that duration is considered as To f f point.
QT interval is dependent upon the RR interval of the person. Hence, the corrected QT
interval is used as a feature instead of directly using the QT duration. The corrected
QT duration is given by Bazett’s formula [15] as follows:
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 9

QT
QTc = √ (1)
RR

The upper limit of QTc duration is 460 ms. Similarly, the lower limit 390 ms
[15]. If the values of the QTc interval are beyond the threshold values, then it is
considered as prolongation of QT duration (Hypocalcemia) or shortening of QT
duration (Hypercalcemia).
For the detection of Atrial Fibrillation, P wave and R-R features are utilized. For
1 min of data, we have considered the Interquartile range (IQR) of R-R intervals
as the variability measure. The missed detection of P waves for large number of
cycles also signifies the Atrial Fibrillation condition. If the IQR is above 50 sample
points (@360 Hz Sampling frequency) or in the case of missing P waves, it has been
considered a probable case of Atrial Fibrillation.
The PR interval measurement is crucial for the detection of AV node anomalies.
Ideally, the PR interval must exist between the range of 120 ms to 200 ms. The PR
interval has been calculated for Lead II of standard 12 Lead ECG. If the values do
not exist in the specified range, various AV node conduction anomalies such as AV
block, AV nodal rhythm may exist.

4 Results and Discussion

For the validation of proposed algorithms, we have utilized various standard databases
available on Physionet [16]. For example, for the detection of QRS complexes and
arrhythmias, we have used MIT-BIH Arrhythmia Database (MITDB) [17]. Simi-
larly, other databases such as PTB database (PTB) [18], QT database (QTDB) [19],
MIT-BIH database with P wave annotations (PWAVE) have been utilized for specific
features as shown in Fig. 2. Table 1 shows the QRS detection evaluation results at
MIT-BIH arrhythmia database. The efficiency metrics are typically defined in terms
of True Positives (TP), False Positives (FP), False Negatives (FN), and True Negative
(TN) values. Based on these values, we have measured the Sensitivity (Se), Speci-
ficity (Sp), False Detection Rate (FDR), Positive Predictivity (PPV), etc. to validate
our proposed algorithms.
Based on the detected R peaks, R-R interval has been utilized to detect various
arrhythmias. A detailed discussion regarding the same can be found in our previ-
ous work [13]. Table 2 shows the present and detected arrhythmias on MIT-BIH
arrhythmia database.

Table 1 Overall QRS detection evaluation


Total annotated Beats detected Average PPV% Average Se% Average FDR%
beats
109864 109391 99.29 99.49 1.29
10 N. Arora and B. Mishra

Table 2 Results for arrhythmias detection


Record Arrhythmia present Arrhythmia detected
100 PVC PVC, Interpolated PVC
101 APC PVC
102 PVC PVC
106 Tachycardia, PVC, Tachycardia, PVC,
Bigeminy Trigeminy
108 PVC, Interpolated PVC PVC, Interpolated PVC
114 PVC, Ventricular Couplets PVC, Asystole, Bradycardia,
Trigeminy
119 Trigeminy, Bigeminy Trigeminy, PVC, Interpolated
PVC
200 Tachycardia, Bigeminy, PVC, Trigeminy,
PVC, APC Tachycardia
201 Trigeminy Trigeminy
203 Trigeminy, Tachycardia, Tachycardia, Trigeminy, PVC,
Multiform PVC Interpolated PVC
207 Bigeminy, Tachycardia Tachycardia, Trigeminy,
PVC, Interpolated PVC
214 Trigeminy, PVC, Tachycardia Trigeminy, Tachycardia,
PVC, Interpolated PVC
221 Tachycardia, PVC Tachycardia, PVC, Trigeminy
222 Bigeminy Tachycardia, PVC
223 Trigeminy, Tachycardia, Tachycardia, Trigeminy,
Bigeminy, PVC PVC, Interpolated PVC
232 Bradycardia, Asystole Bradycardia, Asystole, PVC
234 Tachycardia, PVC Tachycardia, PVC

Further, T wave detection results for the proposed algorithm have been validated
on QT database and are shown in Table 3. The database consists of 15 MITDB signals
and provides the annotations for the normal beats. Table 3 shows the evaluation of the
T wave detection algorithm, where the overall sensitivity of the proposed algorithm
is found out to be 97.78%.
Similarly, the P wave detection utilizes the MIT-BIH arrhythmia database with
P wave annotations. The database consists of 12 databases from MITDB that are
subcategorized to 5 normal and 7 abnormal waveforms consisting of various pathol-
ogy conditions. Records #100, #101, #103, #117, and #122 belong to the normal
category and Records #106, #119, #207, #214, #222, #223, and #231 belong to the
abnormal categories. Mean Se, PPV, and FDR for normal waveforms are found out
to be 99.39%, 99.42%, and 1.13%, respectively. Similarly, for the abnormal wave-
forms, the respective values are 94.23%, 82.74%, and 25.6% so, we conclude that
the abnormal pathologies of waveforms lead to missed detection of P waves. Table 4
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 11

Table 3 T wave detection results for QT database for MITDB signals


Record Annotated beats TP FN Se%
sel100 30 30 0 100
sel102 85 84 1 98.82
sel103 30 30 0 100
sel104 77 74 3 96.10
sel114 50 50 0 100
sel116 50 48 2 96
sel117 30 30 0 100
sel123 30 30 0 100
sel213 71 69 2 97.10
sel221 30 26 4 86.6
sel223 30 30 0 100
sel230 50 50 0 100
sel231 50 47 3 94
sel232 30 30 0 100
sel233 30 30 0 100
Total 673 658 15 97.78

Table 4 P wave detection results for MIT-BIH Arrhythmia database with P wave annotations
Record TP FP FN Se% PPV% FDR%
100 2245 23 9 99.60 98.99 1.4
101 1835 15 27 98.55 99.19 2.2
103 2065 15 16 99.23 99.23 1.4
117 1527 3 4 99.74 99.80 0.45
122 2469 2 4 99.84 99.92 0.24
Mean 99.39 99.42 1.13
106 1491 528 13 99.14 73.85 35.9
119 1566 455 52 96.79 77.49 31.3
214 1765 382 234 88.29 82.21 30.8
222 1118 265 135 89.22 80.83 31.9
223 2048 418 48 97.71 83.05 22.2
231 1878 18 114 94.27 99.05 1.6
mean 94.23 82.74 25.6

shows the P wave detection results for the normal and abnormal signals present in
the PWAVE database. Based on the detected P, Q, R, S, and T feature points, we have
also detected the PR duration and QT duration of the signal that can be used to detect
various cardiac conditions such as Hypocalcemia, Hypercalcemia, and AV Blocks.
12 N. Arora and B. Mishra

Table 5 Interquartile range for atrial fibrillation detection for MITDB 1 min dataset
Record no. IQR (R-R) Record no. IQR (R-R)
Normal cases 100 27 Atrial 201 79
fibrillation
cases
101 33.25 202 90.75
103 37.75 203 279.25
117 35.5 210 101.50
122 23.25 217 50
219 149
221 155.75
222 100.50

Table 6 Performance evaluation of proposed MI detection system with and without user specific
information (USI)
USI TP FP TN FN PPV% Se% Sp% Acc%
 123 7 45 23 94.6 84.2 86.5 85
X 123 22 30 23 84.8 84.2 57.6 77.27

In addition to the formerly mentioned diseases, Lead II data can also be used to
detect the Atrial Fibrillation condition by combining the P wave and R-R features.
Irregular R-R interval and missed P wave signify atrial fibrillation. For a smaller
set of ECG data (1 min of data), interquartile range (IQR) provides better variations
approximation compared to the standard deviations. We have also observed that in
the case of atrial fibrillation, R-R interval distribution is skewed instead of Gaussian
and IQR provides the range of central 50% of data and removes the outliers. Table 5
shows the IQR for normal cases and atrial fibrillation cases for 1 min of duration.
Along with the above-mentioned diseases, we have utilized 2-Lead ECG data
to detect the Myocardial Infarction condition. A detailed description of Myocardial
Infarction detection algorithm can be found in [14]. To validate the proposed algo-
rithm, we have utilized the PTB Database. The database provides the 15 lead signals,
but we have utilized the lead I and lead II signals to validate the Myocardial Infarction
detection algorithm. We have utilized certain person specific parameters such as age,
gender, and smoking status to validate our algorithm along with the 2-Lead ECG
signals. It adopts and separates Myocardial Infarcted signals out of Normal Sinus
signals and the efficiency metrics for the proposed system with and without using
the user specific information (USI) are shown in Table 6.
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 13

Fig. 7 Real-time P, QRS, and T detection; output sample report generated by the system and to be
mailed to the doctor

4.1 Real-Time P, QRS, and T Point Detection

During real time, the system takes 60 s of data from the two leads as the input
at a sampling frequency 184 Hz. However, for proper visualization of the detected
points, a smaller dataset of 1250 samples, i.e., 6–7 s is used. Detected ECG outputs
and generated output report are shown in Fig. 7. Along with this, our designed system
is capable of sending the generated report to the doctor via email.

5 Conclusion and Future Work

Along with the system design, we quantified various feature points (P, QRS, and T)
of the ECG signal and proposed various algorithms for ECG signal processing. The
proposed algorithms are validated on the standard databases. QRS complex detection
provides PPV, Se of 99.29% and 99.49% respectively. The proposed T wave detection
approach based on local search spaces provides the detection Se of 97.78% for the
manually annotated MITDB waveforms from the QTDB database. The proposed
P wave detection scheme obtains the PPV, Se, and FDR of 99.42%, 99.39%, and
1.13%, respectively, for the normal cases. Similarly, for the abnormal cases, the
values are 82.74%, 94.23%, and 25.6%. Myocardial Infarction detection is done on
14 N. Arora and B. Mishra

the available PTB database with PPV, Se, Sp, and Acc of 94.6%, 84.2%, 86.5%,
and 85%, respectively. Our system provides efficiency of 100% in detecting Atrial
Fibrillation. On the set of databases that we have considered, the effectiveness of both
hardware and software, affordability and user friendly approach make this system
an effective candidate for remote monitoring in resource constraint regions. The
system’s validation for cardiac patients is left as future work.

Acknowledgements We would like to acknowledge the contribution from Mr. Tejas Vasava and Mr.
Smit Bhagat for the development of android mobile application. We would also like to acknowledge
the contribution from Mr. Swaresh Phadke who helped us in porting the ECG signal processing
algorithms to Raspberry Pi.

References

1. World Health Organization, Cardiovascular Diseases, World Health Organization Cardio-


vascular Diseases Fact Sheet available at: https://www.who.int/news-room/fact-sheets/detail/
cardiovascular-diseases-(cvds) Last accessed 10 May (2021)
2. D. Prabhakaran, P. Jeemon, A. Roy, Cardiovascular diseases in India: current epidemiology
and future directions. Circulation 133(16), 1605–1620 (2016)
3. I. Silva, G.B. Moody, L. Celi, Improving the quality of ECGs collected using mobile phones:
The Physionet/Computing in Cardiology Challenge 2011, in 2011 Computing in Cardiology
(IEEE, 2011), pp. 273–276
4. V. Krasteva, I. Jekova, QRS template matching for recognition of ventricular ectopic beats.
Ann. Biomed. Eng. 35(12), 2065–2076 (2007)
5. T. Jeon, B. Kim, M. Jeon, B.G. Lee, Implementation of a portable device for real-time ECG
signal analysis. Biom. Eng. Online 13(1), 1–13 (2014)
6. Ö. Yakut, S. Solak, E.M.İ.N.E. Bolat, Measuring ECG signal using e-health sensor platform
(2014)
7. A. Abdullah, A. Ismael, A. Rashid, A. Abou-ElNour, M. Tarique, Real time wireless health
monitoring application using mobile devices. Int. J. Comput. Netw. Commun. (IJCNC) 7(3),
13–30 (2015)
8. J. Pan, W.J. Tompkins, A real-time QRS detection algorithm. IEEE Trans. Biomed. Eng. 3,
230–236 (1985)
9. W.J. Tompkins, Biomedical Digital Signal, Processing: C-Language Examples and Laboratory
Experiments for the IBM PC., Englewood Cliffs (PTR Prentice Hall, NJ, 1993)
10. Y.C. Yeh, W.J. Wang, QRS complexes detection for ECG signal: the difference operation
method. Comput. Methods Programs in Biomed. 91(3), 245–254 (2008)
11. S. Ansari, N. Farzaneh, M. Duda, K. Horan, H.B. Andersson, Z.D. Goldberger, B.K. Nal-
lamothu, K. Najarian, A review of automated methods for detection of myocardial ischemia
and infarction using electrocardiogram and electronic health records. IEEE Rev. Biomed. Eng.
10, 264–298 (2017)
12. L. Maršánová, A. Němcová, R. Smíšek, M. Vítek, L. Smital, Advanced P wave detection in
Ecg signals during pathology: evaluation in different arrhythmia contexts. Sci. Rep. 9(1), 1–11
(2019)
13. B. Mishra, N. Arora, Y. Vora, Wearable ECG for real time complex P-QRS-T detection and
classification of various arrhythmias, in 2019 11th International Conference on Communication
Systems & Networks (COMSNETS) (IEEE, 2019), pp. 870–875
14. N. Arora, B. Mishra, Characterization of a low cost, automated and field deployable 2-lead
myocardial infarction detection system, in 2020 International Conference on Communication
Systems & NETworkS (COMSNETS) (IEEE, 2020), pp. 41–46
An Efficient and Affordable R-Pi Based Cardiac Disease Detection System 15

15. A. Houghton, D. Gray, Making Sense of the ECG: A Hands-on Guide (CRC Press, 2014)
16. A.L. Goldberger, L.A. Amaral, L. Glass, J.M. Hausdorff, P.C. Ivanov, R.G. Mark, J.E. Mietus,
G.B. Moody, C.K. Peng, H.E. Stanley, PhysioBank, PhysioToolkit, and PhysioNet: components
of a new research resource for complex physiologic signals. Circulation 101(23), e215–e220
(2000)
17. G.B. Moody, R.G. Mark, The impact of the MIT-BIH arrhythmia database. IEEE Eng. Med.
Biol. Mag. 20(3), 45–50 (2001)
18. R. Bousseljot, D. Kreiseler, A. Schnabel, Nutzung der EKG-Signaldatenbank CARDIODAT
der PTB über das Internet. Biomedizinische Technik/Biomed. Eng. 40(s1), 317–318 (1995)
19. P. Laguna, R.G. Mark, A. Goldberg, G.B. Moody, A database for evaluation of algorithms for
measurement of QT and other waveform intervals in the ECG, in Computers in Cardiology
1997 (IEEE, 1997), pp. 673–676
Performance Evaluation of IoT Enabled
Pedometer for Estrus Detection in Dairy
Cows in India

Yasha Mehta and Biswajit Mishra

Abstract Insufficient herd fertility due to poor estrus detection significantly affects
the dairy cattle sector. In India, the conventional way of detection relies heavily on
herder’s visual surveillance, which makes the quality of estrus detection highly sub-
jective to herder’s experience in identifying estrus signs. This method lacks automa-
tion and may not be effective in monitoring larger herds. Current devices used for
cattle monitoring are imported, limited in services, and are often expensive. Con-
tinuous monitoring also suffers from poor battery life and is a deterrent for wide
adoption. A low-cost system with long battery life and real-time monitoring could
provide tremendous benefits to the Indian dairy industry. In this context, this paper
presents our efforts to develop an IoT-based estrus monitoring and detection system
that utilizes ICT technique to provide an effective map of the cattle reproduction
phase and identify underlying health issues to determine the wellness of the ani-
mal specifically for Indian conditions. The system is developed to adapt to various
deployment challenges such as frequent topological changes due to cattle’s mobility,
limited power source, and short transmission range of sensors. In this context, the
performance of the proposed system has been evaluated through laboratory trials.

Keywords Estrus detection · Internet-of-things · Dairy · Activity monitoring ·


Pedometers

1 Introduction

Dairy industry is an essential part of the Indian economy. According to the latest
research, the market size of dairy products in India was approximately INR 11,360
billion in 2020, with the milk being the major source of income [1]. Despite being

Y. Mehta (B) · B. Mishra


VLSI and Embedded Systems Research Group, DA-IICT, Gandhinagar 382007, India
e-mail: yasha_mehta@daiict.ac.in
B. Mishra
e-mail: biswajit_mishra@daiict.ac.in

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 17
B. Mishra et al. (eds.), Artificial Intelligence Driven Circuits and Systems,
Lecture Notes in Electrical Engineering 811,
https://doi.org/10.1007/978-981-16-6940-8_2
18 Y. Mehta and B. Mishra

the largest producer, India exports relatively smaller volumes of dairy products and
has managed to capture only a small share in the global dairy trade because, high
milk production is due to larger bovine population [1]. Dairy farming forms an essen-
tial part of rural India, but milk production per animal is relatively low. Therefore,
to increase profitability and satisfy the growing demands, it becomes necessary to
increase the milk production of dairy cattle.
Insufficient cattle fertility poses a major threat that limits the reproductive effi-
ciency of eligible cattle. It is due to poor detection of estrus [2] that negatively impacts
milk production and affects fertility rates. This results in loss of livelihood of Indian
farmers, leading to major economic losses. Therefore, effective detection of estrus
is of prime importance as herd fertility, milk production, and farm profitability are
strongly correlated.
During estrus, cattle show various symptoms like mucus discharge, sniffing, chin
resting, mounting, restlessness, and increased movement. Various methods to detect
estrus, reported in literature, require manual intervention and often cause discomfort
to the animal. Owing to technological advancement, automated solutions for estrus
detection are reported [3], which processes behavioral responses of the animals using
state-of-the-art classifiers and machine learning algorithms. With the cattle showing
various behavioral signs during estrus, high activity is considered as a prominent
behavior and an effective indicator of estrus. Companies like Afimilk, DeLavel [4],
Rumiwatch [5], ALT pedometer [6] have developed pedometer-based automated
estrus detection systems that have become an essential part of herd management
solutions. In India, the devices that are used are imported for large farms and are
often expensive which are not affordable by small or medium farms. There have
been a few experiments on Indian-specific studies [7] but they lacked automated
data collection, and the device required frequent battery recharging. Therefore, in
India, there is still no concrete solution available that addresses the concern of small-
or medium-scale Indian farmers.
To facilitate Indian dairy farmers with efficient herd management tool, we propose
precision livestock farming approaches combined with ICT to monitor and process
activity behavior of cattle. The goal of this project is to develop a low-energy auto-
mated activity monitoring system for estrus detection that would help small- and
medium-sized Indian farmers. In addition to providing remote real-time assessment
of animal’s behavior, the solution also offers easier integration of the existing dairy
database [8] forming a part of a larger information network that would help to analyze
regional trends. To the best of author’s knowledge, this is the first complete solu-
tion developed indigenously and specifically for Indian farmers severely impaired
by resource constraints.
Performance Evaluation of IoT Enabled Pedometer … 19

2 Activity Monitoring System

2.1 General Description

Figure 1 shows the basic configuration of the proposed monitoring system. The major
elements are a battery-powered pedometer sensor, sink nodes that can either be wall-
powered or battery-powered and a cloud server. The pedometer is enclosed in a
small package mounted on the animal leg to acquire activity information or steps
taken. This recorded information is stored in the sensor buffer and is transmitted
periodically to the sink nodes using a 2.4 GHz wireless link. The receivers or sink
nodes are deployed for data aggregation at the server that contains algorithms and
thresholds to interpret the data associated with animal_id. The results of the forecast
are displayed on a front-end display to enable farmers and veterinarians to understand
the status of the animal and help them to act on the next course.

2.2 Pedometer Design and Operation

The block diagram of pedometer shown in Fig. 1 is a custom hardware built that uti-
lizes a micro-electro-mechanical system-based ultra-low power inertial measurement
unit (IMU). To record the steps and activity of the animal, the system is initialized to
step detection mode. A determined wake-up and sleep period is employed for energy
conservation such that data transmission takes place only during active mode of the
sensor and remains in power-saving mode for the rest of the time. Timers are used to

Fig. 1 Activity monitoring system


20 Y. Mehta and B. Mishra

Fig. 2 Pedometer sensor operation

schedule the duty cycle of the pedometer. In order to provide a reference time for the
timers to operate, when a pedometer tag is switched on as shown in Fig. 2, it remains
in the wake state to receive the reference timestamp broadcasted by the synchronizer
board. Upon its reception, the pedometer synchronizes its internal clock and start
their transmission cycle. The pedometer tags calculate their sleep period according
to their defined schedules and undergo a deep sleep mode for a calculated duration of
time. Upon time-up, an interrupt routine is executed to wake the module for uplink.
Figure 2 shows the complete operation of the pedometer tag.
In our developed system, the movement data is sampled at a rate 25 Hz and a step
is detected only if more than 3 s of activity is observed to avoid misinterpretation
of flexes with a step count. The on-board memory stores the acquired information
during sleep periods and upon wake-up, the discrete data containing step counts and
animal identification number(animal_id) is broadcasted. Since the receiver node is
always awake, the wireless connection between the pedometer tags and the receiver
(sink node) is established, and the data is successfully offloaded to the receiver.

2.3 Receiver Node Design and Operation

The receiver acts as a bridge between the pedometer tags/networks and the cloud
server. It consists of a six-channel transceiver module operating over a frequency
range of 2.400 GHz to 2.525 GHz with a separation of 1 MHz [9]. Each channel is
designed to receive multiple pedometer’s data simultaneously. They play a major role
in expanding the system’s capacity that is detailed in Sect. 3. The received data is
uploaded to the server using a GSM module interfaced to the microcontroller using
UART protocol. The communication between the GSM module and the database
server utilizes HTTP protocol.
Performance Evaluation of IoT Enabled Pedometer … 21

2.4 Cloud-Based IoT Platform

The cloud-based software environment utilizes the Microservices architecture as


shown in Fig. 3 to classify the data and provide insights about the behavioral patterns
of the animals. The microservice architecture is developed with the help of a Docker
container. Each process is isolated into a different service that doesn’t rely upon each
other and has different functionality. They communicate with each other through
the help of the REST API (Application Programming Interface) and GraphQL API,
which both leverage the HTTP protocol. The three microservices in this architecture
forming the back-end system are Hasura GraphQL Engine, PostgreSQL Database,
and the back-end server based on Flask(Python). These services work in tandem to
process the received raw data and it contains algorithms that records and processes the
clean step entries for estrus detection and lameness detection. These microservices
also communicate with the front-end UI based on NextJS which enables the server-
side rendering of React Applications through the help of REST API and GraphQL
API. Through the front-end UI, the processed data is displayed in the form of graphs,
texts and important events (estrus status and health status) to be easily understandable
by the veterinarians and farmers. Details of this architecture are discussed in [11].

Fig. 3 Software architecture [10]


22 Y. Mehta and B. Mishra

3 Performance Evaluation

It is important to ensure that the system functions reliably in monitoring the cattle’s
health along with estrus and present them remotely to the veterinarians and farmers
at any time. Therefore, the functionality of each and every component of the system
is to be evaluated and specific challenges owing to the implementation of sensor
networks are to be addressed. The important parameters that need to be tested include
Data Communication, Network Lifetime, Real-time assessment for identification of
aberrant behavior and Scalability for large-scale deployment.

3.1 Data Communication

Network connectivity is important for data communication, but the problems such
as cattle’s mobility, changes in network topology, and radio interference from other
animals significantly affect the system’s performance. Therefore considering cattle’s
mobility, a one-hop star topology is employed at the network layer. Secondly, it may
become possible that the radio transceivers at the sink node have a limited range of
frequencies available to serve all the pedometers present within its vicinity. Thus,
owing to the non-uniform distribution of the cattle, time division multiple access
(TDMA) based channel access strategy is adopted such that each pedometer will
have its own defined time-slot, which will alleviate data overlaps at the sink node.
To test for data communication, we set up six pedometers and three sink nodes in
the laboratory. The pedometers were powered by a 3.7V LiPo battery and sink nodes
were powered by wall socket plug-in power supply. The pedometers were placed
randomly with one of the pedometer tied to a human leg for recording activity data,
and the sink nodes were deployed 3 m apart such that at least one sink node was always
present within the vicinity of the pedometer. The pedometers were programmed to
have a duty cycle period, D = 30 min. The elapsed time ‘D’ is programmable and is
a useful measure for system’s scalability explained in Sect. 3.4.
At server’s side, each time a new data is inserted, activity graph changes, and
results of processing are displayed on the front-end in tandem. The graph in Fig. 4
maps the step data (analogous to the walking activity of the cow) recorded during
a 7-day period. Based upon the results shown in Fig. 4, it is ensured that all the
components of the monitoring system are working in accordance with the designed
pathway of both hardware, software, and the network.
Performance Evaluation of IoT Enabled Pedometer … 23

Fig. 4 Activity graphs at front-end

3.2 Network Lifetime

Network lifetime is an important attribute of any sensor networks and solely depends
upon the sensor node’s battery life. The life span of the sensor node relies on the
current consumption of its components and is given by (1):

Ccapacit y
Li f etime = , (1)
Iavg

where Ccapacit y is the battery capacity of the pedometer tag given in mAH and the
average current, Iavg depends upon the power-saving strategies adopted. Since duty
cycle based power-saving strategy is adopted, average current consumption can be
calculated using (2).

dactive ∗ Iactive + dsleep ∗ Isleep


Iavg = , where, D = dactive + dsleep . (2)
D
Here, Iactive is the average current consumed for dactive seconds during active mode
of the tag and Isleep is the average current consumed for dsleep seconds during the
sleep mode.
Initially, the pedometer remains in the receiving mode; hence the current con-
sumed during the receiving cycle is measured to be approximately 18.6 mA as shown
by the highest current spike in Fig. 5. The periodic spikes in Fig. 5 shows the cur-
rent consumed during data transmission which is measured to be around 4.8 mA.
In between subsequent transmissions, the module remains in a deep sleep mode to
conserve energy thus, the sleep mode current consumption is measured to be around
120 µA.
In reference to Fig. 5, after synchronization the module continues to function in
the transmission phase; therefore the pedometer’s lifetime majorly depends on the
average current consumed during the transmission cycle. Thus Iactive = 4.8 mA and
24 Y. Mehta and B. Mishra

Fig. 5 Current consumption profile of pedometer sensor

Isleep = 120 µA. Now, with the duty cycle period D = 30 min, the wake-up period,
dactive is observed to be 4 s and sleep period, dsleep is observed to be 1796 s. So,
the average current calculated using (2) is around 0.1304 mA per duty cycle. The
pedometer tags are powered by 3.7V/1500 mAh LiPo battery, thus the lifetime of the
pedometer tag is calculated to be around 2.63 years and based on lab experiments, it
is observed to be approximately 2.84 years.

3.3 Real-Time Assessment for Identification of Aberrant


Behavior

In order to determine the performance of the system in identifying estrus or possible


ailment, actual activity values of the animal are compared to its baseline behavior
in a particular time frame. The baseline behavior is the threshold values obtained by
computing the average of the step counts plus twice the standard deviation within
the similar time-interval of the preceding 10 days. Voting algorithm is used for the
detection such that if the number of steps exceeds its respective threshold values for
a defined number of times in a 3 h period, estrus is concluded while the lameness
is detected if reduction of 5% or more is observed in daily activity when compared
to the threshold values in a 5 h period. The daily activity mapped in ‘dark’ traces is
compared against the ‘gray’ traces of baseline activity of the same animal to analyze
the behavioral responses. The beginning of estrus is indicated with the increased
Performance Evaluation of IoT Enabled Pedometer … 25

Fig. 6 Activity analysis of individual Animal—Estrus

Fig. 7 Activity analysis of individual Animal—Lameness

Fig. 8 Activity comparison against average herd behavior

activity as illustrated in Fig. 6 while the lameness is indicated with decreased activity
as illustrated in Fig. 7. In addition, average herd behavior is also recorded which is
used to compare an individual animal’s behavior against the behavior of the herd it
belongs, to determine its performance. Figure 8 shows particular animal’s activity in
‘dark’ bars and average herd behavior in ‘gray’ bars.
26 Y. Mehta and B. Mishra

It is to be noted from Figs. 6 and 7 that the activity analysis of two different
animals is from two different farms, i.e., Farm 1 and Farm 3. This verifies the system’s
feature which helps the farmer in managing all his farms remotely from one single
point. Also, a facility to download the recorded information, in the form of a pdf
file is provided, which contains details of the animal of interest and activity trends
associated with them. This enables the farmer to share the recorded data to the
veterinarians to determine well-being of the animal or to the buyers to know the
history of the animal before a purchase.

3.4 System Scalability

Reliability and scalability to large-scale deployment are of prime importance for


large farms accommodating hundreds of cows. The sink nodes incorporated with
multi-channel transceiver module play a major role in expanding system’s capacity by
exploiting all the available data pipes for data reception. The transceivers have six data
channels [9] and are capable of receiving ‘n’ pedometers transmission simultaneously
per time-slot where, 1≤n≤6. Experiments were conducted with the six pipes enabled
for reception such that n = 6. Each pedometer had its unique pipe address for the
time-slot assigned to it as illustrated in Fig. 9, so it is certain that a packet will be
successfully received on each transmission. For example, out of 1000 packets, 776
packets were received correctly. Thus the sink node’s throughput efficiency was
observed to be around 77.6%.
It should be noted that the transceiver operates over a frequency range from 2.400
to 2.525 GHz. With the separation of 1 MHz, 125 frequency channels are available for
communication [9]. By utilizing different communication frequency, Fs for different
sections ‘s’, we can increase the capacity of the system for larger farms having
tens of hundreds of cows. For implementing this, it was important to check for data
interference between the adjacent frequency channels. To verify this, we set up one set
of sink node and two pedometers communicating at F1 = 2.440 GHz and another

Fig. 9 Demonstration of multi-channel reception


Performance Evaluation of IoT Enabled Pedometer … 27

Fig. 10 Setup to check for


data interference

at F2 = 2.442 GHz as shown in Fig. 10. The performance of a complete system


was examined for a month. No data interference was observed between adjacent
communication frequencies. Thus, it can be concluded that the system capacity to
accommodate number of cows denoted by ‘C’ depends upon the number of data
pipes (n), duty cycle period (D), and number of frequency channels (F) used and is
given as


125
C = n ∗ D ∗ F, where, F = Fs . (3)
s=1

Suppose, if n = 6, D = 30 min and s = 2, system can accommodate 360 cows.


Similarly if n = 6, D = 60 min and s = 125, the theoretical limit that the system
can accommodate is 45000 cows. Thus, it is ensured that the system can support
monitoring tens of hundreds of cows within one farm. With frequency reuse, more
number of cows is possible to be accommodated and is left as a future work.

4 Conclusion

In India, there are few farms/organizations that have adopted modern herd monitor-
ing technologies. The available devices are imported and often expensive which are
not affordable by small- or medium-scale farmers. Since the opportunity to identify
a cow in estrus is relatively short, lasting a few hours in the 21-day cycle, with the
cows showing fewer signs of estrus and for a shorter duration, it becomes difficult
for farmers to identify estrus by only visual observations. Therefore, this research
aims to develop an activity-based estrus detection system with ultra-low energy con-
sumption to be advantageous for small- or medium-sized Indian farmers in resource
constrained regions. The IoT-based software infrastructure will facilitate the farmers
with remote real-time tracking of the cattle and identify estrus triggers; giving them
a chance to prevent missing cycles. In addition, performance of the individual cattle
can be tracked and compared to the herd it belongs such that deviation from normal
will enable early diagnosis of underlying health issues. The research also facilitates
28 Y. Mehta and B. Mishra

data integration with existing regional or national databases to form a part of larger
information network. To make this system more usable, the proposed system has the
potential to be extended for other applications such as lameness detection or stand-
ing time analysis, rumination monitoring and body temperature measurements that
could be useful for dairy farmers.

Acknowledgements This work is a part of project (19/STDST/BSM/DFTEA) funded by Depart-


ment of Science and Technology under S&T scheme. The authors acknowledge the efforts of interns
Mohammed Shadab and Vedant Thakkar in developing the cloud platform for data analysis and
visualization.

References

1. IMARC Group: Dairy Industry in India, Edition: Market Size (Procurement and Distribution,
Growth, Prizes, Segments, Cooperatives, Private Dairies, 2021), p. 2021
2. P.L. Senger, The Estrus detection problem: new concepts, technologies, and possibilities. J.
Dairy Sci. 77(9), 2745–2753 (1994)
3. B. Sharma, D. Koundal, Cattle health monitoring system using wireless sensor network: a
survey from innovation perspective. IET Wirel. Sens. Syst. 8(4), 73–80 (2018)
4. P. Lovendahl, M.G.G. Chagunda, On the use of physical activity monitoring for estrus detection
in dairy Cows. J. Dairy Sci. 93(1), 249–259 (2008)
5. Alsaaod, M., Niederhauser, J.J., Beer, G., Zehner N., Schuepbach-Regula, G., Steiner, A.,:
’Development and validation of a novel pedometer algorithm to quantify extended character-
istics of the locomotor behavior of dairy cows. J. Dairy Sci. 98(9),(2015)
6. U. Brehme, U. Stollberg, R. Holz, T. Schleusener, ALT pedometer-new sensor-aided mea-
surement system for improvement in estrus detection. Comput. Electron. Agric. 62(1), 73–80
(2008)
7. S. Kerketta, T.K. Mohanty, M. Bhakat, A. Kumaresan, R. Baithalu, R. Gupta, A.K. Mohanty,
M. Abdullah, S. Kar, V. Rao, A. Fahim, Moosense pedometer activity and Periestrual hormone
profile in relation to Oestrus in crossbred cattle. Indian J. Animal Sci. 89(12), 1338–1344
(2019)
8. NDDB Database. https://www.nddb.coop/services/sectoral/national-database
9. Nordic Semiconductor: NRF24L01 Single Chip 2.4GHz Transceiver Product Specification
(2nd edn., 2007)
10. Cloud Services. https://www.digitalocean.com/
11. Software Architecture Description. https://medium.com/code-dementia/yoctosehns-software-
architecture-cff81658351c
12. R. Firk, E. Stamer, W. Junge, J. Krieter, Automation of estrus detection in dairy cows: a review.
Livest. Product. Sci. 75(3), 219–232 (2002)
13. A.T. Peter, W.T.K. Bosu, Postpartum ovarian activity in dairy cows: correlation between behav-
ioral estrus. Pedometer measurements and ovulations. Theriogenology 64(8), 111–115 (1986)
14. J.B. Roelofs, F. Eerdenburg, N.M. Soede, B. Kemp, Pedometer readings for estrous detection
and predictor for time of ovulation in dairy cattle. Theriogenology 64(8), 1690–1703 (2005)
15. I. Andonovic, C. Michie, P. Cousin, A. Janati, C. Pham, M. Diop, Precision livestock farming
technologies, in 2018 Global Internet of Things Summit (GIoTS) (IEEE, 2018), pp. 1–5
16. K.H. Kwong, T.T. Wua, H.G. Goh, K. Sasloglou, B. Stephen, I. Glover, C. Shen, W. Du, C.
Michie, I. Andonovic, Practical considerations for wireless sensor networks in cattle monitoring
applications. Comput. Electron. Agric. 81, 33–44 (2012)
Toward an Adversarial Model for
Keystroke Authentication in Embedded
Devices

Niranjan Hegde and Sriram Sankaran

Abstract The increase in computing and communication capabilities has enabled


embedded devices to perform a wide variety of sensing and monitoring tasks in
diverse domains. Keystroke authentication for embedded devices leverages the
unique typing patterns of users toward creating behavioral profiles for identifica-
tion. However, keystroke authentication can be bypassed by manipulating inputs thus
causing machine learning models to operate on perturbed data resulting in higher
error rates. In this paper, we propose to develop an adversarial model for keystroke
authentication in embedded devices. The proposed model leverages Feature Impor-
tance Guided Attack (FIGA) [1] method to add perturbations in the inputs which
in turn increases error rate thus allowing the attacker access to the system. Evalua-
tion using RandomForest, XGBoost, and SVM models shows an average Error rate
across all the users to be 60%, 55%, and 63%, respectively. Our analysis shows that
it takes an average of 3.18 s for the attacker to generate samples per user and that the
generated samples are realistic in nature. Finally, the computational complexity of
our proposed model is O(n) which is comparable with existing adversarial models.

Keywords Adversarial model · Keystroke authentication · Embedded devices

1 Introduction

The rapid proliferation of embedded devices with advanced computing and com-
munication capabilities has given rise to applications in diverse domains [2]. These
devices range from tiny resource-constrained sensors to more powerful data centers
typically known as Internet of Things which are capable of remote monitoring and
response. The advent of machine learning for embedded devices is considered a
“double-edged sword” which means that developing machine learning models can

N. Hegde · S. Sankaran (B)


Center for Cybersecurity Systems and Networks, Amrita Vishwa Vidyapeetham,
Kollam, KL, India
e-mail: srirams@am.amrita.edu

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 29
B. Mishra et al. (eds.), Artificial Intelligence Driven Circuits and Systems,
Lecture Notes in Electrical Engineering 811,
https://doi.org/10.1007/978-981-16-6940-8_3
30 N. Hegde and S. Sankaran

be used to both enable as well as disable security for embedded devices [3]. These
scenarios are increasing prevalent given the increasing challenges of embedded appli-
cations coupled with the varying capabilities of attackers.
Passwords are typically used in embedded devices for authentication [4]. These
devices are typically pre-configured with default passwords which can be guessed
with ease and authentication process bypassed. Thus, passwords involving a com-
bination of numbers, capital letters and special characters are required to increase
the difficulty of password guessing. Despite these requirements, attackers are able
to bypass them since the requirements are typically not followed.
2-Factor Authentication (2-FA) has increasingly become the de-facto standard for
system requiring high-level security for systems such as banking. The idea behind 2-
FA is to authenticate users based on what they know (password) and what they have
(token) [5]. With the advancement in technology, biometric-based authentication
has become a viable option. Authentication system may use information such as the
retina eye scan and fingerprints for authentication. However using biometric form of
authentication raises privacy concerns [6].
Keystroke-based authentication is considered a form of biometric authentication
with low risk to the privacy of the user because the users are identified based on their
typing patterns. It can also be combined with mouse dynamics to further enhance
the authentication process. In a keystroke-based authentication, system stores the
keystroke timing patterns of the users while typing. Based on their typing patterns,
users are authenticated by the system. To facilitate keystroke authentication, system
typically uses machine learning algorithms to classify the typing patterns of users.
However, keystroke-based authentication models are not secure since they can be
bypassed to allow attackers access to the system. In particular, data or the model can
be manipulated to cause perturbations thus causing the machine learning models to
misclassify data. This emphasizes the need for modeling the adversarial behavior on
keystroke authentication and analyzing its impact on embedded devices for varying
number of users.
In this paper, we propose to develop an adversarial model for keystroke authen-
tication in embedded devices. In particular, we leverage the FIGA [1] method to
add perturbations in model inputs thus causing the machine learning models to mis-
classify data and further enable the attackers access to the system. We also compare
the computational complexity of our proposed adversarial model against existing
adversarial model. Evaluation using machine learning models such as RandomFor-
est, XGBoost and SVM shows an average Error rate across all the users to be 60,
55 and 63% respectively. Our analysis shows that it takes an average of 3.18 s to
generate samples for all users and the average change in generated samples across
all users is 1 s which denotes the realistic nature of the samples. Finally the computa-
tional complexity of our proposed model is O(n) which is comparable with existing
adversarial models.
Toward an Adversarial Model for Keystroke Authentication … 31

2 Related Work

There exists numerous forms of authentication for embedded devices such as authen-
tication using Photo Response Non-Uniformity (PRNU) of a smartphone described
by Nimmy et al. [7]. In our work, we focus on the existing literature on keystroke-
based authentication systems and developing adversarial models.
Keystroke authentication: In [8], authors crowdsourced using Amazon’s mechan-
ical turk [9] where they collected typing patterns of different users. They selected the
password “.Generation!8765” which users need to type for 5 times. They selected 3
features: Key-hold, Key-flight, and Key-latency. Key-hold refers to the time until the
key is pressed. Key-flight refers to the time key is pressed and the next key is pressed.
Key-latency refers to the time between leaving the key and next key is pressed. In
their paper, they have set the threshold as 75% of the hold time and standard deviation
for the feature should be within 2.5 standard deviations. With this approach, they
had a false acceptable rate of 0.00% however false rejection rate was 13.89%.
In [10], authors used the dataset provided in [11] and evaluated various machine
learning such as KNN, SVM Random Forest, and XGBoost. They found that
XGBoost got the highest accuracy of 93.5%. Similarly, in [12], authors also showed
that XGBoost was the preferred model for keystroke-based authentication model.
Authors have obtained an accuracy of 90.91% using their approach. In our paper, we
have launched the attacks against these machine learning models to demonstrate our
approach.
In [13], authors used the dataset provided in [11] and developed an Extreme learn-
ing model (ECM) and Extreme learning model combined with evolving cluster model
(ELM-ECM). Using the dataset, ECM achieved an accuracy of 89.27% whereas
ELM-ECM obtained 86.97%. Since the accuracy is not higher than XGBoost, we
did not launch the attack against this machine learning model as a part of our current
work. In future work, we propose to consider these deep learning models as well.
Data Perturbation: Negi and Sharma [14] performed adversarial attack on
keystroke authentication model where they took the average of the typing patterns
and created an artificial sample. Further, k-means clustering was utilized to generate
artificial samples per user. The authors found that they were able to compromise 50%
of the users using their proposed mechanism. Negi et al. [15] proposed a novel algo-
rithm K-mean++ which generates adversarial samples by iterating over the samples
and passwords. They also show that this algorithm could be used to attack other forms
of biometric-based authentication such as touchscreen swipes. They compromised
40–70% of the users by using varying degree of their attack strength.
Khan et al. [16] developed a mimicry attack on smartphones which guides attack-
ers to provide inputs similar to the victims entering passwords. In particular, authors
assumed that they have access to the victim’s password and typing patterns after
having captured them while typing. They also train attackers via help of a visual
aids to enter the password similar to that of a victim. However, this paper does not
focus on exploiting machine learning but instead on accurate mimicry of the victim’s
typing password once the attacker has learnt about the password.
32 N. Hegde and S. Sankaran

Sun and Upadhyaya [17] created a master key using a large dataset where wolves
and lambs were identified. Using lambs and wolves, they create a master key which
is used to increase the Equal Error Rate (EER). The authors were able to increase
EER by 22%. Their approach requires knowledge of the dataset to determine the
lamb and wolves to attack the model.
In contrast to the existing approaches, we develop a novel adversarial model
for keystroke authentication in mobile embedded devices using Feature Importance
Guided Attack (FIGA) method.

3 Keystroke Authentication: Background

Figure 1 describes the working of a keystroke-based authentication model. First,


users would register their typing patterns of their passwords. Further, the system
would preferably ask the user to type in the same word a few times to account for
any deviations. Then, the system would keep track of the time taken for pressing a
key, releasing the pressed key, and total time until the next key is pressed after a key
is released. Finally, the system would store the typing patterns of users and learn to
differentiate among them.
The system for uniquely identifying the typing patterns of users can be developed
using statistical methods and machine learning models. Once these models are able
to differentiate among the users based on their typing patterns, they are further used
for identification.
When the user tries to login to the system by typing in the password, it would first
check for the validity of the password. If the password is successful, system would
track the time taken by the user to press the key and latency between the keys. These

Fig. 1 Keystroke authentication model


Toward an Adversarial Model for Keystroke Authentication … 33

Fig. 2 Feature extracted from typing pattern

details would then be fed to the trained model to predict whether the typing pattern
matches. If it matches, then the system would enable the login. On the other hand,
authentication is rejected if password is invalid.
Figure 2 shows the information captured during the process of typing. Three fea-
ture values that are computed are the following:
• Hold time denotes the time between pressing the button and release.
• Time between pressing the first and the next button.
• Time elapsed after releasing the first button and pressing the next button.
For every adjacent key-pair in the password, above feature values are computed.
In Fig. 2, we have assumed that the key-pair to the period button and character “t”
button. If the next button to press is the character “e” button, then the key-pair would
become “t” and “e” button. The feature values would be calculated again for this
key-pair. It is necessary that the sum of hold time of the first button and time between
releasing the first button and pressing the next button is equal to the time between
pressing the first and the next button. Thus, feature values are extracted from the
typing patterns which is then used for training the model.

4 Proposed Adversarial Model

In our proposed adversarial model, we have made the following assumptions:


• Attacker knows the password and typing patterns of the users.
• Attacker can provide generated samples as input.
Figure 3 contains the pictorial description of the proposed adversarial model. The
attacker utilizes the testing samples to compute the importance of each feature and
rank them accordingly. After performing feature ranking, attacker needs to select
two parameters  and n which are used to control the amount of perturbation that can
be added in the samples. Finally, attackers use FIGA to generate artificial samples
of typing patterns that are then used to attack the keystroke authentication model.
34 N. Hegde and S. Sankaran

Fig. 3 Flowchart of
proposed adversarial model

FIGA is modified such that the relationship between features as explained in


the earlier section is maintained thus ensuring realistic perturbations. The modified
samples when provided to the keystroke detection model cause the Error Rate to
increase thereby allowing the attacker to bypass the model and gain access to the
system.

Algorithm 1 Modified FIGA Algorithm


Input: f, n, , x
Output: x ∗
1: f n ←f[n]
2:  ← n ∗ sum(x f eatur es )
3: for i ← 0, n do
4: x[i] ← x[i] + ( ∗ d[n])
5: if x[i] < 0 then
6: x[i] ← 0.01
7: end if
8: end for
9: return x ∗

Dataset The CMU dataset [11] is being used in the paper. The dataset contains
the typing pattern of several subjects. Subjects typed the password “.tie5Roanl” in
the keyboard. Each row contains the typing information of a single repetition of the
password by a single subject. The feature set contains three types of features: the
hold time for a key, the time between pressing the next key when the current key is
released, and time between when the key and the next key is pressed.
Modified FIGA Algorithm Before the algorithm is utilized, feature importance
ranking of the feature is performed using any ranking method. In this paper, Gini-
impurity method is used for feature ranking. Once the feature has been ranked, the
direction of each feature is computed by taking the sign after subtracting the mean of
Toward an Adversarial Model for Keystroke Authentication … 35

the feature values of the target users and those of the feature values of the rest of the
users. The computed direction can be negative or positive through which direction
of perturbation is known.
After computing the feature ranking and its direction, modified FIGA algorithm
is applied. Inputs to Algorithm 1 are f (Rank and attack direction of the feature), x
(samples to perturb)  (the total percentage to modify) and n (number of feature to
modify).
Step 1: Given n features from the list of ranked feature importance, additional
features are added into the list to maintain the relationship. For example, if the feature
to be modified is hold time of the key “e”, then the feature where the time between
the pressing of key “e” and the next key is also added into the list.
Step 2: In this step, sum of the feature vectors is computed and multiplied with
. The resulting value denoting the total amount of perturbation is divided by n to
effectively distribute the perturbation among the selected features.
Step 3: In this step, the resulting perturbation obtained from Step 2 is multiplied
with the direction of the feature before being applied to the features. Perturbations
are being added to only n features ranked by the feature importance ranking method.
Step 4: In this step, negative values in samples are replaced with the value of 10
milliseconds to ensure that samples generated are realistic in nature.
Step 5: In this step, we obtain the adversarial samples created by the modified
FIGA algorithm which are then tested against machine learning models.
Experiment Setup: We use the Scikit-Learn library [18] running on a laptop
machine with 16 GB RAM 8 core process for implementing the machine learning
models. We utilize 80% of the dataset for training and 20% for testing the trained
model. The adversarial samples were created from the testing dataset. The value of
 and n is set to 0.01 and 3 so that the perturbations added are minimal and optimal.
We select Random Forest Classifier, XGBoost Classifier, and Support Vector
Machine (SVM) for Keystroke Authentication. The rationale behind selecting the
above models is that they provide better performance compared to other approaches.

5 Evaluation

In this section, we evaluate our proposed adversarial model using Random Forest
classifier, XGBoost classifier, and Support Vector Machine. Further we perform the
analysis of the overhead and the crafted adversarial samples. Finally, we compare
the computational complexity of our proposed model against existing adversarial
models.
Metric: The goal of our proposed model is to increase the classification error in
keystroke authentication model. The chances of the attacker authenticating as another
user increases with the increase in classification error. Therefore, we use Error Rate
as our metric to evaluate the performance of keystroke authentication model and our
proposed adversarial model. Error Rate is computed by dividing the sum of False
36 N. Hegde and S. Sankaran

Fig. 4 Error rate per subject before and after attack (XGBoost Classifier)

Fig. 5 Error rate per subject before and after attack (Support Vector Machine)

negatives and False positives by the total number of predictions. Error Rate can be
expressed using the following equation:
False Positives + False N egatives
Err or Rate = (1)
T otal N umbero f Prediction

Results: We test our modified FIGA algorithm against machine learning models
such as RandomForest, XGBoost, and Support Vector Machines (SVM). Using the
RandomForest model, Error rate for a few users such as user 53, user 42, user 39,
etc. is 99% as shown in Fig. 6. Similar results can be seen for SVM (Fig. 5) and
XGBoost models (Fig. 4) as the error rate for the same set of users was over 95%.
This shows that these users are a good target for the attacker to use in order to bypass
the system. Thus, making minimal changes to the typing patterns of any of these
users, the attacker can masquerade as another user for authentication.
On the other hand, there are few users whose typing patterns are unique and
resilient to manipulation such that it does not cause a significant increase in error
rate. From Figs. 4, 5 and 6, it is evident that there are users whose error rates are
less than others. This means that the attacker masquerading as another user for
authentication would likely be unsuccessful.
Toward an Adversarial Model for Keystroke Authentication … 37

Fig. 6 Error rate per subject before and after attack (Random Forest Classifier)

Fig. 7 Time taken to create adversarial samples from all samples of a user

5.1 Overhead Analysis

Figure 7 contains the results for overhead analysis of the proposed adversarial model.
In particular, we estimate the time it takes to generate adversarial samples from all
the user samples. On an average, it takes 3.18 s to generate adversarial samples across
all users. From the figure, we observe that generating samples for the first 27 users
requires more than 3.5 s, but incurs 2.9 s for the remaining users.
This difference can be attributed to the computation of feature importance using
a Random Forest Model. This model is first fitted onto the samples provided by
the attacker. The model uses Gini-Impurity to compute importance values for each
feature. For users before 27, Random Forest Model requires more time to fit itself
onto the samples.

5.2 Analysis of Crafted Adversarial Samples

In our work, we use modified version of FIGA to generate samples that increase
the error rate in the keystroke authentication model. This allows the attacker to
Another Random Scribd Document
with Unrelated Content
horgonyába. Végre is ez az egyetlen ember, a kihez egész életén át
jó volt és összes barátai között ez az egyetlen egy, a ki most már
reális téren mozog. Ki tudja, mi történt vele? Talán elvette egy
kiskereskedő leányát és kapott néhány ezer forint hozományt. Ebből
ki lehetne szakítani egy-két ezret. És Samu meg is teszi, ha teheti.
Az ujságíróra egyszerre úgy hatott ez a kölcsön-ügy, mint egy
reveláczió. Lázasan, mohón kutatott Samu után.
Nem volt nehéz megtalálni. A «tejelde» a városnak egy ujonnan
csinált nagy vérerén és igen exponált ponton állott. Igen elegáns. A
czímtáblája fekete üveg, arra arany betükkel kiírva, hogy az üzlet
mikor alapíttatott. Az évszám a Samu születési évét jelezte. Ezen
mosolyogva nyitott be a boltba Asztalos Aurél, a kit a gazda a
tüntető figyelem minden jelével fogadott. Kaviáros hordót bontatott,
szardiniás skatulyákat nyittatott, kikergette az aludttejért jövő
cselédeket, a segédszemélyzetet – egy varas állú szép kis fiatal
leányt meg egy nagy mafla fehér ruhás szolgát – csaknem
megőrjítette ellentétes parancsaival. A nagy örömtől alig tudott
meghiggadni maga is. Az Aurél ajánlatára végre is bementek a bolt
melletti szobába, a mely raktárnak és egyszersmind magánlakásúl
szolgált. A vörös képű, gyöngyöző homlokú Samu itt is alig engedte
szóhoz jutni barátját.
– Tudok, mindent tudok. Megveszek minden ujságot, a mibe irsz,
vagy a miben rólad van szó. Hogy annyit tud írni a magad fejéből?!
Az iskolában is szabadkézből nagyon szépen rajzoltál. Ha te akarnád,
lehetnél festő is. Te vagy a mi büszkeségünk; én csak kínlódom. Ma
hajnalban már kint voltam a vásárcsarnokban, kötöttem: ezer forint
ára tojást! Ki is fizettem. Én mindent készpénzzel veszek. Ez az üzleti
titkom.
– És van készpénzed?
– Van. Honnan, azt kérded?
Egy roppant valószinűtlen történetet adott elő, a mibe csakhamar
belezavarodott. Abba is hagyta a következő rövid kijelentéssel:
– Elég az hozzá, hogy van. Kell? Ha meg nem sértenélek…
– Nem azért jöttem éppen – mondá az ujságíró boldogan – nem
annyira ezért, hanem látni akartalak és hogy bizalmasan közöljek
veled egyet-mást. Valakinek el kell mondani a titkaimat és a kiben
biztos vagyok, hogy el nem árul, az te vagy. Legegyszerűbb, de
legbecsületesebb barátom!
Az ujságíró most – kissé népiesen – de elmondott mindent,
szerencséjét, terveit, czéljait. Igen csekély habozás után megmutatta
a Nuszi leveleit, a Juli sürgönyét. Mit sem hallgatott el, sőt talán
mert belejött a beszédbe, – valamivel többet mondott. A Samu kis
szeme könnyezett; kifeszített, a sok tejivástól csaknem kicsattanó
arcza ragyogott. A szája is nyitva maradt, a mint hallgatta. Nem
kedvetlenedett el akkor sem, a mikor az ujságíró kijelentette, hogy
már most neki pénz kell, sok pénz kell.
– Adom, mind adom. Többet adok, mint a mennyi van. Ha kell –
itt van a sarkon egy vén, de gazdag trafikosné, – elveszem. Az estve
átmegyek hozzá, várj, mindjárt. Eladom ezt a boltot, ma sokat ér. Az
a baj nálatok íróknál, hogy pénz nélkül hagyjátok magatokat. Nincs
forgó tőkétek, csak a mi az eszetekben van, az mindennél többet ér.
Ha nekem olyan eszem volna, mint neked! Te – mondá a boltos
susogóra változott hangon – nem lehetne a herczegtől mindjárt
valami készpénzt kérni kis kamatra. Azon megvennénk mind a tojást,
a mi csak van Magyarországon…
A «tejeldés» fantáziája kezdett a maga útjára kikalandozni, ámde
az ujságíró nem engedte.
– Mennyit adhatsz most?
A Wertheim-kasszát – az is volt – kikotorták, úgyszintén a
fafiókot. Egy cselédkönyvben úgynevezett «letett» pénzek voltak, azt
is kiszedték. Ezer és néhány forintot adott Samu – minden írás nélkül
– a barátjának, a kit ölelgetések között kisért el egy fiakker-
állomásig.
Föl, föl a gummitalpú alkalmatosságra, ujra föl. Ezek a lovak
vágtattak csak büszkén és vidáman. Gorombáskodva menekült előle
a nép, majd a belvárosba érve hódolattal nyitott előtte utat.
Megálltak egy új szabó előtt, egy új divatárús-bolt előtt, a havanna-
szivaros boltja előtt. És a fiatalember nem is emlékezett többé a
délelőtti megaláztatásra. Úgy tetszett neki, hogy nincs könnyebb
valami, mint pénzt szerezni. Csak egyszer találjon az ember a
nyitjára.
– Ez a Samu, ez nagyszerű evangéliumi szív és milyen szeme
van!
Elhatározta, hogy később nagy vállalatokhoz fogja segíteni. Addig
is nem neheztel többé arra a vallásfelekezetre, a melyhez a Samu
tartozik. «Kitűnő emberek vannak közöttük, annyi bizonyos!» Így
állapította meg és megtevén nagy bevásárlásait, kimentette magát a
szerkesztőségben, a melynek tagjai azonnal megérezték rajta a
jómód szagát és bizonyos tisztelettel bántak vele. Ő szórakozott volt,
gőgös volt, táviratot várt és kapott is. Sőt váratlanul levelet is. Annie
írt.
«Gyer vissza szerelmem!» ezt írta Annie francziáúl. Majd
magyarúl folytatta: «Én azóta mindig rosszat «álmodok» te rólad. Az
éjjel angol katona voltál, Zsiga – bátyám nekem – volt a tiszted és
egy hétágú macskával az arczodra, neked az arczodba vágott! A
drága véred mind a fehér piquet szoknyámra folyt és te erre nem
panaszkodtál, csak azt mondtad: bocsánat comtesse, hogy ilyen
rendetlenséget csináltam, de az én vérem az nem fog. Én a Zsigát –
nagyon szeretem őt – nagyon megróttam: hogyan meri ezt, nem
látja, ez a katona nekem vőlegényem? Mire Zsiga így: «mért nem
tett akkor maga rá piros selyemből egy kis jelet!» Aztán reggel
megint volt egy kicsiny álmom. Nagy, csunya, lompos, nem faj
kutyák között te álltál és lehajoltál a földre a kavics miatt, hogy őket
elkergetnéd. De a föld olyan volt, mint a parquette…»
«Oh bűbájos leányszív, előre megálmodta a mai kálvária
járásomat!» Gondolta magában Asztalos Aurél és némi hideg futott
át a hátgerinczén: «Ez a leány engem a kelleténél jobban szeret;
nem volna-e jobb abbahagyni, magamba utáltatni. A rossz álmokat –
a melyek rám vonatkoznak – nem szeretem. És ez a Zsiga
kellemetlen alak. Nem ismerem, de az ösztönöm nem mond róla jót.
Ezek az emberek iszonyúan exkluzivek; sok bajom lesz a családdal…
Telhetetlenség tőlem, nem elég nekem a herczeg barátnője?!»
Már-már meghátrált a tiszta, nagy szerelem elől.
«Konczentrálnom kell magamat egyre!» javítgatta a programmját, de
csak egy pár perczig, a míg a rossz álmoknak friss volt a hatásuk.
Hamar fölbátorodott, látta, hogy minden csaknem úgy megy, mint
akarja és kipéczézte. A Juli sürgönye csupa jót jelentett, már csókot
is küldött a villamon és a sürgős parancsot: «Siess!»
És a fiatalember sietett, a hogy csak tőle tellett. Még este
fölment a klubba, hogy a kormányzat fejének távollétében
jelentkezzék annál a politikai személyiségnél, a kinek az volt a
kötelessége, hogy a kezét folyton rajta tartsa az ország érzületén. A
jámbor aggastyán kedves mosolylyal – és félálomban – hallgatta a
fiatalembert, a kit felületesen, személyesen is ismert. Nyár volt, a
klub elhagyatva, senki sem informálta másképp, a szeretetreméltó
öreg elhitt mindent. Különben is ő volt az, a ki, bár körülötte mindig
és mindenki hazudott, mégis mindig elszörnyüködött és meglepődött
a hazugságon. Tetette-e magát, vagy csakugyan gyermek és tiszta
maradt közepette azoknak a politikai machináczióknak, a melyek
csupa csatak, iszap és összekuszáltság, mint a mocsár flórája?
Keztyüt tartott a lelkén, fehér keztyüt a kezén, a melyben az egész
rothadt és büdös flóra meg fauna összefutott? Álomban járt, így tett
és vett, valaki elaltatta vagy fanatizálta? De huszonöt évig
egyfolytában aludni és egyszersmind helyettes feje lenni egy
czéljaiban talán tiszteletreméltó, de eszközeiben nem szemérmes
maffiának, mert minek másnak lehet nevezni…
Nem tetette magát; valóban felnőtt és megőszült baba volt.
Elárulta a hangja is, a mint naivul és siránkozva mondta: «Pénzem
nincs, azt nem adok!»
Az ujságiró felfutta mellét:
– Pénz! – mondá mélységes megvetéssel. – Nekem nem kell
pénz!
– Nohát akkor menj isten hirével!
Ment is, még az éjjel és a hogy az uradalom székhelyére
érkezett, látta rögtön, hogy itthon, elemében van itt. Volt diadalkapú
– oh jogászkori álom – egy fiatal fenyőerdőt ledöntöttek miatta. És
voltak – bár a hirtelen jövetel okán csak rögtönzötten – fehér ruhás
lányok és volt különösen sok nép a munka mellől elhajtva és
beterelve a jószágigazgatói kastély udvarába, a melyet egy részük
ismert, főképpen az onnan kirohanó idegen kutyák harapása után…
A forró napban levett kalappal állott a nép és az ünnepi, tehát téli
ruhában szinte párolgott. Bután, alázatosan, de nem titkolható
rosszakarattal néztek az idegen szájára, a kinek csak egy-egy szavát
ha érthették. Azt, a mely az ő nyelvükbe is átment és a minek
megörültek mindig, egy-egy csoportjuk az ilyen szóra «éljen»-zett is.
Tetszésüknek különben kifejezést – szegények – nem igen adhattak,
bárha látták, hogy az uradalom ezt így akarja. Egyáltalán hirtelen
jött az egész; nem adták ki nekik, hogyan viselkedjenek és ha egy-
egy katonaviselt béres, csizmás boltos nem tartja őket össze, el is
szélednek mindannyiszor, a mikor a szónok nyugovót tart vagy iszik.
Meglátszott rajtok, hogy nem jól érzik itt magukat, félnek. És ennek
a csuf érzésnek a rút moraja töltötte be az udvart, a melynek állattá
aljasított, boldogtalanná tett nehány száz emberét egyetlen herczegi
szó emberré és boldoggá tehette volna. És nem is a nagylelküség
szavára, csakis az okosság egy mondására lett volna szükség. Csak
arra, hogy a ki kis földet akar bérelni drága pénzért, az jőjjön! Az
ujságiró pedig a «nemzeti állam kiépítéséről» beszélt nekik,
mindazonáltal – a mi a nemzeti minőséget illeti – mérséklettel, de
oly hosszasan, hogy az uradalmi tisztnek arczán a rizspor megindult
a forróságban. Az idegen ajkú jószágigazgató, a ki mint
nyugalmazott törzstiszt nem rokonszenvezett a politikával, már-már
a száját huzogatta, de nem a megvetés érzelmeivel, hanem
határozottan gutaütésesen. Az ujságiró azonban még egy jó fertály
óráig tovább szerette a hazát és erre felszólította a népet is, a mely
jóakaratú morajjal majd kieszközölt és közvetített éljenekkel felelt.
Végül imádkoztak a herczegért, mert ez minden választási
campagne, minden nagyobb mozzanatában így volt kiadva. Nehány
öreg paraszt le is térdelt, a pópa közbe jajgatott, a csizmás boltosok
lekapták fövegeiket, hátukat meggörbesztették. Mintha valami
epidemia szállta volna meg az egész udvart.
Az ujságiró fent nagyon is felül érezte magát. És e pillanatban
talán szerette is a hazát, mely ilyen felséges jogok gyakorlásához
segítette őt. Lázas, élénk és boldog volt egész napon át. Látta, hogy
itt minden biztos; sunyin, de babonás tisztelettel néznek itt reá. A
főtiszt talán tudja is, hogy mit jelent ő a fenségre nézve. Már
nagyságolja, a kastélyba a herczegi szobákba kvártélyozza be. Ott
van, a míg le nem vetkezik és a mikor elbúcsúzik tőle, sugva
mondja:
– Ő fensége a fenséges asszonyhoz utazott. Ő fensége barátnője,
ha éjjel meg nem jön, reggelre megjön.
Sokkal jobban meg volt elégedve magával, mintsem hogy
lemondjon öntudatáról és elaludjék. Rendkívül élvezte a módot, azt,
hogy a barokk fodros párnákon tenyérnyi nagyságú czimer-állatok
vannak kihimezve és ő a képét épp arra fektetheti. Horgolt, sárga
selyem takarója is czimeres és «az anciennitás illata van rajta»
gondolta magában az ujságiró egész komolyan. Ám, hogy a régi jó
módnak különösen jó illata van, az közismert dolog, valamint az is,
hogy a szegények ha hozzájutnak ehhez az illathoz, rendszerint
megszédülnek. Korunknak most kéjesen nyujtózkodó hőse is ezuttal
egyszerre megértette és átérzette azt, a mit a történelemből –
főképp a magyarból – eddig nem jól értett és nem érzett semmikép.
Azt, hogy luxus-kedvelő állatok vagyunk vagy leszünk csakhamar. A
hazaárulás technikáját az ujságiró ezuttal látta, még pedig nem is
mint valami nehéz, természetellenes és szövevényes, hanem mint
egészen egyszerű és plauzibilis dolgot. Azonban nem a nagy
hazaárulást, csak a kicsinyt és alkalmit kell itt érteni…
A vendég szivarozgatott és hallgatózott. A régi butorok bizony
nyikkantak egyet. A fenséges család házi kisértetei is neszeltek az
ujságiró tiszteletére. Vagy puderezett haju ősök és vékony, hosszu
szájú, lapos mellű ősnők restelték, hogy egy papszabó fia fekszik a
koronás vendégágyban. Az ujságirónak talán még sohasem volt ily
kellemetlen az eredete és igen erősen képzelődött, hogy e miatt itt
még bajba keveredhetik. Félálomba merülvén, egyszer föl is rettent
arra a képzelődésre, hogy egy tubákos orrú komornyik megérinti a
vállát és szól:
– Micsoda szemtelenség, ide tolakodni ebbe az ágyba! Mars ki, az
alsó cselédekhez, a pecsenyeforgatókhoz!
Azonban nem a komornyik állott előtte, hanem Juli. És ah, most
nem cselédruhában, hanem herczegnői csipkepongyolában és ehhez
képest puderezett hajjal. Ha a fiatal ember ebben a pillanatban és ily
viszonyok között látni tudott volna részleteket is, okvetetlen
megfigyeli, hogy az asszony öltözetének, a mi a harmoniát illeti, nem
használt a gyorsaság.
Herczegnői csipkepongyolában, puderezett hajjal.

Juli ki is nevette magát.


– Komikus vagyok, mi, de szép vagyok, jó vagyok, he? A
herczegnő nagy hozzám képest és nem hoztam hálóreklit. Aztán
nekem ez juxot csinál. Beszélj hát. Örülsz, hogy eljöttem?!
A fiatalember kissé készületlen volt mindenféle szónoklatra és
most már csakugyan tartott attól, hogy nem hagyja el a herczegi
kastélyt a maga lábán. Zavartan speechelt és engedelmet kért, hogy
ágyban fogadja, mire Juli dühbe jött:

É
– Ti irók mind affektánsok vagytok. Én egy napra lejövök. A
herczeg mondja: Juli, menj és lássad, nehogy annak a
fiatalembernek baja legyen. Ha belefogtunk, emeljük föl, ezt követeli
a férfi becsület. És te, te gyáva!
Az asszony befogta a férfi száját, nehogy az szónokolhasson, ő
beszélt:
– Tudod-e, hogy féltem: megszököl. Azt hittem, hogy te is olyan
férfi vagy, a kinek nincs kitartása. Kezdenek, kezdenek, aztán
elfáradnak mindjárt és tovább mennek. Mi csak nem szaladhatunk
utánatok?
Erős althangja, teli kaczagása fölverte a bolthajtásos termek
csendjét. A férfi intette, hogy «csendesebben!»
– Eh – mondá a nép leánya – ez nem olyan hely, a melyet
tisztelni kell; itt nem születtek herczegi gyermekek. Ide ő fenségeik
nem hozták el hires családi erkölcseiket. Itt vannak a
hitbizományban a legjobb ágyak és a legjobb borok! Te – fujj a
szemembe – ha a te apád is nem volt mesterember. Látom arról,
hogy a kezedet minden ok nélkül a szájadhoz viszed. Az én apám
kőmives volt!
Juliban egyszerre kitört a szinésznő. Belékezdett egy utczai dalba,
a melynek a hőse tiz mesterlegény és egy málter-hordó leány volt.
És a régi belga csipkékkel csinálta hozzá a dekorativ elemet.
Óriási kedve támadt; mindenen nevetett, mindent kicsufolt, főkép
a fiatalembert, a kinek nehéz – és alapjában sivár – kedélye nem
tudott fölemelkedni a helyzet szokatlanságához. A férfinak még
imponált a magas uraságok e kiczifrázott, kivattázott óriási vaczka. A
szinésznőnek azonban nem volt egyéb, mint diszlet.
Megparancsolta a fiatalembernek, hogy öltözzön föl – a
fönségnek egy katonaköpenyege épp kéznél volt – aztán karonfogva
mutatta meg neki a díszlet jeles helyeit.
– Ez az otromba nagy hálóterem, ezt nézzed meg. A herczeg
atyja itt végezte be életét, itt és nem a családi házban. Az öreg,
tudni kell, nagy ravasz volt, vidáman élt nyolczvanéves koráig; de
akkor leesett a lábáról, mindég feküdnie kellett. Nem volt beteg,
csak elgyengült. Ennek okáért, hogy a gyöngeség erőt ne vegyen
rajta, hat pár tizenhat éves ruthén leányt hozatott föl a faluból. Azok
éjjel-nappal itt tartózkodtak e szobában az öreg körül, fiatal
lélekzetüket, hogy beszívja és ezzel ő maga is fiatalodjék. Keskeny
ágyaiknak a helye itt van a pallón… Csak én éltem volna akkor, csak
én lettem volna az egyik ruthén leány…!
A családhistoria orgiáinak rövid, de alapos ismertetése
felgyújtotta a fiatalembernek rendszerint nem könnyen szárnyaló
fantáziáját.
Fölkiáltott:
– Nagyszerű ember volt. Éljen! Ezek mind nagyszerű férfiak!
– Gyalázatosak! – mondá Juli harag nélkül.
– Miért?
– Mert a hat pár ruthén leányt aztán nehány ezüst tallérral
fizették ki. Azokból hat pár uri dámának kellett volna lennie. Azért a
förtelmes nehány hónapért legalább is nyugdíj járt volna nekik. Fuj,
az öreg, az isten nyugtassa – nem volt korrekt ember, csak mindig
mondta – alighanem ő is mondta – hogy az. Hej, az enyém is, ha én
nem volnék én…!
Juli népies mozdulatot tett kezével és leülvén az óriási ágy
szélére, ennek a káváját verte, úgy harsogott:
– Ezek mind azt hiszik, hogy az isten pusztán az ő kedvükért
teremtette a földet és minden csak értük van. Ha egy gyerek ürgét
fog a mezőn, az az ürge az övék. A mikor a herczeg engem
«magának elvett», egy félévig kölcsön pénzből kellett élnem. Gránát
ékszereket és inggombokat hozott; mind oda adtam barátomnak a
baritonistának. Nyilván azt akarta ez az öreg férfi, hogy önmagáért
szeressem. A mikor el akart hagyni, negyedrész értékükre leszállott
részvényeket hozott az udvarmestere. Kidobtam, még a lábammal is
segítettem rajta, hogy gyorsabban haladjon. Arra a fenség visszajött
és azt mondta: te egy fess lány vagy, csak próbára akartalak tenni!
Kapott egy jó nagy gorombaságot. De úgy látszik, ez kell nekik.
Néha föl kell ezeket világosítani, hogy nem istenek. Csakhogy ez
nagyon bajos. Az okos embereket nem igen türik a maguk
közelében. Én kivétel vagyok. Azért van, mert én nagyon kedves
vagyok!
Komolyat, tréfát így vegyített össze az asszony, a ki ezuttal
nyilván haragudott a herczegre. Minden esztendőben megtörtént ez
egyszer-kétszer, a mikor a fenség felnőtt leányai apjukat el akarták
mellőle intrigálni és az egy kissé bele is ment a tervükbe. Rendesen
nyáron és karácsony táján szokott ez megtörténni, de eddig nem volt
baj: ő fensége kiélvezte a családi boldogságot és egy hét mulván
visszatért.
A fiatalember szivében kissé elhidegedett az asszony iránt, a
mikor meghallotta, hogy ilyen válság köszöntött be hozzá. Azonban
nem volt hozzá kevésbbé szives, alázatos és szerelmes. Az idegen, a
regényes, sőt történelmi hely ugyan elfogódottá tette a hangját,
mégis tőle telhetőleg udvarolt. Ám a nő megismerte hangjának
hamisságát, megérezte érzelmeinek csináltságát és megvetőleg
hárította el magától:
– Menj – mondá – ostoba vagy; nem tudsz velem bánni.
– Hát hogy kell?
– Sehogy. Megfogni a kezemet. Nem szólni egy szót sem. Csak
nem szavalni. Néha sóhajtani egy nagyot. Mindig a szoknyám szélén
ülni. Ha alszom, aludni, ha ébren vagyok, ébren lenni. Ha eszem,
enni. Mindent úgy tenni, a hogy én akarom. Én csak azt a férfit
szerethetem, a ki nekem engedelmeskedik.
– Jól van, engedelmeskedem!
– Mondani azt nem szabad; tenni kell.
Egymásba fonták az ujjaikat és így lóbálták a kezüket, mint
vasárnap estve a sétányon egy altiszti pár. Mint egy posztó csillagos
katona és bekötött fejű cseléd. És olyanformán beszéltek is, mert a
herczeg barátnője így akarta. Beszéltek a köztük megtörténendő
egybekelésről, pedig épp oly biztosan tudták, hogy ebből nem lesz
semmi, mint a hogy biztosan tudja ezt az egymást bolondító
asszonyi pár. Közben egy-egy puha csók is esett.
Nem volt péntek…
A szolgai eredet erőt vett mind a kettőn és a nagyszerű
interieurök immár nem voltak kedvökre valók. Elindúltak keresni
valami egyszerübb szobát, miközben Juli megéhezett.
A nagy ebédlőben terítve volt a számára, óborok, pezsgő az
asztalon. Leültek lakmározni és az asszony ugyancsak rendesen
evett és ivott. Belé is szédült a mámorba mind a kettő, az egyik, a
férfi, életében először. Ám nem lett vidám, inkább fölötte szomorú.
Az anyját siratta, a hazát siratta és oly erősen érezte a lelkiismeretét,
mint gyomrában a tüzes folyadékot.
– A jó herczeg… Kitünő és nemes ő fensége, ezt nem érdemli
tőlem… Hogy én a családi otthonában… A mikor ő jót tesz velem és
nem védekezhetik, bünt kövessek el ellene!
A szó szoros értelmében sírt. Az asszony méltányolta fájdalmát,
törülte könnyeit, de szintén könnyezett, mert alapjában jó szivű volt,
a mint az már eddig is kitünt. De egyszerű természete és jókedve
még is győzött. Elismerte a férfi előtt, hogy nála becsületesebb
embert még nem ismert és hogy ezt alkalomadtán meg fogja
mondani a herczegnek. Megnyugodtak valahogy és miután a hajnal
közelléte a befüggönyözött és bespalettázott termek daczára
érezhető lett, még egy sétát tettek, mielőtt elpihennének.
Utjok közben egy kedves kis fehér szobára bukkantak. Juli itt
tréfásan szalutált:
– Itt volt a herczeg fiatal, ide hozta az első feleségét. Jó estvét!
A fiatalember szintén szalutált, de nem tréfásan, hanem némileg
meghatva.
… A szegény emberek rongyos paraszt kakasai odakint rekedten
üdvözölték a hajnalt. Vagy hivták vissza a fiatal tyukokat, a melyeket
mind az uradalmi asztalon fogyasztottak el? Nem lehet tudni. A
legegyszerűbb dolgok is olyan rejtelmesek.
Szerda volt.
… Délfelé a kasztellánus, sőt maga a jószágigazgató is jelentette,
hogy ő méltósága – ez volt a Juli czíme itt – az éjjel megérkezett és
szivesen látja a jelölt urat magánál.
Asztalos Aurél frakkot is hozott magával, és mert eszébe jutott –
egy franczia regényből – hogy az előkelő és hálás szerelem, ha csak
lehet, frakkot ölt: a déli óra daczára gálának öltözködött és úgy
jelent meg Julinál. Az igazgató is jelen volt, tehát szép szertartással
folyt le a vizit, a mely alatt az ujságiró úgy tisztelte a meghódított
kedvest, mint egy herczegnőt, a kinél először van audienczián.
Formás volt, alázatos és uri volt – legalább ő úgy érezte. És a mikor
egyedül maradtak, akkor sem lépett az asszonyhoz közelebb, tetette,
mintha az agyából kiesett volna a kis intermezzó, a méltóságos
asszony utjáról és Gasteinról tudakozódott. Juli, a becsületes és
egyszerű lény nem állhatta soká, megtapintotta az orrát és szólt:
– Maga a legnagyobb hazug. Ki nem állhatom a jezsuitákat
polgári ruhában!
Aurél még mindig udvariaskodott és hajlongott, miglen az
asszony megcsipte a karját:
– Ugyan kinek szinészkedel? Kinek? Nekem? Magadnak? És ez
ostoba frakk?! Ti férfiak mindig csaltok. Már nem is kezdhet veletek
egy becsületes nő!
A kis vitába és a morálprédikáczióba – a melyet más
viszonylatban a herczeg szokott ebéd előtt tartani és nem Juli – nem
vegyült üröm egy cseppnyi se. Az asszony vidám volt, mind a harczi
kanári madár, a mikor friss tyúkhurt, uj eledelt tesznek be a
kalitkájába, a férfi pedig méltóságos, elégedett és tisztesen – sőt
kedvesen – szomorú. Egyébként úgy érezte magát, mint egy igazi
uriember, a mely elnevezéssel az ifjú publiczista és társai – kik vele
kongeniálisak, annyit veszkődnek a legujabb időkben. Asztalos Aurél
most is példálózgatott vele, hogy ő uriember és bizonyára
jóhiszeműen ejtette ki ezt a jellegző szót, nem gondolván arra, hogy
mily felfuvalkodott és oktalan majomkodás ez és hogy igazi úr ő meg
a fajtája sohasem lehet. Ahhoz legelsőbben ancienitás kell az, hogy
már a nagyapjának az apját is annak nevelték legyen, vagy ha ez
nem történt meg, akkor egy kemény és puritán karakter, a ki minden
pillanatban halni kész e szóért és bizonyára meg is hal e fogalomért.
Uri ember! Nyavalyás jelző, melyet ujra kitalált a magyar társadalom
számára valami munkátlan alak, hogy magának hivatást, pozicziót és
jogkört találjon. Ám a mikor megállapította a fogalomnak
kriterionját, már eleve el volt telve uratlansággal és azzal az
elhatározással, hogy noblesset fog követelni a szobapinczértől is; de
neki, ha úgy tetszik, előre meg vannak bocsájtva a szobapinczéri
hangulatok, sőt – ha nem nyiltak – a cselekedetek is. Úri ember!
A mi kis hősünk valóban érezte most, hogy ő az és az emberi
lények túlnyomó részét – rosszszivűség és öntudatosság nélkül – a
szolga szerepére kárhoztatta. A kiválás korszakában így készülnek az
új meghatározások, a melyekről Széchenyi még nem tudott, pedig
meg volt benne az úrnak egyetlen igazi qualitása: a nagylelkűség és
az önfeláldozás!
Az ujságiró azonban nem mélyedt el ezuttal sem a társadalmi
tudományba, sem a politikába. Csak úgy ösztönszerűen érezte, hogy
az utolsó tizenkét óra alatt, ha nagyon közvetett uton is, de sokkal
előkelőbb lett. Valami grand seigneuri vonást fedezett föl magában,
mintha valami – merőben ismeretlen – cselekedet révén a
kiválasztottak közé jutott volna fel. Mint irónak – sok társával
egyetemben – az volt a végczélja, hogy két orosz trapperrel, a maga
fogatával vágtathasson ki majdan a tavaszi lóversenyre. A divatos
nagy munkának megirása és az irodalomnak más térre való hasznos
átjátszása után – természetesen. Most már inast – még pedig
borotváltat – képzelt az osztrák kocsis mellé. (Csak azok tudnak
hajtani!) Édes álmai, kedves és könnyed fantáziái a mint fejében
jártak és ott ült Juli előtt szótlanul, az meghuzta a haját:
– Föl, föl! Menj a dolgodra. Az után járjunk, a mi után élünk. El
kell menni a főispánhoz (morfinista). A szolgabiróhoz, a ki mindig a
nyulakat zavarja és kutyáit csókolja vagy veri (mint a feleségét).
Különben el se menj, egyszer végig kucsirozunk a sandlauferen – a
jószágigazgató is velünk jön: megtudják vagy látják és minden
rendben van.
Végig kocsikáztak: minden rendben volt.
– Most el innen! – szólt az okos Juli. – Én visszamegyek
Gasteinba, a fenség már visszaszökőben – te is oda jösz és ott várod
meg az eredményt. Ezekre többé rá se fujsz! Ezekkel pedig, nekem –
gőgösebben bánj, különben az ujjadra néznek és látják, hogy tintás.
Bár nehezére esett, megfogadta, gőgösebb volt, a füstölő szivar
mellől beszélt velük. Kezet sem adott mindig és mindenkinek,
ellenben a főbirótól elfogadott ajándékba egy kölyök agarat. Alig
várta azonban, hogy négy-öt állomásnyira lehessen, egy ismeretlen
község határában kieresztette a szegény tájékozatlan ebet. Csak a
hozzávaló korbácsot tartotta meg a kezében; ezzel jelent meg a
szerkesztőségben.
Ám ott többé nem nevettek rajta.
Egy-kettőre Gasteinban volt ujra. Csak ez, csak is ez hiányzott,
hogy egy kis forgó, vagy a mint a nemes Samu mondotta: üzleti
tőkéje legyen. Nem mosolyogtak többé rajta a kartársai és mivel a
jólét illata terjedt el körülötte, megbocsájtották neki, hogy
parfumirozta magát és otkolonnal meg teával mosta a haját, hogy
szőkébb legyen. «Ez valaki, mégis csak valaki» – mondta ki reá az
itéletet egy igen elmés régi és intim ellenfele azon az estén, a
melyen először érzett meg rajta a szolid üzleti alap. Mert hajh, az így
van, hogy kinek mije nincs –: annak az imponál. És ha tagadja, ha
csufolja, annál jobban! És e társaságban – a fiatalember virágzó és
terjedő czéhjében – egyszerre felszökött az értéke Asztalos Aurélnak
és azok a kiknek – különösen az anyagi – szoliditás nem
sajátosságuk: szivesen csinálták neki a jó hirt, mert rendezett férfiú
és szolid. És ebben aligha vezette őket anyagi érdek, hanem
alkalmasint a realizmusra törő eszményiség, mint a hogy Asztalos
Aurél a publiczista irná, a kinek, – mint annyi társának – e két szó s
a vele való játék ott volt mindig a tolla alatt.
Kivonatban ugyan, de kiadták a programmbeszédjét és bár csak
telefonon, nem személyesen kérte, jó sorokat is irtak hozzá. A saját
lapja természetesen hasábokat; mert ki becsülje meg az embert, ha
nem saját maga?!… Különben ez estvén nemcsak ellenfelei, de
barátjai is azt mondták reá: «Ez valaki!» Előleget nem kért, csak
jelentette, hogy nehány hétre elmegy külföldre, majd visszajön. A
lapot ő azért szereti és mihelyt ráér… Hogy repült volna ki, ha nem
érzik meg rajta, hogy létének van anyagi alapja és ha ma kiejtenék a
lap kötelékéből, holnap, sőt holnapután sem szorulna a pinczérek
jóakaratára, nem lenne szerencsétlenebb mint a faczér boltos segéd,
a ki legalább részt vehet tüntető körmenetekben, kirakatokat zuzhat
és ha ma letör, holnap visszatörhet.
Egyszerre mintha a munkája is becsesebb lett volna. És az
ujságiró – dicséretére legyen mondva, mosolygott az orra alatt és
megismerte, hogy ez az úgynevezett «poziczió». «Ezt ugyan eddig
szépen megcsináltam», gondolta magában, sőt a nemes Samunak, a
ki Gasteinba menet a vasuthoz kikisérte, el is mondta:
– Samu, a helyzetem oly jó, hogy már jobb nem is lehet. Fölül
vagyok, egészen fölül. Poziczióm van, még pedig ritka poziczió!
A felhuzott ernyőjű kocsiban ölelgette ifjú pajtását, bárha kissé
idegenkedett tőle, mert ez testét és külsejét nem részesítette a kellő
ápolásban. De most ezt is elnézte neki:
– Ha nem lennék veled régen te, most mindjárt testut innék
veled. Pedig ha tudnád…!
A fiatalember valószinüleg a jövőre czélzott vagy a közeli multra
utalt? Samu nagyon szeretett volna tudni valamit, de azért nem volt
tolakodó. Szeretett volna ugyan némely praktikus kérdésekbe
belemenni; de látván, hogy Aurél erre gőgös lesz és hideg – hagyta.
– Csak csináld. Úgy csináljad, a mint te akarod!
A tervszerüség, hogy valamit akar – kissé zavarta a fiatalembert.
Még mindig szeretett volna biztosítani magának egy kis
becsületességet is, főképpen Nuszi miatt, kinek ajkára és
elismerésére olyannyira vágyott, a mennyire ő már vágyakozni
tudott. És a szerelemnek néhány ismérvét fedezvén fel szivében:
szinte boldog volt, szinte hitt magában. «Eh, majd leszek én még
becsületes ember is, csak módom legyen hozzá!» Ebben és így
bizakodott, de a míg Gasteinba ért, nem erre vonatkozó terveken
törte az eszét. Azt komponálgatta ki, hogyan kell Annieval bánnia és
mit kell cselekednie, hogy Juli idő előtt meg ne tudjon semmit.
Valami összeütközés félét szimatolt a levegőben és ezért, mint nagy
elővigyázó, elhatározta, hogy nem fognak levelezni és a találkozások
tekintetében minél óvatosabbak lesznek.
Mégis a mint megérkezett, volt annyi emberi érzése, hogy
legelsőbben is a missnek írt; megérkezését tudatta vele. Egy félóra
mulva már meg volt a találka, egy szinte nyilt és fölöttébb
veszedelmes helyen. A leány már várta; fehéren, a boldogságtól
szinte betegen.
– Ide jönni, ilyen vakmerőség. Menjen haza! – mondá a férfi és
ezuttal csakugyan nem tetette magát.
– Miért? – kérdé a leány szomorúan.
– Mert nem engedem, hogy kompromittálja magát.
– Jól van szerelmem – mondá a leány, mert okos volt –
elmegyek. Csak egy csókot kapok. És még meg kell mondanom
valamit: a kis bátyám megjött és engedd meg, hogy azt is
szeressem!
– A testvérét? Hogyne!
A leány könnyes szemmel nézett föl reá. És ő megengedte, hogy
a testvérét is szerethesse.
Egy elaggott, úgyszólván összesorvadt erdő tisztásán – egy
helyecskén, a hol fa és fű valamely ismeretlen okból kikopott –
leültek mind a hárman. A miss, a mint illik, elfordult tőlük és az
avarral játszott. A háta azonban, mintha látta, érezte és élvezte
volna a mögötte lefolyt tilos és titkos idillt: meg-megrezzent, sőt a
mikor a csók megtörtént, meg is rándult.
– Úgy-e nem haragszol? A kis bátyám nekem mindjárt a jó Isten
után következett, eddig. Perfid nem lehetek hozzá.

– Ugy-e, nem haragszol?

– Hol volt eddig?


– Berlinben. Két éve nem láttam; de mindennap irtunk
egymásnak.
– Nem szimpatikusak nekem azok, a kik Berlinben tanulnak. A
protestantizmus rideg dolog. Nincs benne költészet. Osztozni sziven
nem szeretek! – mondá a férfi komolyan, csaknem rosszkedvüen.
– Meg fogok neki mondani mindent. Meglásd, ő velünk lesz.
Beavatjuk!
– Ezt pedig megtiltom, – mondá a fiatalember és elhalványodott.
– Nekem nem lehet megtiltani semmit! – szólt a leány harag
nélkül, de elvörösödve.
Elhallgattak. Némán, de valósággal összepöröltek. A meghasonlás
azonban nem tarthatott itt sokáig; kevés idejük volt és mindketten
hetek óta készültek és szomjuhoztak erre a találkára. A férfi
megcsókolta a leány szemét:
– Szépen megkérem magát… Nem mondhatom még el mind a
titkomat… Kell, hogy még egy darabig senki se tudjon semmiről.
Másokat döntenénk szerencsétlenségbe.
Ez az érv, a melyet nem értett, meggyőzte Anniet. Megfogadta,
hogy nem szól, ellenben kitervelte, mint fogja bátyját és szerelmesét
összebarátkoztatni.
– Milyen ember? – kérdé az ujságiró.
– Az első maga után! – felelte a leány.
– Komoly?
– Nagyon is. Semmi sem jár az eszében, csak a kötelesség. Még
minden iskolában az első volt. Berlinben technikát és jogot tanult.
Minden reggel imádkozik. Aztán egy óráig czélba lő.
– Gondoltam. Persze nagy vadász.
– Nem annyira, mint mások. Szeret olvasni régi könyveket és a
mit olvas, mindenhez jegyzeteket csinál. Gyönyörű gondolatai
vannak. Nekem egy egész tékát teleírt. Este az evangéliumból olvas
föl nagymamának. Olyan szelid, mint egy baby.
– Talán beteges volt, elmaradt?
– Engem a tenyerére állít, úgy emel föl a magasba.
– Nem szeretem az ilyen tökéletes férfiakat, rendszerint rossz
szivüek, – mondá a fiatalember kissé elfogódottan.
– Különben meglássuk. Azt hiszem, össze fogok vele barátkozni!
– tette hozzá békítő szándékkal. De valahogy, mintha a szivét
elvesztette volna: ideges lett, siettette a hazatérést. Felszólította a
misst, hogy menjenek és máskor legyenek óvatosabbak. És egyedül
engedte őket haza, azzal válván el, hogy majd találkoznak a
tennisnél.
Napokig nem mutatkozott; egész napon a herczegéknél ült. Ő
fenségének valóban kitünő memoriája volt, nagyon jól emlékezett
reá és tüntető szeretettel fogadta, bárha természete ellen volt
minden tüntetés. Csak a vezetéknevét cserélte föl olykor Kovácsra, a
minél mi sem lehet természetesebb. Egyébként elhalmozta – a maga
módja szerint – szeretete minden jelével:
– Sok minden jót hallottam önről. A konduitlisztája a legjobb. Az
igazgató el fogja intézni az ön képviselőségét. Remélem, ön meg
fogja azoknak mondani az igazat. Ideje, hogy vége legyen annak a
bizonyos zsidó wirtsaftnak.
A vallásügyi vita – mely nélkül nem mulhat el e komisz hitetlen
korszakban egy jóravaló szellemi dialogus sem – Julit is érdekelte és
a magas körökkel való sűrű érintkezése és néhány nem sikerült
bevásárlása miatt ő is ellensége volt ama ideges fajtának, a mely
állandóan nyugtalanságra izgatja a világot. Közöttük az ujságiró volt
még a legmérsékeltebb. Nem lehetetlen, hogy ama bizonyos Samu
miatt.
Türelmeseknek kell lennünk, fenség – mondá. Azon felekezetbeli
emberek dolgoznak, mohók, kapzsik, álutakon járók; de ügyesek.
Mindenhová befurakodnak! de építenek és gyűjtenek. Csak hadd
kaparjanak, vakoljanak, építsenek és gyűjtsenek még egy darabig, a
míg nagyon szemtelenek lesznek és akkor…
Egy kézmozdulattal jelezte, hogy akkor mindent el lehet majd
tőlük venni. Hazaszeretetüket, mely nyilván adopczió, nemzeti
érzületüket, a melynek hevessége nyilván renegát vonás. Kilökni
őket az intézményekből, a melyeknek megalkotásához hozzájárultak,
elzárni őket a kulturától, a melynek erősödésében benne van vérük,
minden törekvésük és tradicziójuk: van-e ennél valami egyszerűbb és
igazságosabb? Hogy a mit gyűjtöttek, azt kasszálják tőlük: – ez
pedig egyenesen humor.
A herczeg hangosan fölnevetett, úgy tetszett neki a kézmozdulat
és az, a mit jelentett.
– Ördöngős ficzkó ez a politikus. Ennek még Magyarországon
szobrot emelnek. És ez lesz az első szobor, a melyhez pénzt adok! –
mondá a fenség, miután az ujságiró elment. A mi az idő tekintetében
való elszámítását illeti, ő fensége ebben nem volt hibás; hozzá volt
szokva, hogy mindenki előbb haljon meg, mint ő.
Ezentúl, ha ébren volt, lehetőleg minden idejét lefoglalta a
fiatalembernek. Egyenruhákat kombinált és rajzoltatott vele. Sőt arra
is méltatta, hogy családjáról, gyermekeiről beszélt vele. És nem
ismervén kellőképen a magyar közjogot, megigérte neki, hogy ha
majd egyszer Pestre megy, szól a miniszterelnöknek – vagy a
polgármesternek – hogy a hivatalos lap szerkesztését neki adják át.
Ő legalább majd megmutatja abban a lakosságnak a helyes utat.
Juli, szinte féltékeny volt – mindkettőjükre. De azért nem tartott
semmitől:
– Még kitursz engem! – mondá tréfásan, a fenség délutáni álma
alatt, a mikor mindig együtt voltak. Az asszony paróka és füző
nélkül, visszanyerve normális alakját; folytathatta régen megszakított
életét és boldog volt. Kínozhatta titkos barátját, féltékenykedhetett
reá, szidhatta és ijesztgethette.
– Tudom, te nem becsülsz engem! – mondá. Tudom, abban a
szép buta homlokodban micsoda terv jár. Ki akarsz használni, aztán
megakarsz szökni. A mikor már nagy úr lettél – az segítségemmel –
el akarsz venni valami tejben mosdó comtesset. Barátom, korábban
kelj! Meg se mocczanj, különben repülsz!
A fiatal ember elfordult, méltatlankodott. Az asszony megérintette
vállát és kérdé:
– Mit csinál Annie?
Ettől fogva Annievel kínozta fiatal barátját. Hol tiltotta, egyszer s
mindenkorra letiltotta tőle, hol meg küldte:
– Menj, tudod az összeköttetések miatt. De jól viseld magad!
A fiatal ember meg is jelent a tennisnél, a hol megismerkedett az
Annie bátyjával. A mikor megpillantotta a nagy, nyurga, leányos
teintü, de mészáros kezű, nála egy-két évvel fiatalabb férfiút: nem
egészen jól érezte magát. Az ösztöne azt sugta neki, hogy ez az
ember még kellemetlenséget fog neki okozni. Mit? Hogy? Ennek a
keresésére sem ideje, sem módja nem volt; a fiatal gróf jó erősen
kezet fogott vele és mindjárt bevonta a játékba. Nem
szertartáskodott vele, de nem is tegezte le mindjárt, mint a hogy
szokás egy úri társaságban, mely tudja egymásról, hogy már az apja
sem volt szolga. Ez a tegeződés azonban aligha az egymáshoz való
szoros összetartozandóságot jelenti, mint inkább azt, hogy mások, a
kik a te-re nem születtek, külön veendők, megbélyegzendők, vagy ha
nem, legalább is meg kell éreztetni velök, hogy ők más fajtájúak.
Tudatos összetartás és öntudatos ellenségeskedés nincs e
viszonyban. A «te»-k, ha okuk, módjuk, alkalmuk van rá, minden
született és kifejezett intimitásuk daczára megfojtják, lelövik,
megcsalják, elejtik és semmibe sem veszik egymást. Egy reakcziós
szabadkőművesség – ha ugyan lehetséges az ilyen – jele az egész,
inkább csak egy ártatlan tüntetés a mellett, hogy még sem mozog a
föld. Különösen pedig nem ez a drága, ez a mienk, ez a magyar, a
mely néha-nap ugyan nagyot lendül, aztán pedig mintha
megbánná…
Az ujságíró hideg volt és ott helyben csaknem megrikatta Nuszit,
a ki a játék angol jelszavai mellé vakmerően – de ügyesen – hozzá-
hozzá tett az ő számára egy-egy franczia szót: szeretlek, várlak
holnap az erdőben! A fiatal ember nem reagált, erősen vigyázott
magára és valóságos dedektiv szemmel nézte a fiatal mágnást, hogy
annak mely oldala hozzáférhető. Azt hamarosan megkülönböztette,
hogy valamint a lónak, ennek is elejbe kell kerülni, különben
gyanakszik és védekezik, a közeledőt elrugja magától. Különben
semmi mód nem komplikált ember. Az úgynevezett első eminensek
közül való, azok közül, akik mindig és mindenben korrektek, de ez
nem esik nehezükre, mert szopós koruktól a kötelességre vannak
trainirozva és rendszerint sohasem állíttatnak nehéz föladatok elé.
Ők megtartják igéreteiket, de sohasem igérnek semmit, a mi reájuk
nézve terhes; ők nyiltak, mert nincs fantáziájuk és mert magukon –
és családjukon – kívül nem becsülnek meg élő lényt senkit is, és
ezért nem tartják érdemesnek, alakoskodni. Azonfelül elméjük, mely
úgy szolgálja őket, mint egy kitűnően fizetett és pompásan
disciplinált huszár: bizonyos mindenben, de legfőképen bizonyos
magában. Az emberi méltóságnak e kemény kőből faragott, kissé
esetlen és mindig művészietlen oszlopai nem ritkák a protestáns
mágnások között és ha van közöttük olyan, a ki gránit szívében némi
érzelmességet, hideg fejében a predesztináczió elszántságát is
hordja; az ilyen – mint az Annie bátyja is – megmarad örök első
eminensnek.
Mindezekből az ujságíró inkább csak érzett egyet-mást. A
politikánál fogva igyekezett hozzája közel férkőzni:
– Mikor jön a gróf haza politizálni? Várják!
– Nem vagyok készen a tanulmányaimmal. Egy évre Amerikába
megyek!
– Értékes gondolat. Mikor?
– Egy hónap múlva.
Az ujságíró – belül – elmosolyodott. «Ime, minden véletlen őt
szolgálja. Ez az amerikai út főképp!» Elhatározta nyomban, hogy egy
hónapig nyugton lesz. Csakis arra szorítkozik, hogy Annievel e
viszony végkép meg ne szakadjon. És ezentúl alig jelent meg a
délutáni játékokon. Igaz, ideje is alig volt rá; Bécsbe kellett mennie.
Állítólag azért, hogy az udvari levéltárban valamely munkájához adat
után kutasson, tényleg pedig azért, hogy lakást vegyen, mert Julival
elhatározták, hogy Bécsben is kell laknia, másképen nem lehet.
Juli a kezébe számlálta a lakásért és a berendezésért járó pénzt
és valóban megbocsájthatatlanúl kicsinyes volt e pillanatban s
végkép elfeledkezett arról, hogy egy férfiúra nem mindig megtisztelő
az, ha egy nőtől pénzt kap.
– Csupa új pénz. Vigyázz rá, el ne lumpold. Nem muszáj neked
sem a «Sacher»-ba, sem a «Bristol»-ba szállnod. Jó a «Krausz» is.
Ez a pénz itt külön, lakásra. Tudod milyen lakás kell. A konyhát
nézzed meg, kaparj le egy kis falat, nincs-e…? A bútorokat nem a
belvárosban veszed, arra nincs semmi szükség. Haszontalanságokat
se magadnak, se nekem ne vegyél. Meglephetsz engem egy
varrógéppel, ha egy jót és egy olcsót találsz. Ott varrni fogunk.
(Végre varrhatok.) Ez a szép új százas itt a tied. Meglátom, mennyit
hozol belőle haza.
A férfiú tartotta a markát, mosolygott. Belül sem sírt. Mégis,
valami igen keveset alterálta a helyzet szokatlansága: először kapott
életében pénzt nőtől. Tudta, hogy a megtanult erkölcsi fogalmak
szerint ez szégyen, sőt férfi-bűn. Sőt némely auktorok szerint
aljasságnak is mondatik. «Végre is egy kis pozitiv bűn a sok hamleti
kisérletezés után!» – gondolta és inkább örvendett, mint
szomorkodott. Ám a hogy Juli a pénzt kezelte, a hogy megfogta, a
kezébe tette meg visszavette: ez a modor mégis bosszantotta egy
kissé és elméjében végkép kialakult egy meghatározás: «Ez a Juli
egy közönséges lény!» A miből azonban a herczeg barátnéja nem
csinált titkot:
– Tudd meg barátom, hogy én nem lopom a pénzt. Keserves
véres verejtékem ez nekem. És ne gondoljad, hogy csak úgy rá fogsz
járni a kasszámra… Beosztással kell élni. Végre is, nekem szegény
rokonaim vannak, a kiknek életemben nem adok ugyan egy
rézgarast sem, de halálom után kell, hogy hálásan emlegessenek és
imádkozzanak értem. Vagy eszembe jut és férjhez megyek… Van
apró pénzed vasuti jegyre? Persze nincs. Minden pénzeteket rongy
ruhákra és nyakravalókra költitek. Itt van külön.
Perelt, szemrehányásokat tett; de mindenféle czímeken tömte
pénzzel. Igaz, hogy kis összegekkel. Természetesen a herczegtől is
kapott komissiót, titokban. A fenség istállójából egy pár lovat kellett
átszállíttatni barátnéja istállójába. Gyors járású, erős orosz lovakat,
mert a Juli lovai lassúak és feltűnően fehérek; ismeri Bécsben
minden gyerek, incognitó nem mehet rajtuk sehova. Azonfelül a
fenség egyik fölöttébb tekintélyes rokonához is el kellett mennie.
Arról volt szó, hogy annak a legnagyobb fia, egy rendkívül tettvágyó
herczeg, magyar közjogot akart tanulni, titokban és titkos czélból…
Az ujságíró mindent elvégzett. Megegyezett az öreg herczeggel,
hogy a fiatalt minden héten egy napon át tanítja a magyar jogra és a
hazai viszonyokra, a melyekről apának és fiának csak annyi fogalmuk
volt, hogy az kuszáltabb mint valaha és az országot nemsokára újra
meg kell menteni… A fiatal herczegben különösen nagy jóakarat volt
erre nézve. Kár, hogy a «bajusz» és még egy másik magyar szón
kívül egyetlen egy honi hangot sem ismert, azonban egyéb
nyelveken is csak a legszigorúbb tőmondatokban beszélt. Az
ujságírót ez is hamar megszerette. Különben sem volt exkluziv lélek:
ismeretségük első estélyén elvitte egy fölöttébb zártkörű éjszakai
társaságba, a hol kis karlsbadi czipőkből franczia konyakot ittak, még
pedig sokkal több nő, mint a mennyi férfi.
A bécsi utazás, a Samu biztosan megérkező pénzküldeményei, a
jószágigazgató jelentése, hogy a mandátumért nem is kell – nem
érdemes – lejönni, a szerkesztőség, a mely könyörögve kért egy
ünnepre szánt vezérczikket: az előkelőség biztos és boldog érzésével
töltötték el Asztalos Aurélt.
«A révben vagyok!» – mondá magának (Csaknem mindig leírásra
alkalmatos metaforákban gondolkozott.) Most már csak egy pénzes
állásra kellett törekednie. Elégtételére a mesterségéről sem
feledkezett meg és az járt az eszében, hogy egy mérsékelten
szabadelvű nagy napilapot alapíttasson magának. Mindenesetre
némileg liberális színű legyen, mert a szabadelvűséggel végkép
szakítani zsenáns. Karlsbadból, utókúrára nagypénzű kereskedők és
nagyiparosok jöttek át Gasteinba, a kik hamarosan hozzája
tolakodtak. Faiseur is akadt, egy nagyszabású ágens, a ki üzlete
érdekében képviselőséget is vett magának. Az ujságíróval nyomban
egymásra ismertek és a lapalapítás rohamosan ment előre. Az egyik
polgár csak azt kivánta, hogy a lapnak biztosítva legyen a herczeg
jóakarata. A fenség, miután Juli megértette vele, hogy ez a jóakarat
pénzbe nem kerül, semmiféle kötelezettséggel nem jár, szóval adta
tudtára az egyik alapító úrnak, hogy «bár az ujság neki általában
nem rokonszenves – mert hazudik – a tervezett időszaki közlöny
legnagyobb jóakaratát birja, a mennyiben megfelel hivatásának!»
Minden rendben volt. A megjelenés napját is meghatározták és
az ujságírónak Pestre kellett volna mennie, hogy a szükséges
dolgokat elvégezze, a szerződéseket megkösse. Ámde Julin hirtelen
kitört a féltékenység és barátját halálra kinozta Annie miatt:
– Ismerlek benneteket; tudom, te is ki vagy! Én jó vagyok arra,
hogy kihasználj. De engem nem tisztelsz. Engem nem tisztel senki!
Ő vele persze úgy beszélsz, mint egy istennővel!
– Alig látom. Mindig azzal az utálatos bátyjával van! Különben
sem ad rám semmit. Ezek gőgösebbek, mint maga a fenség!
A nép leánya elvörösödött a haragtól:
– Talán rosszat beszéltek rólam? No megálljatok! Megmutatom,
hogy ide fognak jönni. A leányok kezet fognak nekem csókolni, Annie
is. És együtt foglak látni benneteket. Vigyázz, te meg vigyázz!
A herczeg barátnője, bár nem volt protestáns, nem szokott
tágítani attól, amit kimondott. Másnap a herczeg már üzent az Annie
családjának, hogy látni akarja őket, főként a leányokat. A nagyapa
felelt írásban, nyilván diktandóra. Alázatos, folyamodványszerű
levélben értesítette a herczeget, hogy mind betegek.
Valami kellemetlen szövődés kezdetét szimatolta meg az ujságíró
és szeretett volna elmenekülni egy kis időre; de immár nem volt a
maga ura, nem eresztették. Borús, izgatott, villamossággal terhes
napok következtek és volt egy óra, a mikor a fiatalembert a
legnagyobb csapás környékezte: az, hogy kedvese ott hagyja a
herczeget. Akár akarta, akár nem, a szomszéd szobából hallotta az
erre vonatkozó tárgyalásokat:
– Én megyek el. Bocsásson engem el. Ha nem bocsájt, akkor is
elmegyek. És ha egy ezredet állít ide elém, azon is keresztül megyek.
Ki vagyok én itt? Egy rossz nő. Tisztel engem valaki? Véd engem
valaki? Azért a kis falásért, pár rossz ruháért, komiszabb a
helyzetem, mint a legutolsó cselédnek!
A fenség csitította, pénzzel kinálta; Juli még dühösebb lett:
– Vigye innen a rongyos büdös pénzét! Ha engem a
művészetemnél hagy, az nekem többet jövedelmez, mint a mennyit
a fenség nagynehezen kiizzad. Tudja meg, hogy engem
megkárosított. Engem becsületemben degradált. Velem tisztességes
nő szóba nem áll. És mindez azért van, mert nem tud tiszteletet
szerezni annak, a ki hozzá tartozik. Egy káplár tud a szeretőjének; az
altábornagy nem! Mindenki lenéz, mert a herczeg szeretője vagyok.
Pedig nem is vagyok az. Jó volna, ha az lennék. Egy öreg ember
ápolója vagyok!
Erre az utolsó kijelentésre kinos csönd állott be. A herczeg
köhögni kezdett oly hevesen, hogy Juli egy pillanatra békét kötött,
vízért futott és megitatta. Midőn a fenség magához tért, szigorúan és
emelt hangon szólott:
– Kehrt euch, marrsch!
Az ujságiró visszafojtotta lélegzetét. Juli azonban csak most
eresztette ki a hangját:
– Nekem ugyan kommandirozhat! Akkor megyek el, a mikor
nekem tetszik. Olyan botrányt csinálok, hogy tele lesz vele Bécs,
Pest, de Berlin és Páris is! Úgy se valami nagy már a tekintélyünk. Az
utczán kezdenek már nem köszöngetni. Az emberek kezdenek
okosodni. A legfőbb ideje. A magyar arisztokráczia fuj önökre; a
cseh, az még nem. Inkább beleavatkozik a családi állapotjukba.
A herczeg barátnője valami igen erős szót mondott, de halkan; az
ujságíró nem hallotta. Mindazonáltal a legnagyobb mértékben a
herczeg pártján állott és kedve lett volna a nőt – a ki az ő karrierjét
is veszélyezteti – testileg megfenyíteni. Mit akart? Nem értette, mit
akar? Társadalmi pozicziót? El akarja választani a herczeget a
herczegnőtől és ő akar a felesége lenni? Ez lehetetlenség,
képtelenség, erre még Juli sem vállalkozhatik. A nem erkölcsös férfi
nem értette a szintén nem erkölcsös nőnek társadalmi tisztesség
után való vágyát, mert nem voltak ebben az irányban megfigyelései
és nem látta meg soha, hogy a legelvetemedettebb asszony sem
tudja kiirtani a szívéből az ideált, hogy hű feleség és jó anya
lehessen valaha.
Utóvégre is a fenség engedett. Elismerte az áldozatokat, a melyet
barátnője érte hozott. Megígérte, hogy Bécsben haladéktalanúl
férjhez adja. De neki – az asszonynak is utána kell nézni… Valami
rendes, csinos embert, valami becsületes, igazi férfit, a kit ő – a
fenség – nem volna kénytelen utálni. És most, ritka pillanat, a
herczegnek ötlete támadt: mint férjjelöltet fölállította az ujságírót, a
ki, nevét hallván, elvörösödött, de aligha szeméremből.
– Egy ilyen semmit, egy ilyen senkit! méltatlankodott Juli, majd
félig haragosan, félig nevetve hozzátette:
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and


personal growth!

ebooknice.com

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy