See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/343041537

IoT-based Enterprise Resource Planning: Challenges, Open Issues,

Applications, Architecture, and Future Research Directions

Article  in  Internet of Things · July 2020

DOI: 10.1016/j.iot.2020.100262

CITATIONS READS

2 892

3 authors:

Madjid Tavana Vahid Hajipour

La Salle University Islamic Azad University West Tehran Branch
358 PUBLICATIONS   7,190 CITATIONS    61 PUBLICATIONS   763 CITATIONS   

SEE PROFILE SEE PROFILE

Shahrzad Oveisi
University of Tehran
20 PUBLICATIONS   58 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Humanitarian Supply Chain Network Design View project

Graph Analysis in Social media/ Recommender System View project

All content following this page was uploaded by Shahrzad Oveisi on 28 February 2021.

The user has requested enhancement of the downloaded file.

 Internet of Things 11 (2020) 100262

Contents lists available at ScienceDirect

Internet of Things
journal homepage: www.elsevier.com/locate/iot

IoT-based enterprise resource planning: Challenges, open

issues, applications, architecture, and future research
directions
Madjid Tavana a,b,∗, Vahid Hajipour c,d, Shahrzad Oveisi d
a
Business Systems and Analytics Department, Distinguished Chair of Business Analytics, La Salle University, Philadelphia, USA
b
Business Information Systems Department, Faculty of Business Administration and Economics, University of Paderborn, Paderborn,
Germany
c
Department of Industrial Engineering, Islamic Azad University, West Tehran Branch, Tehran, Iran
d
Research Center, FANAP Co., Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: In today’s highly competitive markets, organizations can create a competitive advantage
Received 30 June 2020 through the successful implementation of Enterprise Resource Planning (ERP) systems. ERP
Revised 13 July 2020
works with different technologies, including the Internet of Things (IoT). IoT uses a unique
Accepted 18 July 2020
Internet protocol to identify, control, and transfer data to individuals as well as databases.
Available online 25 July 2020
The data is collected through IoT, stored on the cloud, and extracted and managed in
Keywords: through ERP. In this study, we review the challenges, open issues, applications, and archi-
Internet of things tecture of the IoT-based ERP. For this purpose, we review and analyze the latest IoT-related
Enterprise resource planning articles to present the unique features of the IoT and discuss its impact on ERP. The results
Cloud computing show sensors and devices connected to the Internet can manage the stored data processed
in the cloud through ERP without human intervention. We also discuss the challenges and
opportunities in the relationship between ERP and the IoT risen by the introduction of the
cloud.
© 2020 Elsevier B.V. All rights reserved.

1. Introduction

Companies in various industries collect different types of data, including invoices, customer information, project sched-
ules, shipping, and payment information, among others [89]. Enterprise Resource Planning (ERP) plays a critical role in
streamlining processes and improving business operations across industries [68]. Today, the Internet of Things (IoT) and
massive data influxes are stirring up the ERP landscape, introducing new possibilities for operational excellence and work-
flow automation [10,27]. IoT is evolving rapidly, and according to McKinsey’s report on the global economic effect of IoT, the
annual economic impact of IoT in 2025 will range from 2.7 to 6.2 trillion dollars [8,98]. However, in the midst of the ex-
citement for increasing connectivity in manufacturing, the impact of IoT has on ERP systems, and business processes should
not be overlooked [12].
Every system component in the IoT has an IP address, which enables the computers to identify, control, transfer data to
other individuals, and related databases [104,105]. The collected data, stored in the cloud, enables ERP to manage, control,

∗
Corresponding author.
E-mail addresses: tavana@lasalle.edu (M. Tavana), hajipour.v@wtiau.ac.ir (V. Hajipour), sh.oveisi@fanap.ir (S. Oveisi).

https://doi.org/10.1016/j.iot.2020.100262
2542-6605/© 2020 Elsevier B.V. All rights reserved.
2 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Table 1
ERP history.

Type of system Period Goal Systems

Reorder point systems The 1960s Available historical data is used to predict the It is used to manage the production of
expected inventory items in the future; when some products in high volume and at
an item reaches a level less than the constant demand; in these systems,
predetermined level, more products are the focus is on cost.
ordered.
Material requirement The 1970s Provides a method based on the need for Focus on marketing; Emphasis on
planning system (MRP) planning the production of products and more Integration and planning
ordering the raw materials
Manufacturing The 1980s Planning the newly added capacity; can be Focus on quality, focusing production
resource planning used for planning and monitoring the strategy on process control, overhead
systems (MRP-II) implementation of production plans. costs reduction, and precise cost
reporting.
Manufacturing The 1990s Enabling the adaptation of production plans Focus on the ability to adapt new
execution systems to meet customer needs; providing more services and products on a timely
(MES) feedback in relation to sales activities basis to meet specific customer needs
Enterprise resource The late 1990s Integrating the production process with Integrates supplier, manufacturer, and
planning (ERP) and beyond supply chain processes in companies; customer information throughout the
designed to integrate enterprise business supply chain.
processes to provide a seamless flow of
information from suppliers
IoT 1999 Everything has a digital identity and allows The term “Internet of Things” was
computers to organize and manage them. coined by Kevin Ashton, executive
director of the Auto-ID Center

and process the data. Overall, the integration of ERP into IoT provides a great number of opportunities, the most important
of which includes better management, automation, product traceability, and reduce ERP implementation costs [133]. IoT
creates a connection between the product and the customer by reducing human intervention and promoting automation
using sensors. However, the integration of IoT into ERP faces a lot of challenges, and the biggest challenge is the lack of
trust in IoT as a relatively new technology trend. Data security is a top concern, and companies heavily investing in ERP
are very wary of the accuracy and reliability of the data coming from an IoT system [98,108]. After almost three decades
since their introduction, ERP systems are so well-oiled that a single record could represent a few million dollars. It is simply
risky to trust new technology, and in a large-scale ERP implementation, nobody wants to save money unless it is millions
of dollars [27,30,62].
In this study, we review the challenges, open issues, applications, and ERP-based IoT architectures. The latest journal
papers at the intersection of ERP and IoT are reviewed and analyzed. We further demonstrate the relationship between ERP
with cloud-based services. In Section 2, we present a brief history of ERP and IoT. In Section 3, we present the technologies
and platforms in cloud ERP. In Section 4, we present the IoT, and in Section 5, we present our conclusions and future
research directions.

2. History of ERP and IoT

Demonstrating a history of ERP systems, Table 1 shows that IoT was introduced in 1999 by Kevin Ashton [5]. The growth
of technology in ERP is also shown in Fig. 1. As can be seen in Fig. 1, the use of Internet technology in ERP has increased
from 20 0 0 with the use of the Internet and extranet technology [70].
The evolution of ERP systems came just after the development of hardware and software systems. During the 1960s, most
organizations designed, developed, and implemented centralized computing systems, most of which automated warehouse
inventory control systems using inventory control (IC) packages. Essential Requirements Planning (ERP) systems were de-
veloped in the 1970s. In these systems, sectoral or product requirements planning is based on a comprehensive production
plan. New software programs, such as product resource planning (MRP II), were introduced, focusing on the synchroniza-
tion of the production process with the production requirements to optimize the production processes. Systems with ERP
systems technologies first appeared in the late 1980s and early 1990s and had the power to coordinate and integrate opera-
tions throughout the company. Using MRP technology, ERP systems achieve the integration between business processes such
as manufacturing, distribution, accounting, finance, human resources, project management, Inventory management, mainte-
nance, and Logistics. This integration also creates accessibility and visibility, as well as unity throughout the company.
Developments in ERP include planning and scheduling (APS), Internet business solutions, including customer relation-
ship management (CRM), and supply chain management (SCM). Fig. 2 shows the major historical events related to ERP.
During this decade, ERP has been associated with mainframe and mini and macro computer technologies, as well as data
warehousing, data mining, knowledge management, and client-server networks with a distributed database and knowledge
management. The first decade of the 20th century also saw ERP working with low-end IoT technologies, mainframes, client-
server systems, distributed computing, Internet technology, and knowledge management. Today, there are many different
systems on the market, most of which are migrating outside the cloud.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 3

The Evolution of ERP

(Enabling Technology)

Real time processing

Batch processing Mainframe and mini
Mainframe computers computers
Traditional file systems Database management
systems(relational)

Inventory MRP MRPII

1960s 1970s 1980s

IRP
Enterprise Suite ERPII ERP
New Trends 2000s 1990s
2020

IOT Main Frame

Internet Internet Client server networks with
Mobile wireless Client server systems distributed data base
WEB Service Architecture Main frames Mini and micro computers
Knowledge management technology(includes Data warehousing
Artificial intelligence intranets and extranets) Knowledge management
Grid computing Distributed computing MRPII
Knowledge management 1980s

Fig. 1. ERP evolution.

Platforms
Application

Interactive Microsft Microsft

Statistical BI/Reporting Interprise IOT SAP oracle
Azure Azure
Epicor SAGE Netsuite PeopleSoft
Web&Mobile
Analysis Adohoc Analysis Application
Computing

Platforms
Block Microsft
SAP Oracle IBM
Chain Azure
Percent
Platforms
SAP Integration Machine Platforms
Microsft
SAP Oracle IBM
Oracle Cloud Learning Azure

Infor
Platforms
Microso
Microsft
AI SAP Oracle IBM
Other Azure

Platforms
Microsft
Big Data SAP Oracle
Human Azure
Source

Social CRM Supply

Sensor resource Finance Sale manage
ERP manageme ment
Media Systems Data manageme
nt System
accounting
nt system
Technology

Fig. 2. Cloud ERP, technologies, and platforms

Note: Some of the data presented in this figure are obtained from 2016 the Report on ERP Systems and Enterprise Software : A Panorama Consulting
Solutions Research Report available at: https://www.panorama- consulting.com/wp- content/uploads/2016/07/2016- ERP- Report- 2.pdf.

The company’s hardware provides the bases for the local installation of the conventional on-site ERP systems, while
cloud ERP does not require such hardware. In other words, it is considered to be a service that works through providers
and programs and is accessible through web applications. Each company can choose from a range of various ERP systems
to deploy, and of course, the choice depends on a number of factors such as the data storage location as well as software
hosting. The changes made to ERP implementation over time led to changes in the way the companies use ERP. Initially,
the systems were small and not very functional, and they were developed to perform a specific task. So companies have
no choice but to install and use various systems to perform all the tasks and provide management that a company needs.
In many cases, the existence of small systems is indicative of the fact that there will be a lot of integration and commu-
nication among these systems. It also means that there is a lot of traffic coming through the network, and there are more
risks involved, that is caused, ERP works with Internet technologies, mobile wireless, artificial intelligence, grid computing,
knowledge management, web service architecture over the years and decades.As can be seen in Fig. 1, the use of Internet
technology in ERP has increased from 20 0 0 with the use of the Internet and Extranet technology, which will reach its peak
in 2020.
4 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Table 2
Advantages and disadvantages of cloud ERP vs. on-premise ERP systems (-, +).

Cloud ERP On-Premise ERP

Cost + • Costs identified over time – • High risk of primary investment

• Inexpensive preliminary investment
• Limited investment in hardware (e.g., server
infrastructure)
– • High property costs over the entire system + • Reduced initial system risk
lifecycle
Security + • Data is stored on vendor servers, and – • Providing the organization with data
security is usually in the vendor’s hand. security may keep the data secure.
• Prospect of continuing stability resulting in • The runtime may be a function of
regular updates from the vendor as little customization.
customization is necessary.
– • While data security is in the hand of + • The data is stored on the customer’s server,
vendors, security is not fully guaranteed. In and security is usually provided by the
addition, prospects may not go well in some orGANization.
orGANizations.
Customization + • Facilitation of customization for its – • Customization may lead to problems after a
flexibility. software update by the vendor
– • There are a few customization options. + • Facilitation of customization
Implementation + • Lower implementation time is needed. – • The implementation time sometimes lasts
longer than usual.
– • Shorter implementation time may be a + • The enterprise exerts a higher level of
function of lower customization options. control.

3. Cloud ERP: technologies and platforms

This section will discuss cloud ERP technologies and platforms. First, the advantages and disadvantages of cloud ERP
versus on-premise ERP are discussed in Table 2; then, different cloud ERP vendors are introduced, and ERP technology
explored [75,89].
As can be seen in Fig. 1, distributed, web-based, artificial intelligence and other technologies have been more frequently
used in recent years. Therefore, this report highlights five IoT technologies, namely blockchain, artificial intelligence, machine
learning, and big data. The following will explore what platforms can work with these technologies. As shown in Fig. 2, the
platforms that can work with IoT include SAP, Oracle, Microsoft, Epicor, Sage, Netsuite, and People Soft.
The platforms that can work with blockchain technology include SAP, Oracle, Microsoft Azure, and IBM. SAP, Oracle,
Microsoft Azure, and IBM are the technologies that can work with machine learning technology and artificial intelligence.
Finally, the platforms capable of working with big data technology are SAP, Oracle, and Microsoft Azure [27].

4. Internet of things (IoT)

The term Internet of Things (IoT) was coined by Kevin Ashton in 1999. He described a world where everything had a
digital identity for itself and enabled computers to organize and manage things. For Kevin Ashton, “IoT” is a tool to overcome
the domination of time and place. This view was first popularized by the Center for Automatic Identification and relevant
market analysis publications [73]. RFID is considered as a prerequisite for this technology. From another perspective, the
IoT is an environment where everything, whether human, animal or inanimate, possesses a unique Internet Protocol (IP)
that has the ability to detect, control, send and transfer data to others and their respective databases. The data collected
from objects will be visible through different tools such as mobile phones and a variety of computers and tablets. When
IoT is implemented, data can be transferred across several objects. IoT is a function of the convergence and evolution of the
three elements: Internet, wireless technology, and microelectromechanical systems. International Telecommunication Union
defines IoT as follows: at any time and place, we will have something to which anyone can connect [13].

4.1. Integration ERP and IoT

Since the semantic web is considered to be an effective method for achieving the integration among systems, a model
for integrating ERP and IoT is introduced here. Model of integration of cloud ERP APIs and things as a service [8]. As shown
in Fig. 3, the IoT includes sensors, actuators, and applications (on mobile or Windows). Accordingly, the sensors collect data
based on the actuator’s act. This information provides a common framework that allows data to be shared and connected
to the cloud through the Semantic Web. The data is also linked to the IoT application via the cloud. ERP also manages and
controls data through the cloud. The integration of IoT and ERP has many benefits, including the following:

1. Events such as re-orders, replenishment, out of stock inventories and missed deliveries could be informed automatically
through sensors and devices connected to the internet.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 5

Cloud ERP API 1

IOT
Services

Semanc Web

Sensors Actuators

API
Cloud ERP
2

Fig. 3. IoT and cloud ERP integration model.

2. IoT makes it possible to send notifications and warning to the manufacturers informing them that, for example, some
of their products require their attention or some machines are down. Nevertheless, such a functionality is provided only
when the business processes are adapted to this new model and be responsive to it.
3. IoT provides a considerable amount of data that needs to be gathered, processed, and analyzed appropriately to gain
the maximum benefits from IoT ERP. The robustness of this software program, however, is a genuine concern. IoT ERP is
expected to deal with the massive influx of data from different devices and products. Thus, a lot of pre-preparation is
demanded of the manufacturing companies while they need to take into account the size of their current ERP program
and whether it is capable of connecting to IoT.
4. Real-time information is a prerequisite for an instant solution. The integration of ERP and IoT paves the way for real-
time data and immediate solutions at the same time; this is possible through ERP, which provides a clear understanding
of the situation while IoT offers the potential solutions. In fact, IoT will create junk information that can be extracted
through ERP in an attempt to improve the business.
5. Through the sensors, this integrated system is able to protect company products from theft. It does this through an alert
message such that if the package is manipulated before it gets to the destination, an alert message or an email is sent
to the registered number or email address of the customer.
6. As for special or sensitive products whose maintenance demands special circumstances such as specific temperature or
pressure, some IoT solutions safeguard these kinds of products against the environmental risks and keep the quality the
same during the shipment.
7. According to the traditional methods, the manufacturer could only check on the quality of a product that has been
shipped to the destination only through some primitive techniques such as customer service calls or field visits. However,
the advent of technologies such as IoT makes it possible for the customer to trace the product’s status and find out about
its use, behavior, wear, and tear. Even some powerful IoT devices demand to replace their parts after they detect wear
and tear.
8. There are two cases for selling the products: first, when the product is sold directly to the customer and second, when
the product is sold through dealers. Of course, adopting the second approach poses a challenge to the ERP system in
terms of keeping track of the end-users’ data. However, by enabling the products to communicate with their customer
immediately after they are initiated, IoT resolves this problem.
9. The automatic scan and entry of the barcode data are achieved through IoT in real-time. In other words, IoT reduces
human intervention to a minimum through the sensors embedded inside the machines leading to improved production
efficiency.
10. In a supply chain, IoT allows tracing the products or equipment as they are transferred from one stage to the next, and as
a result, it provides real-time information. Subsequently, the real-time data is fed into the ERP system providing all those
who are engaged with the supply chain with the feature which enables them to track the interdependencies between
product life cycle and the material flow on the other hand.

4.2. Analysis of published researches in IoT

As we have seen, IoT is one of the key technologies working with ERP. In analysis review sections, according to our inves-
tigation on the IoT field, 140 articles were randomly selected from recent IEEE, ACM, and Elsevier journals from IEEE Explore,
ACM Digital library, Arvix repository, and Elsevier. By reviewing these papers, the journals in this field in our database are
6 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

18

16

14
NUMBER OF PAPERS

12

10

JOURNAL NAME

Fig. 4. The papers published in each journal.

shown in Fig. 4. As shown in this figure, a majority of articles in our database are derived from the IEEE database, IEEE Ac-
cess, ACM database, ACM Transactions on Internet journal, and Future Generation Computer System from Elsevier database.
The IoT review articles are shown in Table 3.

4.2.1. Taxonomy of IoT systems

After analyzing the papers, the following categories were identified based on applications, architectures, technologies,
challenges, and topics to be discussed in the future, which are shown in Fig. 5 and will be discussed in more detail below.

4.2.1.1. IoT system architecture. This section briefly describes the architecture of IoT systems. As can be seen in Fig. 6, IoT
networks include three layers, namely Application, Network, and Perception. Fig. 6 shows the detailed architecture of IoT
[32,35,129]. The following example illustrates how IoT works. Smart parking, including sensors and microcontrollers, is lo-
cated in each parking place. The user receives a live update about the availability of all parking places and chooses the best
one. The ultrasonic sensors are used to relay information and display the available parking places in a web application. The
IoT device consists of an ESP8266 microcontroller and an HC-SR04 distance measurement sensor. The sensor periodically
measures the distance and transmits this data to the microcontroller, which is connected to AWS IoT service via the MQTT
protocol. Periodically, the device sends measurements to the cloud where they are stored in AWS IoT shadow as a sensor
state. A sensor detects a parked car by measuring the distance to the nearest obstacle. We used this example in parts a,b
and c.

a) Physical layer

The physical object surface is involved with physical IoT sensors. The physical object is mainly involved with sensing,
collecting, and sometimes processing information. At this level, diverse sensing functionalities are implemented through
adapting sensors and actuators such as temperature, motion, or acceleration sensors. In the smart parking example, the
ultrasonic sensors are in the physical layer.

b) Connectivity
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 7

Table 3
IoT review articles.

Paper year Brief description

A Review on the Use of Blockchain for the 2018 This paper presents a focus on the way blockchain is adapted to the specific
Internet of Things needs of IoT in an attempt to develop Blockchain-based IoT (BIoT)
applications..
A Survey of Machine and Deep Learning 2018 this paper presents a comprehensive review of machine learning (ML)
Methods for the Internet of Things (IoT) methods as well as the recent advances in the field, which can be deployed
Security in the development of advanced security methods of IoT systems.
A survey on the application of machine 2018 This paper examines how machine learning works in the IoT domain intends
learning for the Internet of Things to shed light on the recent advances in the IoT ML techniques and surveys of
various IoT applications.
Deep Learning for IoT Big Data and Streaming 2018 This paper focuses on Deep Learning (DL), i.e., a class of advanced ML in an
Analytics: A Survey attempt to smooth the learning process in the IoT domain. The authors in
this paper mention the characteristics of IoT data and introduce two
significant IoT data treatments, i.e., IoT big data analytics as well as IoT
streaming data analytics.
Enabling Deep Learning on IoT Devices 2017 This paper looks into the integration of deep learning and low-potheyr IoT
products. IoT-based devices are enabled by DL, and this helps them to render
the unstructured multimedia data and be responsive no only to the
environmental events but also to the users as well as the potheyr
requirements.
Evaluating Critical Security Issues of the IoT 2017 This paper focuses on three major layers of the IoT system model, namely,
World: Present and Future Challenges Perception, Transportation, and Application providing a taxonomic analysis
from these three perspectives. The authors offer, in the end, the most critical
issues as a guide for future research directions..
Feature Reduction Method for Cognition and 2019 This paper explains the potheyr amplifier radio frequency (RF) fingerprinting
Classification of IoT Devices Based on as an example.The paper mainly focuses on studying the RF fingerprint
Artificial Intelligence dimension reduction method while taking into account the uncorrelated or
redundant features as well as the high dimensionality of RF fingerprint
features.
From Cloud Down to Things: An Overview of 2019 This paper examines the role of machine learning plays in IoT, as well as the
Machine Learning on the IoT various functions of machine learning in application data processing and
managerial tasks.
IoT for Smart Healthcare: Technologies, 2017 This paper explains the most recent research on different areas of the model,
Challenges, and Opportunities taking into account their pros and cons and discusses whether the model
can be fulfilling for a wearable IoT healthcare system.
IoT-Based Big Data Storage Systems in Cloud 2016 This paper introduces the research topics and publications which have been
Computing: Perspectives and Challenges conducted in the domain of critical IoT application.
Mobile Phone Computing and the IoT: A 2016 This paper provides a selection of more than 100 major up to date studies in
Survey the field. Based on their domain and area of application such as health,
sports, gaming and etc., and the type of focus, whether it includes
participatory sensing, eco-feedback, actuation, and control, and the involved
communicating agent such as things or people, these papers are categorized
to 10 domains.
The paper also presents an analysis of the open issues and research gaps
after they are identified.
Modeling of Aggregated IoT Traffic and Its 2019 This paper provides an insight into the accuracy of the Poisson
Application to an IoT Cloud approximation model.
Securing the IoT in the Age of Machine 2018 In this paper, a taxonomy and a review of the most recent research on IoT
Learning and Software-defined Networking security are provided while presenting a roadmap of the real research
challenges of employing machine learning and software-defined networking
to focus on the current and future IoT security issues.
Smart Electricity Meter Data Intelligence for 2015 The focus of this paper is on the smart electricity meters along with their
Future Energy Systems: A Survey use while taking into account the significant aspects metering process, the
benefits of the shareholders as well as the technologies dealing with the
interests of the shareholders. The paper put the spotlight on the challenges
as well as the opportunities brought about by the big data and the rising
interest in cloud environments.
Survey of platforms for massive IoT 2018 This paper is devoted to IoT platforms focusing mainly on their components
and features.
Data fusion and IoT for ubiquitous smart 2017 By focusing mainly on the mathematical methods, e.g., probabilistic
environments: A survey methods, artificial intelligence, and theory of belief as well as particular IoT
environments, e.g., distributed, heterogeneous, nonlinear, and object tracking
environments, this paper presents a literature review of IoT data fusion.
Big IoT data analytics: architecture, 2017 This paper proposes a state of the art architecture for big IoT data analytics
opportunities, and open research challenges after the most recent researches on the topic are examined, and the relation
between IoT and big data analytics are explained.
"IoT (IoT): A vision, architectural elements, 2017 This paper discusses the implementation of IoT across the world from the
and future directions, cloud-centric vision perspective.
Secure integration of IoT and cloud computing 2018 In this paper, IoT and cloud computing are examined while concentrating on
the security issues of these two domains.
(continued on next page)
8 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Table 3 (continued)

Paper year Brief description

A survey: Attacks on RPL and 6LoWPAN in IoT 2015 This paper discusses the potential attacks on RPL as well as the 6LoWPAN
network while examing the countermeasure against them and pointing out
the consequences on network parameters.
Program Analysis of Commodity IoT 2019 This paper provides a survey of the privacy and security challenges which
Applications for Security and Privacy: demand program-analysis techniques in IoT focusing on the attacks which
Challenges and Opportunities have been launched against these systems as well as the defense deployed
against them.
Leveraging User-related IoT for Continuous 2019 More than 58 papers on the major components, such as a CA system, are
Authentication: A Survey thoroughly examined in this paper. The paper also studies the industry
status focusing on 32 market contributions, research projects, and related
standards.
Blockchain for the IoT and Industrial IoT: A 2019 This paper starts with examing the underlying structure as well as the key
review features of blockchain and then focus on the security requirement for
Industry 4.0 and IoT development. Finally, the paper investigates the way IoT
security tools and techniques are employed to apply blockchain to IoT for
industry 4.0.
A survey of edge computing-based designs for 2019 In an attempt to motivate edge-based IoT security designs, the present paper
IoT security provides a thorough study of the current IoT security solutions.
Enabling IoT platforms for social IoT 2019 In this paper, after the OS standard features and hardware IoT platforms are
applications: Vision, feature mapping, and explained, an OS-to-hardware architectures features-mapping is proposed
challenges focusing on the investigation of the unique requirements that IoT application
has.
The dual effects of the IoT (IoT): A systematic 2019 This paper explains the advantage and disadvantages of IoT after analyzing
review of the benefits and risks of IoT its adaptation by the organizations. Actually, the advantages and
adoption by orGANizations disadvantages of the IoT are categorized into Big, Open, Linked Data
(BOLD)as a result of a thorough review of the literature.
Privacy preservation in blockchain-based IoT 2019 In this paper, the integration of blockchain technology in IoT scenarios is
systems: Integration issues, prospects, analyzed, emphasizing the privacy challenges resulted from this integration.
challenges, and future research directions
The paradigm of IoT Big Data Analytics in 2019 This paper aims at examining the impacts of IoT Big Data Analytics paradigm
Healthcare Industry: A Review of Scientific (IoT BDA) on the implementation and application of IoT-based technologies
literature and Mapping of Research Trends in the area of healthcare services. A comprehensive review of the literature
in the domain is presented, and the research trends are mapped onto the IoT
BDAin the healthcare industry to achieve this. In fact,46 papers regarding IoT
BDA, as well as 84 papers about fog computing in the healthcare industry,
are reviewed from qualitative and quantitative perspectives in the paper.
Blockchain-based IoT: A Survey 2019 This paper presents a thorough examination of the BCoT by providing
summarized explanations of what IoT is and what challenges it poses. Then,
a brief review of the blockchain is presented to introduce the integration of
blockchain with IoT, an attempt which led to a proposal regarding BCoT
architecture.

One of the main goals of the IoT platform, communicate Heterogeneous sensors cooperatively, and subsequently provide
smart services. Since the sensors deployed in IoT use battery for their power and their computation as well as storage
capacity are limited, they are considered resource-limited. Therefore, IoT sensors should work with low-power sources under
a lossy and noisy communication environment. Recent IoT communication technologies: 6LoWPAN, Bluetooth, IEEE 802.15.4,
WiFi, ultra-wide bandwidth, RFID, and near-field communication. In the smart parking example, the AWS IoT core is the
service that manages the connection with the cloud.

c) Middleware

Middleware aims at demonstrating the complexity of a hardware system, and of course, this means more concentration
on the problems by the developers without being disturbed at the hardware level. Several interfaces can be used to connect
us to sensors and other objects, including Bluetooth, WI-Fi, RFID, and NFC. In a smart parking example, the AWS IoT Platform
is middleware.

d) Hardware-level

Hardware complications are usually caused by related to communication and computation issues. A software level is
provided across operating systems, applications as well as the network communication levels by the middleware, enabling
cooperative processing. From a computational standpoint, a middleware provides a level that sits between application and
system software and has the following main functions. First, it allows for collaboration between heterogeneous IoT objects
so that diverse IoT classes communicate with each other without interruptions through the middleware. Actually, the mid-
dleware plays a significant role in enabling the interoperability among IoT devices. The second major role played by the
middleware concerns the devices, which are potential for interaction within the IoT domain. In fact, it provides these de-
vices with scalability.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 9

Fig. 5. Taxonomy of IoT systems.

Fig. 6. IoT system architecture.

The evolution of IoT devices is expected to be the outcome of middleware through critical reforms being made when
scaling up the organization. The third role is the device discovery and environmental awareness, which has to be provided
by middleware to support object knowledge of other surrounding IoT objects. Middleware is also in charge of providing
the type of computing that is environmentally conscious so that the sensors’ data become understandable, which can be
employed to provide context and present intelligent services to users. Finally, in another role, the middleware addresses
10 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

100
90
80
70
60
50
40
30
20
10
0

Fig. 7. IoT applications.

IoT security as well as its privacy, and that is because of human or industry involvement with the data presented by IoT.
In other words, it is essential to take into account the sensitive issues of security and privacy; moreover, the firmware is
expected to consider mechanisms to create a secure IoT system [6,94].

e) Big data analytics

The noticeable amount of data captured by IoT is highly valuable. Several physical objects existing in the different IoT
applications are in charge of creating such big data. Thus, the large amount of data generated by the physical devices must
be analyzed online to acquire useful knowledge. The timing of knowledge acquiring is useful. According to researchers,
there are various ways to integrate big data analytics with IoT design so that this data becomes understandable. ML and
DL go beyond the traditional analytic methods; in fact, they are capable of absorbing the invisible insight of big data and
transforming it into useful data with a minimal human contribution [6].

4.2.1.2. Applications of IoT in ERP systems. IoT has several applications. The analysis of articles available in our database
identified the applications of IoT shown in Fig. 7 and Table 4 along with the articles in our database (Smart grid, health
care, monitoring, smart home, security, wearable devices, smart city, and energy). The most common applications of IoT
were in Smart City (road management, smart agriculture, smart transportation, smart parking, and smart university), health
care, and security.

a) Smart city

Cities inherently face complex, widespread, and interconnected challenges that can only be solved using a systematic
approach. In other words, the massive crowding of residents leads to chaos, creating conditions that not only bring down
the equilibrium of cities but prevent sustainability with current methods of urban management and development. Therefore,
urban planners all over the world try to develop models of 21st-century urban development, satisfying the new demands
and expectations of today’s world with an integrated view for all aspects of urbanization. The development of a smart city
is a new concept to address the current challenges of cities in the field of urban planning, which has attracted a lot of
attention in recent years. A smart city has been the focus of the millennium transformation and development, which means
opening new concepts in urban planning that combine the capabilities of real and virtual worlds to solve urban problems
[69].
The ERP systems govern rules for purchasing, inventory usage, permitting, and billing. The city finance department tracks
all public expenditures and incoming payments via the ERP system. Public works and city engineering utilize the work order
system within the ERP system to respond to city building issues or infrastructure issues reported by citizens or identified
by the real-time event management and analytics system.
Planning, permitting, and inspections use geographical information systems (GIS) to manage the planning process for
land usage throughout the city. The permitting process utilized with the ERP system allows permits to be requested online,
and then the citizen or business can track the permit and inspection process as it is completed.Permits are issued online via
the ERP permitting web portal. The permit process includes inspections of retention ponds throughout the city. The sensor
network IoT will be utilized to provide water quality information and foliage growth data [105].

b) Smart health care

 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 11

Table 4
IoT applications.

IoT application domain References

Smart Grid& Smart meter Guan [40], Afzal et al. [1,106], Ukil et al. [119], Tabrizi and Pattabiraman [113], Marjani et al. [69],
Rholam and Tabaa [93],

Health care Wu et al. [125], Subahi [110], Baker et al.(2016), Yang [128], Tekeste et al. [115], Sra Amendola
et al.(2014), Rathore et al. [91], Afzal et al. [1], Saheb and Izadi [94], Zhu et al. [137].

Smart Monitoring Jabbar et al. [50], Tekeste et al. [115], Shevtsov et al. [101], Md. Sanwar Hossain et al. [45].

Smart Home (smartwatch, Fang et al. [32], Laput and Harrison [60], Chauhan et al. [22], Tiloca et al. [117], Laput and Harrison
smartphone, smart building) [60], Saeed et al. [140], SIBONI et al. [103], Rholam and Tabaa [93].

Smart Contract Novo [80], Dai et al. [138].

Smart City(road management, Barcelo et al. [13], Metzger et al. [72], Gharaibeh et al. [37], Choi and Choi [24], Perera et al. [86],
smart agriculture, smart Mohammadi et al. [73], Martin Strohbach et al. [109], Furqan Alam et al. [5], Furqan Alam et al. [4],
transportation, smart parking, Chuang Hu et al. [46], Calderoni et al. [141], Paul Brous et al. [15], Bu and Wang [16], Shih et al.
smart university, smart [102], Celesti et al. [142], Vogler et al. [121], Alama et al. [4], Baranwal et al. [143], Marjani et al.
building) [69], Hu et al. [46], Calderoni et al. [141], Rholam and Tabaa [93], Bu et al. [17].

Wearable devices Bennett et al. [144].

Security(Intrusion detection, Mahmudul Hasan et al. [43], Ketheyi Sha et al. [98], Adnan Akbar et al. [2], Xiao et al. [126],
Attack detection, Chauhan et al. [22], Saxena et al. [97], Emami Naeini et al. [31], Chatterjee et al. [21], Tabrizi and
Authentication) Pattabiraman [113], Shen et al. [100], Li et al. [65], Guan [40], Siddiqui et al. [104], Rathee et al.
[90]. Sha et al. [98]

smartphone Kamilaris & Pitsillides [52], Chauhan et al. [22]

Energy [111], Randhawa et al. [88], Pan et al. [83].

Other Rathee et al. [90], Alahakoon and Yu [3], Al-Garadi et al. [6], Ateeq et al. [9], Bertino et al. [14], Bu
et al. [17], Cai et al. [18], Caro and sadr [19], Celik et al. [20], Chen and Hao [23], Chowdhury et al.
[25], Cui et al. [26], Dawe& Paradice [28], Deng et al. [29], Naeini et al. [31], Felemban et al. [33],
Fernández-Caramés and Fraga-Lamas [34], Frustaci et al. [35], Georgakopoulos et al. [36], Gogineni
et al. [38], Gonzalez-Manzano et al. [39], Gubbi et al. [41], Ham et al. [42], Hassan et al. [44], Huang
et al. [47], Huang et al. [48], Islam et al. [49], Junior et al. [51], Karimanzira and Rauschenbach [53],
Khan and Salah [54], Khelifi et al. [55], Ko et al. [56], Kolias et al. [57], Lee et al. [61], Li et al. [63],
Li et al. [64], Li et al. [65], Liu et al. [66], Lv et al. [67], Meddeb et al. [71], Moin et al. [74], Moosavi
et al. [76], Muñuzuri et al. [77], Nawaz et al. [78], Navo [79], Novo [80], Okafor and Delaney [81],
Omitola and Wills [82], Pawlick and Zhu [84], Peña and Fernández [85], Pourvahab and Ekbatanifard
[87], Restuccia et al. [92], Samie et al. [95], Sha et al. [98], Sharma and Wang [99], Siboni et al.
[103], Siddiqui et al. (2018), Siegemund and Turau [105], Soliman et al. [107], Stergiou et al. [108],
Sung and Chiang [112], Tang et al. [114], Tello-Oquendo et al. [116], Truong and Dustdar [118],
Verdouw et al. [120], Wang and Guo [122], Wang et al. [123], Wirtz et al. [124], Wu et al. [125], Xu
et al. [127], Yang et al. [129], Yao et al. [130], Yıldırım and Tatar [131], Yin et al. [132], Zheng et al.
[134], Zhou et al. [135], Zhou et al. [136], Yavuz [145], Mahmud [68], Balampanis et al. [12], Cai
et al. [18], Kumrai et al. [59], Andročec et al. [8], Wanga and Gue [122], AhmadKhan and salah [54],
Chowdhury et al. [25], Junior et al. [139], Gogineni et al. [38], Ko et al. [56], Ahmed Saeed et al.
[140], Konstantinidis et al. [58], Gonzalez-Manzano et al. [39].

The IoT devices have become the heart of key health applications recently, having received a boost from the healthcare
sector. Monitoring the health status of the patients, recording their information, notifying the related healthcare system
of the critical circumstances, and subsequently facilitating the process of treatment for the patient in a timely manner is,
among others, the significant roles of IoT devices in the healthcare systems. With almost 60% of the healthcare sector has
implemented the Internet of medical things (IoMT) devices, the IoMT is believed to be leading to a revolution in the domain
by turning disorganized healthcare into a synchronized one. IoT devices made almost 30.3% of 4.5 billion IoT devices in 2015;
however, the statistic had been predicted to rise to 20–30 billion by 2020 [1,11,69].
ERP, on the other hand, plays a major role in the healthcare sector by achieving an effective integration among the pro-
cesses and the services.ERP would make a big difference in areas such as finance, human resources, and revenue and admis-
sion resources of the healthcare sector if it were implemented successfully. Services such as diagnosis, patient’s home care,
and chronic care are provided by the healthcare system through the information they receive from various sources. Needless
to say, these various sources and processes will be much more helpful for the healthcare system if they are integrated be-
cause integration allows for more information sharing while the traditional models are based on segregated processes and
thus less information sharing. So by achieving integration across various systems in different locations and having access
to them, ERP smoothes the healthcare processes, improve the quality of their services, and help the health fraternity to be
more efficient in their jobs. Implementing ERP in a hospital is particularly beneficial for the front desk as well as the fi-
nancial management sections.ERP provides the front-desk staff with the information they require, such as appointment, bed
availability, specialized services, and the doctors’ schedules leading to the organized and effective performance of the front
desk. The financial management also benefits from ERP because it offers solutions to cut costs, produce more comprehensive
managerial reports, and reduce risks.
12 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Relaon With Other Technologies

30
25

FREQUENCY
20
15
10
5
0

TECHNOLOGIES

Fig. 8. Relation with other technologies.

c) Smart monitoring

IoT monitoring has many advantages, some of which are listed below:

• It helps in analyzing the dynamic systems and processing a large number of events and alerts.
• It also gathers and analyzes IoT data from connected devices and applications and thus create integration between
the devices and business.
• It optimizes the performance across multiple applications, APIs, networks, and protocols and thus bridges the per-
formance gaps
• It provides you with feasible insights to create a better customer experience, mitigate the problems, and increase
the IoT opportunities.

d) Smart home

Human presence detection, activity detection, self-organizing appliances, air conditioning control based on user desire
and convenience, etc. are made possible by machine learning. In particular, the detection of activity due to its use in life
support, home automation, etc., has received much attention [95]. For example, information regarding energy status, includ-
ing its availability and price now or in the future, is provided by energy sources such as photovoltaic appliances and power
suppliers. This information is used by devices which are considered the subscriber in an attempt to modify their patterns.

e) Smart grid and smart meter

Smart grid deals with effective ways to effectively manage and control energy production costs and save energy, which
proves to be more reliable in achieving these goals than the conventional grid. Moreover, smart meters as modern energy
meter makes power consumption measurable and help to monitor and control electrical devices.
The smart grid, which is a flexible system connecting people with technology and natural systems, can be defined from
functional or technological perspectives. It includes an electric grid, a communications network, and controlling and mon-
itoring hardware and software and is capable of providing power, reducing costs, and presenting instant information. For
example, the digital electricity grid which gathers and pass information, and creates electricity using the bilateral technical
direction [69].
In relation to the smart grid and the onslaught of big data, financial transformation is achieved as a result of the de-
mands of business transformation. Since utility companies are encouraging industry-changing advancements of this type,
ERP systems strive to offer strategic insights for the better management and application of the gathered data.

4.2.1.3. Relation with other technologies. Another issue discussed in papers is the relationship of IoT with other technologies.
As can be seen in Fig. 8, machine learning and deep learning technologies make the most use of and have synergy with IoT
technology in reviewed papers. To this end, we also examined a variety of methods used in this field among the articles,
the results of which are shown in Fig. 8.

a) Machine learning and deep learning methods in IoT

Recently, a lot of IoT-based applications have emerged in various vertical fields, namely health, transportation, smart
home, smart city, agriculture, education and so forth most of which are based on intelligent learning mechanism for predic-
tion (i.e., regression, classification, and clustering), data mining, and pattern recognition of data analysis in general. Recently,
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 13

deep learning (DP) has been employed in IoT applications more than other machine learning approaches. According to an an-
nouncement made at the Gartner/ITxpo 2016 Symposium, DL and IoTwere among the top three strategic technology trends
in 2017. This widespread advertising for DL points out the fact that the traditional machine learning approaches do not meet
the newly identified analytical needs in the field of IoT. In other words, according to the hierarchy of IoT data generation
and management, various modern analytical and artificial intelligence (AI) approaches are what IoT systems demand. The
growing interest in IoT and the big data, which have a reciprocal relationship, demands shareholders who clearly understand
its definition, components, potentials, and challenges. On the one hand, IoT is the major generator of big data, and on the
other hand, IoT services are an essential target for analyzing big data to improve processes. In addition, big IoT data analysis
has proven to be valuable to the community. For example, it was reported that identifying the damage to pipes and fixing
them saved nearly one millionUS$ in the water bills of Miami Park Administration [6]. DL models generally bring about two
important advances relative to traditional machine learning approaches in training and prediction stages. Firstly, they reduce
the need for manual, human-made, and engineered feature sets for use in education. As a result, some features that may
not be apparent for men are easily extractable by DL models. Moreover, DL models improve accuracy. Neural networks and
deep learning methods are generally used in ERP-based IoT systems for the following cases:

• Advanced analytics

Machine learning and deep learning methods allow for handling a large amount of data paving the way for real-time and
accurate data insights. As an example, we refer to AI for analyzing different types of customers’ buying behavior, providing
one with the opportunity to adapt products or services to the needs of a specific audience.

• Warehouse management

There are a large number of demand forecasting models which can be tested accurately using machine learning and deep
learning methods to adapt various kinds of variables such as change in demand, supply chain disruptions as well as new
product introduction.BMW, for example, traces an item from the time it is being manufactured until the moment it’s sold
using learning algorithms while monitoring 31 assembly lines in different countries.

• Forecasting

Machine learning and deep learning methods help find solutions that are capable of processing historical data and pre-
dicting the future. In the business world, these methods help identify seasonal patterns and provide suggestions regarding
the production volume.These forecasting methods provide the business world with more accurate predictions in a more
cost-effective way and thus help avoid overproduction or underproduction by guiding through the decisions regarding the
manufacturing volume.

• Financial management

Machine learning and deep learning methods allow the automation of quarterly and monthly processes and the closure
of operations after the accuracy reports are verified, and the account balances are compared. Invoice data can be categorized
into different accounts using these methods, enabling us to differentiate between a phone purchase and a monthly phone
bill.

• Interdepartmental processes

It is complicated to integrate sales, inventory, and accounting. However, Artificial Intelligence is not afraid of massive
amounts and various types of data, making it the perfect solution for creating a centralized platform.

• Customer service

In the case of field services, these methods provide users with the information regarding performance evaluations and
the qualification of the employees so that the planning is facilitated and the service calls are scheduled.

• Production processes

In case of integration with ERP these methods, help to identify the inefficient processes and offer a solution to reduce
costs. Moreover, these methods help to detect energy-consuming processes and facilitate predictive diagnoses and thus con-
tribute to less waste of resources.

• Human resources

ERP software enabled by AI can be proactive, and this is a big advantage because, for example, it can analyze the appli-
cants’ data based on their qualification parameters such as skills and experience, it can find out which employee needs to
be promoted.

• Sales automation
14 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Learning Method
9
8

FREQUENCY
7
6
5
4
3
2
1
0

LEARNING METHOD

Fig. 9. Learning methods.

Chatbots powered by AI help not only customers but also conduct the whole sales triangle. Now, bots can cope with
segmentation and provide responses in real-time. After reviewing our database, the methods used in these articles are as
follows (shown in Fig. 9). Accordingly, SVM, CNN, reinforcement learning, and autoencoder were the most commonly used
tools in the articles. Some of these articles are discussed below;
Another framework operating within the IoT ecosystem was presented by Khelifi et al. [55] for a decision-support sys-
tem (DSS). The proposed framework, which has been compared with three machine learning prediction model classifiers,
i.e., Naïve Bayes, Random Forest, and Decision Tree and empirically evaluated using data sets from a commercial network,
demonstrated efficiency with a complete Bayesian network prediction model. Chen and Hao [23] concentrated on feature
dimension reduction of wireless communication signals and took the power amplifier radio frequency (RF) fingerprinting
as an example of the output of their efforts. The researches conducted on the RF fingerprint feature dimension reduction
method focus mainly on reducing the high dimensionality of RF fingerprint features and the uncorrelated or redundant fea-
tures in the features space. In another paper, Tang et al. [114] describe how CNN inference engines can enable deep-learning
tasks on IoT devices. Ateeq et al. [9] employed historical data from machine learning (ML) and proposed a proactive ar-
chitecture for prediction. Their proposed model named adaptive moving window regression (AMWR) was derived from an
adaptive prediction algorithm and was evaluated using a real-world use case (traffic data provided by the city of Madrid)
with an accuracy of over 96%. Adnan Akbar et al. [2] presented a proactive architecture that exploits historical data using
machine learning (ML) for prediction in conjunction with complex event processing. A semi-supervised deep reinforcement
learning model was introduced by Mohammadi et al. [73], which employs Variational Autoencoders (VAE) as the inference
engine to generalize optimal policies. This model can be deployed in applications related to smart cities using labeled and
unlabeled data to facilitate the performance as well as the accuracy of the learning agent.
Tekeste et al. [115] concentrated on developing technologies such as Machin Learning (ML), cloud Computing as well
as IoT in an attempt to develop a more cost-effective system for water quality monitoring replacing the traditional way
of quality monitoring. Xiao et al. [126], on the other hand, concentrated on the IoT system from different perspectives
such as the attack model, security solutions based on different techniques of machine learning such as supervised learning,
unsupervised learning, and reinforcement learning. Samie et al. [95], Mohammadi [73], Khelifi [55], and Al-Garadi et al.
[6] presented infrastructure toolsets that provide elastic provisioning of application components on resource-constrained
and heterogeneous edge devices in large-scale IoT deployments. After reviewing the papers and results, we have discussed
the machine learning and deep learning approaches in more detail. The advantages, disadvantages, and applications and
references to these methods are presented in Tables 5 and 6.

4.1.3. The challenge related to cloud ERR and IoT

Several challenges have been discussed in the reviewed articles. As mentioned, because IoT systems are associated with
cloud ERP, many challenges, open issues, and future directions are associated with cloud services ([7,15,41,52,53]; Cai et al.
[18]; Samie et al. [95]; [26]).

4.1.3.1.Heterogeneity. A big problem with the cloud ERP-IoT paradigm lies in the extensive heterogeneity of available devices,
operating systems, platforms and services, and possibly recently upgraded applications. cloud platform heterogeneity is not
something to be overlooked. Typically, there is an overlap in ownership of objects for cloud services, allowing for proper
merging and customization of resources according to each provider. This problem is exacerbated when users take advantage
of multiple cloud platforms.
Providers should consider target scenarios, analytic requirements, choice of hardware and software environments, a com-
bination of heterogeneous subsystems, development, and delivery of computing infrastructure, and provision of mainte-
Table 5
Machine learning and deep learning methodsin IoT-based ERP.

Application in
IoT-based ERP/
Method Working principle Advantages Disadvantages methods reference

Decision Tree (DT) • The DT-based approach uses DT to create a • It is actually a clear and simple method. • DT needs a lot of storage space because of its • Intrusion
model (i.e., a forecast model) in which the • Data preparation during the pre-processing structural nature. DT-based approaches are easy detection and
training examples are considered as branches and takes less effort in comparison with other to understand only if few DTs are involved. smart grid
leaves which facilitate the learning process. The algorithms. • The structure of the DT is very sensitive to the • Siryani et al.
new sample class is also predicted using the • No normalization is required for the DT. changes in the data. In fact, if the data changes a [106],
previously-learned model. • No scaling of data is required. little, the structure undergoes a huge change • Hasan et al. [43]
• The loss of values in the data has NO causing instability.
considerable effect on creating a decision tree. • In comparison with other algorithms, the
• It is easy to explain this tree to the team and calculations for a DT may become too
the shareholders. In addition, it is an intuitive complicated.

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

model. • A decision tree often involves a longer time to
train the model.
• Decision tree training is relatively expensive
because of higher complexity and longer time.
• The algorithm of the DT is not accountable for
applying regression and predicting continuous
values.
Support Vector • SVMs create a dividing highway in the feature • SVMs are generalizable and adequate for data • An optimal choice of a core is a difficult task. • Attack detection,
Machine (SVM) dimension of two or more classes so that the with a lot of descriptive features and a few Understanding and interpretation of SVM-based intrusion
distance between the highway and the adjacent sample points. models are also difficult. detection, smart
sample points of each class is maximized. • Hyperline detects optimal segregation. • Its algorithm is not adequate for a large data city, malware
• Maximizes the boundaries between different set. detection, smart
classes. • As for the data with more noise, SVM grid, and sensor
• As for classes with a clear margin of separation, performance is not satisfactory. network
SVMs perform relatively well. • The SVMs underperform when the number of • Yildirim and
• High dimensional spaces contribute to more features for each data point exceeds the number Tatar [131]
efficiency of SVM. of training data samples. • Alama et al. [4].
• When the number of dimensions is bigger than • Since the functionality of the support vector • Chauhan et al.
that of samples, SVM becomes effective. classifier depends in the data points, there is no [22].
• SVM is somehow efficient from the memory probabilistic explanation for the classification • Vogler et al.
perspective. above and below the classifying hyperplane. [121].
• Junior et al.
[139].
Naïve Bayes (NB) • NB calculates posterior probability. NB used • It is fast, simple, and easy to use. • NB independent use of features makes it • smart city, smart
Bayes theorem to predict the probability that a • If the NB conditional independence assumption, impossible for NB to capture useful clues from the grid//
particular feature set of unlabeled specimens fits then it will converge quicker than discriminative relationships and interactions between features. • Alama et al. [4]
into a specific label assuming independence models like logistic regression. • (NB may have a better performance in • Siryani et al.
among features. • Even if the NB assumption doesn’t hold, it applications whose instances have dependent and [106]
works great in practice. related features.).
• It does not require a lot of training data.
• It has high linear scalability with the number of
predictors and data points.
• It is applicable in both binary and multiclass
classification problems.
• It is capable of probabilistic predictions.
•It is capable of handling continuous and discrete
data.
• Irrelevant features do not affect them.

15
(continued on next page)
 16
Table 5 (continued)

Method Working principle Advantages Disadvantages Application in

IoT-based ERP/
methods reference

KNN • The KNN classifies the new sample according to • KNN is a popular and efficient approach to • The optimal value is usually different from one • Attack detection,
the votes of a select number of its closest detect intrusion. dataset to another. Therefore, determining the malware
neighbors. That is, the KNN class identifies • Robust to noisy training data (especially if we optimal value of K can be a challenging and detection//
unknown samples with the majority of votes to use the inverse square of the weighted distance time-consuming process. • Xiao et al. [126]
its nearest neighbors. as the distance). • Needs to specify the value of the k parameter • Alama et al. [4]
• It becomes effective when there is a large (number of nearest neighbors). • Junior et al.
amount of training data. • It is inadequate for distance-based learning [139]
because the type of distance and attribute which
must be used to derive the best results are not

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

specified.
• The high computation cost is necessary to
compute the distance between each query
instance and all training samples. However,
indexing may be helpful in reducing this cost.
Random Forest • In RF, several DTs are built and combined to • RF is extremely resistant to overfitting. It avoids • RF is based on the construction of several DTs. • Hasan et al. [43]
(RF) achieve a well-established prediction model to feature selection and needs only a few input Therefore, it may be impractical for specific
improve overall results. parameters. online programs where the training data sets are
• The overfitting of a single decision tree can be large.
tackled by the process of averaging or combining • A critical problem with the RFs is that they are
the results of different decision trees. very complicated. In comparison with the DT,
• RFs performs much better than a single DT they are more difficult and time-consuming.
when it comes to a huge amount of data items • Because RF includes a lot of decision trees, it is
because they have less variance in comparison less intuitive and demands more computational
with a single DT. resources. This because the large number of the
• They are very flexible and show high accuracy. DTs makes it difficult to understand the
• There is no need to prepare input data because relationship governing the input data.
the data do not need to be scaled. • In comparison with other algorithms, the
• Its accuracy remains the same in case of prediction process becomes longer.
missing a large amount of data.
k-Means Clustering • In k-Means clustering is an unsupervised • Unsupervised algorithms are usually a good • The clustering of k-means is less effective than • Deng et al. [29]
clustering learning approach that identifies clusters of data choice when it is difficult to generate labeled supervised learning methods, especially in
with respect to the similarities of attributes and data. Clustering of K averages can be used to identifying known attacks.
denotes the number of clusters created by the anonymize private data in an IoT system because • Predicting the number of clusters (K-Value) is
algorithm. it does not require labeled data. difficult.
• It is easy to use. • The final results are deeply affected by the
• In comparison with the hierarchical clustering, initial seeds.
K-means may perform faster computations • The final results get affected by order of the
because of the large number of variables it data.
includes (if K is small). • Another disadvantage of the K-Means is that it
• In comparison with the hierarchical clustering, is sensitive to scale. In other words, if you
K-Means may produce higher clusters. normalize, standardize, or rescale your datasets,
• After the centroids are recomputed, an instance your results will change drastically. Although this
can move to another cluster. is not a disadvantage in itself, nit being aware of
the extra time needed to scale the data is
annoying.
(continued on next page)
Table 5 (continued)

Method Working principle Advantages Disadvantages Application in

IoT-based ERP/
methods reference

Principal • NB PCA is a process that converts a number of • PCA can lead to dimensional reduction and • In order to have high efficiency, the PCA, which • Yildirim and
Component possibly correlated features into a smaller mitigation of the model complexity. is a technique for feature reduction, needs to be Tatar [131]
Analysis (PCA) number of unrelated features, called core • It’s very efficient. used in conjunction with the machine learning • Deng et al. [29]

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

components. • It is based on a hierarchy arranged according to methods.
the relevance. • The performance of PCA depends highly on the
• Its performance is excellent for most data sets. non-linearity of the data.
• Results are not always the best for visualization
• Rendering it is difficult
• It is capable of discarding useful information
because it is too focused on the variance, while
sometimes there is not a direct link between the
variance and the predictive power.
SOM • SOM is a kind of artificial neural network • It transforms high dimensional data to • Clustering result depends on the initial weight • Yildirim and
(ANN) which used for dimensionality reduction. It lower-dimensional mao vector Tatar [131]
is trained using unsupervised learning to produce • It contributes to mapping the similar objects • It demands a lot of good training data.
a low-dimensional (typically two-dimensional), onto the locations adjacent to them on the map It is difficult to specify the best map size
discretized representation of the input space of by maintaining the topological properties of the • High computation cost
the training samples named a map. input space.
• It can be efficiently used for visualization.
• It contributes to the topology relations in the
input space.
A multilayer • MLP tends to tackle problems that demand • Guarantee of the possibility of solving of tasks • Low speed of learning • Huang et al. [48]
perceptron (MLP) supervised learning, and it is also employed in • Possibility of overfitting • Chauhan et al.
researches on computational neuroscience and • Impossible to relearning [22]
parallel distributed processing. • Selection of structure needed for solving of a
concrete task I unknown
Bayesian • BN is a probabilistic graphical model that uses • It can readily handle incomplete data sets • It is more complex to construct the graph. • Junior et al.
network(BN) a directed acyclic graph (DAG) to provide a set of • It Allows Ine To Learn About Causal [139]
variables as well as their conditional Relationships.
dependencies. It actually considers an event that • It readily facilitates the use of prior knowledge.
takes place to predict the weather; the
contributing factor is one of the several possible
known causes.
(continued on next page)

17
 18
Table 5 (continued)

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Method Working principle Advantages Disadvantages Application in
IoT-based ERP/
methods reference

ANN • ANN stands for an artificial neuron network • Requires less formal statistical restrictions, able • Big computation burden tends to overfit. • Huang et al. [48]
(ANN) whose underlying bais includes the to model complex nonlinear relationships, able to • Alama et al. [4]
structure and functions of biological neural train multiple algorithms. • Hasan et al. [43]
networks. This computational model is affected
by the information flow through the network
since the changes in a neural network depends on
that input or output.
• Where there are complicated relationships
between inputs and outputs or when patterns are
established, ANNscan is used as nonlinear
statistical data modeling tools.
Regression • Regression is used to estimate the relationships • Linear • Linear • Ateeq et al. [9]
between a dependent variable and one or more • All specifics summarized in one matrix • Doesn’t tell us about a particular parameter •Tekeste et al.
independent variables. • Logistic regression • Mixes up internal and external parameter [115]
• Makes no assumptions about distributions of • More unknowns than the true degree of •Chauhan et al.
classes in feature space freedom [22]
• Easily extended to multiple classes • Logistic regression •Hasan et al. [43]
• Natural probabilistic view of class predictions • Linear decision boundary
• Quick to train
• Resistant to overfitting
Table 6
Potential DL methods for IoT systems.

Application in IoT
based ERP/
Methods Working principle Advantages Disadvantages methods reference

Convolutional • CNNs are mainly used to reduce the data • CNNs are robust, supervised DL methods that • CNNs have high computational cost. Therefore, • Khelifi et al. [55].

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

Neural Networks parameters in dispersed interactions, to share are highly competitive. running them on devices with limited resources • Tang et al. [114].
(CNN) parameters, and to represent the covariances, • The scalability of CNNs has increased with their to support board security systems is challenging. • Li et al. [64]
thereby reducing inter-layer connections to lower new features, and their training time complexity • Mohammadi
values than those found in ANNs. has improved compared to ANNs. et al. [73]
• CNNs have high computational cost. Therefore, • Reduces large data. • Junior et al.
running them on resource-limited devices to [139]
support board security systems is challenging.
Recurrent Neural • The main drawback of RNNs is the issue of • RNNs integrate a time layer to capture •The main drawback of RNNs is the •Intrusion
Networks (RNN) disappearance or explosion of slopes. sequential data and then learn multidimensional disappearance or explosion of slopes. detection and
• Tough settings changes with the recurrent cell hidden unit. •Tough settings prevention
• RNNs and their variants perform well in many mechanism.
applications with sequential data. • Khelifi et al. [55].
• Online processing •Saeed et al. [140].
• Keeping Record Mohammadi et al.
[73]
Deep • AE has a code on its hidden h layer, which • AEs can be useful for feature extraction. Instead • AEs takes long computational hours. Although • Smart grid,
Autoencoders (AE) displays the input. An AE neural network has two of manual engineered features, AEs can be they can learn to effectively record training data smart service,
parts: the encoder function h =, (-), and the effectively used in traditional ML to learn features if the training data set does not smart city//
decoder function, which endeavors to reproduce representation for feature learning reduction of represent the test data set, the AEs may • Mohammadi
input 9: (h). The encoder gets the input and dimensions without prior data knowledge. complicate the learning process rather than et al. [73].
converts it into an abstraction, commonly representing the data set features. [73]
referred to as code. Then, the decoder obtains the
code originally generated for input representation
to reconstruct the original input.
Restricted • The G RBMs are deep production models that • RBMs are deep production models that are • RBMs have high computational cost. Therefore, • Mohammadi
Boltzmann are developed for unsupervised learning. They are developed forunsupervised learning their implementation of IoT devices has limited et al. [73].
machines (RBM) completely non-directional models in which there • They are completely non-directional models in resources to support challenging board security
is no relationship between two nodes in a layer. which there is no relationship between two systems.
nodes in a layer.
(continued on next page)

19
 20
Table 6 (continued)

Methods Working principle Advantages Disadvantages Application in IoT

based ERP/
methods reference

M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

DBN • The G R DBNs consist of stacked RBMs that • DBNs are unsupervised learning methods that • RBMs have high computational cost. Therefore, • Mohammadi
perform greedy layer training to perform robust are trained to display remarkable features with their implementation of IoT devices has limited et al. [73]
performance in an unsupervised environment unlabeled data resources to support challenging board security
systems.
Generative • The GAN framework simultaneously teaches • DBNs are In GANs, sample generation requires • DB GAN training is unstable and difficult. • Mohammadi
adversarial two models (i.e., production and discriminative only a single pass through the model, unlike Learning to generate discrete data using GAN is a et al. [73]
networks (GAN) models) through the opposite process. The DBNs and RBMs where an unknown number of difficult task
production model learns the data distribution and Markov chain repetition is required
produces the data sample, and the discriminative
model predicts the probability that a sample
originated from the training dataset rather than
the production model (i.e., evaluates the case for
accuracy).
Reinforcement • It differs from supervised learning in that • RL solve the complicated problems which • RL cannot be used as a framework, and this is a • Khelifi et al. [55]
learning (RL) labeled input/output pairs need not be presented, remained unsolved by conventional techniques. useful point in this technique. • Yao et al. [130]
and sub-optimal actions need not be explicitly • RL is used mostly for long term results, which • States overload caused by too much RL may • Mohammadi
corrected. Instead, the focus is finding a balance are difficult to fulfill. result in diminishing the results. et al. [73]
between exploration (of uncharted territory) and • It can be considered an ideal model because of • RL is not recommended for simple problems. • Xiao et al. [126]
exploitation (of current knowledge) its similarity to the learning of human beings. • A lot of data and computations are required for • Chowdhury et al.
• The errors of the training process can be RL. Actually, it is a data-hungry model, and this is [25].
corrected by this model. what makes this model excellent for video games. • Bu and Wang
• After an error is corrected, it is very unlikely to When the game is played over and over again, a [16].
encounter the same error again. lot of data are produced. • Xiao et al. [126]
• RL is capable of creating an ideal model for • According to RL, the world is assumed to be
solving a specific problem. Markovian. However, in reality, this is not the
• Robots can learn to walk by implementing RL case, because according to the Markovian model,a
algorithms. sequence of possible events is explained, and the
probability of each event depends only on the
state attained in the previous event.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 21

nance services for each service or application. On the other hand, thanks to the presentation paradigms of cloud services,
it is hoped that the IoT paradigm will easily end up in IoT service provision, although the implementation requires solving
the big problem of heterogeneity. For example, it is necessary to appropriately combine (and manage) a large number of
highly heterogeneous objects (and their correspondingly generated data) across different levels of the cloud platform. This
problem encompasses a variety of dimensions, and its solutions are under consideration based on integrated platforms and
middleware, interactive programming interfaces, duplication tools in the presence of data variety, and so forth [5].

4.1.3.2. Functionality. In this section, we focus on the challenges related to Functionality discussed in the reviewed articles:

a) Performance

Most cases of cloud ERP-IoT paradigm applications provide specific functionality and quality of service (QoS) require-
ments (for example, communication and computing and storage aspects) and, in some scenarios, meet requirements that
are not easily accessible. Specifically, achieving stable and acceptable network performance for the cloud paradigm is a ma-
jor problem as the increasing bandwidth does not lead to an evolution in storage and computing since timeliness is greatly
influenced by unpredictable problems, real-time applications are essentially vulnerable to performance bottlenecks.

b) Latency (Delay)

The term latency, also known as turnaround time, in the context of mobile cloud computing, means the time when the
computation is offloaded, and the results are received from the nearby infrastructure or cloud.

c) Energy efficiency

Energy efficiency refers to a way through which the growth of energy consumption can be managed and prevented. By
delivering more services for the same energy input or for the same services for less energy input, maybe something more
energy efficient.

d) Resource limitations

Since the IoT architecture is resource-constrained, it has been difficult to define a robust security mechanism for it, and
thus in order to work within such constraints, the algorithms have to be limited. For a successful implementation of IoT
security and communication protocols, the storage and energy requirements need to be managed with each broadcast or
multicast necessary for keys or certificates exchange. This means that these protocols should be lightweight and energy-
efficient despite the complicated computations and improved energy harvesting techniques.

e) Cost-effectiveness

A major challenge for the SIoT application is to provide an efficient and profitable service development for energy and
computing costs. A major challenge for the SIoT application is to provide an efficient and profitable service development for
energy and computing costs. Multiple processes can run simultaneously, which may change as per variation in applications
and user requirements.

f) Reliability

In the field of intelligent transportation, cars are usually on the move most of the time, and vehicle networking and
communications are often disrupted or unreliable. When using applications in limited-resource environments, there are a
few problems with a crash or lack of sustained availability of the device. On the other hand, cloud capabilities help overcome
some of these problems; for instance, cloud technology improves the reliability of devices by making it possible to reduce
the burden of heavy tasks and thereby increase battery life or enable modular architecture; alternatively, the technology
itself has uncertainties regarding data center virtualization or resource depletion.

4.1.3.2. Large scale. New applications can be designed using the cloud ERP-IoT paradigm to integrate and analyze the infor-
mation received from (embedded) real-world tools. Some of the presented scenarios implicitly require interaction with a
large number of these tools, which are usually scattered across large environments. The large scale of the mentioned sys-
tems makes it more difficult to solve common problems (for example, when dealing with long-life data that are collected
at high speeds, the requirements of storage capacity and computing capability will be increasingly difficult to meet). In
addition, the distribution of IoT tools complicates monitoring tasks due to dynamic time lag and connectivity issues.

4.1.3.3. Big data. Special attention must be paid to the transfer, storage, and access and processing of the enormous amount
of data that will be generated given 50 billion devices estimated to be connected to the network by 2020. Thanks to re-
cent advances in technology, IoT will be a major source of big data, and our cloud system enables us to store and analyze
complex data for long periods and perform complex analytics on them. The ubiquity of mobile devices and sensors in-
evitably requires scalable computing platforms. Convenient handling of this volume of data is a sensitive issue since the
22 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

overall performance of applications depends heavily on the features of data management services. For instance, cloud-based
approaches to summarizing big data based on extracting semantic features are under investigation. Hence, following the
NoSQL movement, both proprietary and open-source solutions use alternative database technologies for big data, which in-
cludes: time series, key-value, key-value repositories, large column repositories, and graph databases. Unfortunately, there is
no excellent data management solution for the cloud system to handle big data. Moreover, data integrity is considered an
important factor not only because of its impact on service quality but due to the security and privacy dimensions that are
particularly relevant to outsourced data.

4.1.3.4. Sensor network. Sensor networks have been defined as the major enablers of IoT and one of the five technologies that
are shaping our world, giving us the ability to measure, understand and perceive environmental signals; recent advances in
technology have allowed for efficient, inexpensive, and low-cost small tools to be used for remote sensing applications on a
large scale. Furthermore, despite their limitations in battery consumption and reliability, smartphones have multiple sensors
(GPS, accelerator, digital compass, microphone, and camera) that support a wide range of mobile applications in various
areas of IoT. In this context, the timely processing of massively transmitted sensor data, which are vulnerable to power
and network constraints, is a newly introduced problem. The cloud system offers new opportunities to collect sensor data
and use them for widespread coverage and relevance, but at the same time overshadows the privacy and security aspects.
In addition, although IoT devices are not mobile, which is considered a common dimension, the sensitivity of sensors in
smartphones and wearable electronics devices is still a problem.

4.1.3.5. Monitoring. It has been widely documented in resources that surveillance is an essential activity in cloud environ-
ments for capacity planning, resource management, service level agreements, performance, and security as well as trou-
bleshooting. As a result, the cloud ERP-IoT paradigm has the same regulatory requirements as the cloud system; however,
the problems associated with this paradigm are also sensitive to the properties of the volume, variety, and speed of IoT.

4.1.3.6. Fog computing. Fog computing is part of the classic cloud computing operating on the periphery of the network. It
is a system designed to support IoT applications with features such as time delay constraints and requirements concerning
mobility and geographic distribution. Although computing, storage, and networking of resources are both cloud and fog sys-
tems, the latter has certain features as follows: being on the edge of the network and aware of the time-delayed geographic
location; geographic distribution and a large number of nodes compared to concentrated cloud system; support for object
mobility (via wireless communication) and real-time interaction (rather than batch processing); support for interaction. It is
difficult to design cloud computing projects in multiple forms. Certainly, adopting cloud-based approaches requires a variety
of specific algorithms and methodologies that examine the reliability of smart device networks and operate under specific
conditions demanding error-resolving methods.

4.1.3.7. Security. This section discusses the security challenges discussed in the articles [92,96].
• Security and trust management

When it comes to the widespread adaptation of technology, security is always a critical issue. It is essential to estab-
lish a trust to strengthen communication among two or more IoT devices. Efficient authentication and trust management
tasks in different SIoT application scenarios need to be ensured through flexible and strong security protocols within OS,
each of which may require a specific kind of encryption algorithm. However, for implementing these algorithms, they need
to be computationally less intensive; they need to operate in real-time to prevent performance degradation, and finally,
they should take into account the memory and latency requirements. Correspondingly, the memory handling can also raise
security concerns.
• Privacy

Data privacy is of great significance since it poses a major challenge preventing consumers from adopting cloud comput-
ing. The consumers’ trust must be established to meet this challenge, and this can be achieved when the application models
cover application development with privacy protection and implicit authentication mechanisms [69].

4.1.3.8. Complex environment. A complicated IoT ecosystem is made of various elements such as software programs, hard-
ware devices, different connected objects, and, most importantly, a large number of end-users. Some practical applications
are built based on this ecosystem, demonstrating its influence. However, real-time IoT applications mostly demand a large
amount of data, which requires storage and processing resources to extract intelligent information, and this is almost an
uncharted territory in the existing IoT literature.

4.1.3.9. Real-time processing. Since a considerable number of requests from various types of smart applications along with
fast responses are managed in the IoT ecosystem simultaneously, compared with the cloud-based applications, IoT appli-
cations impose multiple requirements regarding the resources for real-time and online processing. This turns the spotlight
on the resource management of such a complicated ecosystem and highlights the critical issue of multiple orders which
demands brand new solutions.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 23

4.1.4. Open issues and future direction

This section discusses open issues and future direction. As mentioned, many issues are related to these three areas due
to ERP’s relationship with cloud and IoT services [95].

4.1.4.1. Open issues. This section focuses on the open issues discussed in the articles:

a) Standardization

The lack of standards has been considered by many researchers as a major issue in cloud IoT. Most objects are now
connected to the cloud via web-based interfaces, which have the potential to reduce complexity in the development of such
applications. Similarly, both clouds and objects are implemented using non-standard heterogeneous interfaces. Although the
scientific community has contributed to the deployment and standardization of IoT and cloud phenomena, the need for stan-
dard protocols, architectures, and API software interfaces is evident to facilitate the interconnection between heterogeneous
smart objects and the creation of advanced services realizing the phenomenon of cloud IoT.

b) Big data

cloud IoT involves managing and processing large amounts of data and events from different situations and heterogeneous
source types so that most applications require complex real-time operations. On the other hand, this means proper syn-
chronization of events from remote sources, reconstructing and correlating their semantics for meaningful deduction of the
status of a particular application. Alternatively, this signifies processing a lot of multimedia data in real-time for the timely
extraction of the information needed to set up affiliate services and assist the user in their current state.

c) Security and privacy

Security and privacy are considered both as a research challenge attracting a lot of attention and an open issue that still
needs further work. Although many users are now worried about privacy and security in cloud-based applications, such
concerns are all the more important as cloudIoTimports real-world data into the cloud and trigger real-world actions. With
respect to privacy, providing the right roles with appropriate design and security is still a challenge, while clearly ensuring
that only authorized people have access to sensitive data, especially when data integrity must be guaranteed in response to
permitted changes.
With regard to security, it remains challenging to deal with various hacker threats since malicious software can be in-
jected into physical sensors to provide manipulated data; raw or processed data can be stolen or tampered with on cloud;
compromised gateways can violate cloud IoT system security; communication channels are vulnerable to leakage of side-
channel information.

d) Intelligence

Focusing on real-time data coming from inhomogeneous objects enhances increased decision-making abilities using in-
formation combination and selection mechanisms. Although there have been researching efforts in this area, maximizing
intelligence in this field is still an obvious challenge.

e) Integration methodology

As long as cloud IoT solutions are built around specific applications, little effort has been devoted to developing a com-
mon methodology for the integration of cloud systems and IoT. Because the suite of applications has defined requirements,
standard multi-tasking can be described. Moreover, a flexible, public platform can be a starting point for easier implemen-
tation of such tasks.

f) Network communications

cloud IoT includes several heterogeneous network technologies where many applications require continuous data transfer,
and overall bandwidth consumption is dramatically increased. On the one hand, the efficiency of access management for en-
abling continuity and optimizing bandwidth utilization is still an open issue. Alternatively, the current bandwidth constraint
cannot support this increasing trend, and further research work is needed to refine large-scale allocation methods.

g) Storage

Storage solutions have been discussed in several articles. For instance, we have already considered them as an outcome
for the integration of cloud and IoT. Nevertheless, articles continue to address this issue as an open one because current
solutions may not provide the support needed for future applications. While data must receive time labels to enable server-
side reconstruction and processing, the transfers need proper scheduling to avoid over-networking and load processing. A
likely orientation to address such a topic involves the introduction of prophetic storage and concealment.
24 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

5. Conclusions and future research directions

The organizations’ awareness of the changes in ERP plays a significant role in customer satisfaction. Intelligent devices
that provide organizations with online data about their product, their quality, their transportation, etc. can have a drastic
influence on not only the customer services but also on better management of the whole organization. However, cloud ERP
integrated with IoT is a modern area promising more fulfilling management as well as customer services.
In this article, because of the impact of IoT on cloud ERP quality, our primary focus was on identifying important IoT
features by analyzing articles from recent scholarly and academic journals. We presented a history of ERP and IoT, along
with cloud ERP and IoT technologies and platforms. We further discussed the integrated ERP and IoT and analyzed recently
published research in IoT. We presented a taxonomy of IoT systems and architecture with applications of IoT in ERP systems.
Finally, we presented the challenge related to cloud ERR and IoT and presented future research directions in the integrated
ERP and IoT research.
Further research efforts are expected in several directions to realize the full potential of the cloud IoT [35,41,95]:
• Proper identification, naming, and addressing of objects to support a large number of objects as well as real-time mo-
bility. While IPv6 may be a viable solution, large scale adoption of it, is still an ongoing process. Further research is
needed in specific scenarios to accelerate this slow process (such as access networks) and in dealing with new drivers
and scalable requirements.
• Solutions for detecting environmental changes based on IoT data will enable the delivery of advanced content-based
services. Such an opportunity leads to research on more effective ways to deliver personalized content and advertising.
• Large-scale support is necessary for multi-network scalers (such as multi-homing, multi-path, multi-cost), connection
delivery, and roaming to improve network reliability and ensure continuous connectivity, QoS, redundancy, and error
tolerance. Software-based network solutions are also expected.
• Many cloud-based IoT applications can utilize efficient and flexible mechanisms to create logical discrete network par-
titions without the need for globally distributed network infrastructures, which could be another important driver for
research into network virtualization and software-centric networks.
• Convergence towards a common open-service infrastructure environment provides APLs with the third-party deployment
of cloud-based applications. This convergence will provide new business opportunities and result in research efforts to
define standard protocols, libraries, languages , and methodologies for the cloud computing Internet.

Declaration of Competing Interest

None.

Acknowledgements

The authors would like to thank the anonymous reviewers and the editor for their insightful comments and suggestions.
Dr. Madjid Tavana is grateful for the partial support he received from the Czech Science Foundation (GAˇCR19-13946S) for
this research, and Dr. Vahid Hajipour would like to acknowledge the support he received from the Research Center at FANAP
Co.

References

[1] B. Afzal, M. Umair, G.A. Shah, E. Ahmed, Enabling IoT platforms for social IoT applications: vision, feature mapping, and challenges, Fut. Gener.
Comput. Syst. 92 (2019) 718–731.
[2] A. Akbar, A. Khan, F. Carrez, K. Moessner, Predictive analytics for complex IoT data streams, IEEE Internet Things J. 4 (5) (2017) 1571–1582.
[3] D. Alahakoon, X. Yu, Smart electricity meter data intelligence for future energy systems: a survey, IEEE Trans. Ind. Inform. 12 (1) (2015) 425–436.
[4] F. Alam, R. Mehmood, I. Katib, A. Albeshri, Analysis of eight data mining algorithms for smarter Internet of Things (IoT), Procedia Comput. Sci. 98
(2016) 437–442.
[5] F. Alam, R. Mehmood, I. Katib, N.N. Albogami, A. Albeshri, Data fusion and IoT for smart ubiquitous environments: a survey, IEEE Access 5 (2017)
9533–9554.
[6] M.A. Al-Garadi, A. Mohamed, A. Al-Ali, X. Du, M. Guizani, A survey of machine and deep learning methods for internet of things (IoT) security, 2018.
arXiv preprint arXiv:1807.11023.
[7] S. Amendola, R. Lodato, S. Manzari, C. Occhiuzzi, G. Marrocco, RFID technology for IoT-based personal healthcare in smart spaces, IEEE Internet Things
J. 1 (2) (2014) 144–152.
[8] D. Andročec, R. Picek, M. Mijač, The ontologically based model for the integration of the IoT and cloud ERP services, 8th International Conference on
Cloud Computing and Services Science (CLOSER 2018), 2018.
[9] M. Ateeq, F. Ishmanov, M.K. Afzal, M. Naeem, Multi-parametric analysis of reliability and energy consumption in IoT: a deep learning approach,
Sensors 19 (2) (2019) 309.
[10] D.M. Bahssas, A.M. AlBar, M.R. Hoque, Enterprise resource planning (ERP) systems: design, trends and deployment, Int. Technol. Manag. Rev. 5 (2)
(2015) 72–81.
[11] S.B. Baker, W. Xiang, I. Atkinson, Internet of things for smart healthcare: technologies, challenges, and opportunities, IEEE Access 5 (2017)
26521–26544.
[12] S. Balampanis, S. Sotiriadis, E.G. Petrakis, Internet of things architecture for enhanced living environments, IEEE Cloud Comput. 3 (6) (2016) 28–34.
[13] M. Barcelo, A. Correa, J. Llorca, A.M. Tulino, J.L. Vicario, A. Morell, IoT-cloud service optimization in next generation smart environments, IEEE J. Sel.
Areas Commun. 34 (12) (2016) 4077–4090.
[14] E. Bertino, K.K.R. Choo, D. Georgakopolous, S. Nepal, Internet of things (IoT): smart and secure service delivery, ACM Trans. Internet Technol. 16 (4)
(2016) 22.
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 25

[15] P. Brous, M. Janssen, P. Herder, The dual effects of the Internet of Things (IoT): a systematic review of the benefits and risks of IoT adoption by
organizations, Int. J. Inf. Manag. (2019).
[16] F. Bu, X. Wang, A smart agriculture IoT system based on deep reinforcement learning, Fut. Gener. Comput. Syst. 99 (2019) 500–507.
[17] L. Bu, M. Isakov, M.A. Kinsy, A secure and robust scheme for sharing confidential information in IoT systems, Ad Hoc Netw. 92 (2019) 101762.
[18] H. Cai, B. Xu, L. Jiang, A.V. Vasilakos, IoT-based big data storage systems in cloud computing: perspectives and challenges, IEEE Internet Things J. 4
(1) (2016) 75–87.
[19] F. Caro, R. Sadr, The Internet of Things (IoT) in retail: bridging supply and demand, Bus Horiz 62 (1) (2019) 47–54.
[20] Z.B. Celik, E. Fernandes, E. Pauley, G. Tan, P. McDaniel, Program analysis of commodity IoT applications for security and privacy: challenges and
opportunities, 2018. arXiv preprint arXiv:1809.06962.
[21] U. Chatterjee, R.S. Chakraborty, D. Mukhopadhyay, A PUF-based secure communication protocol for IoT, ACM Trans. Embedd. Comput. Syst. 16 (3)
(2017) 67.
[22] J. Chauhan, J. Rajasegaran, S. Seneviratne, A. Misra, A. Seneviratne, Y. Lee, Performance characterization of deep learning models for breathing-based
authentication on resource-constrained devices, in: Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2, 2018,
p. 158.
[23] X. Chen, X. Hao, Feature reduction method for cognition and classification of iot devices based on artificial intelligence, IEEE Access 7 (2019)
103291–103298.
[24] C. Choi, J. Choi, Ontology-based security context reasoning for power IoT-cloud security service, IEEE Access 7 (2019) 110510–110517.
[25] A. Chowdhury, S.A. Raut, H.S. Narman, DA-DRLS: drift adaptive deep reinforcement learning based scheduling for IoT resource management, J. Netw.
Comput. Appl. 138 (2019) 51–65.
[26] L. Cui, S. Yang, F. Chen, Z. Ming, N. Lu, J. Qin, A survey on application of machine learning for Internet of Things, Int. J. Mach. Learn. Cybern. 9 (8)
(2018) 1399–1417.
[27] D. Snellman, Difference in Cloud ERP Systems – A comparison, Degree Project In Information and Communication Technology, 2017 Second Cycle, 30
Credits Stockholm, Sweden.
[28] S.N. Dawe, D. Paradice, A Systems Approach to Smart City Infrastructure: A Small City Perspective, 2016, p. 11. Dublin, Ireland.
[29] L. Deng, D. Li, X. Yao, D. Cox, H. Wang, Mobile network intrusion detection for IoT system based on transfer learning algorithm, Cluster Comput.
(2018) 1–16.
[30] M.A.A. Elmonem, E.S. Nasr, M.H. Geith, Benefits and challenges of cloud ERP systems–a systematic literature review, Fut. Comput. Inform. J. 1 (1–2)
(2016) 1–9.
[31] P.E. Naeini, M. Degeling, L. Bauer, R. Chow, L.F. Cranor, M.R. Haghighat, H. Patterson, The Influence of friends and experts on privacy decision making
in IoT scenarios, in: Proceedings of the ACM on Human-Computer Interaction, 2, 2018, p. 48.
[32] S. Fang, L. Da Xu, Y. Zhu, J. Ahati, H. Pei, J. Yan, Z. Liu, An integrated system for regional environmental monitoring and management based on
internet of things, IEEE Trans. Ind. Inform. 10 (2) (2014) 1596–1605.
[33] M. Felemban, E. Felemban, J. Kobes, A. Ghafoor, Threat management in data-centric IoT-based collaborative systems, ACM Trans. Internet Technol. 19
(3) (2019) 42.
[34] T.M. Fernández-Caramés, P. Fraga-Lamas, A review on the use of blockchain for the Internet of Things, IEEE Access 6 (2018) 32979–33001.
[35] M. Frustaci, P. Pace, G. Aloi, G. Fortino, Evaluating critical security issues of the IoT world: present and future challenges, IEEE Internet Things J. 5 (4)
(2017) 2483–2495.
[36] D. Georgakopoulos, P.P. Jayaraman, M. Fazia, M. Villari, R. Ranjan, Internet of Things and edge cloud computing roadmap for manufacturing, IEEE
Cloud Comput. (4) (2016) 66–73.
[37] A. Gharaibeh, A. Khreishah, M. Mohammadi, A. Al-Fuqaha, I. Khalil, A. Rayes, Online auction of cloud resources in support of the Internet of Things,
IEEE Internet Things J. 4 (5) (2017) 1583–1596.
[38] S.K. Gogineni, T. Riedelsheimer, R. Stark, Systematic product development methodology for customizable IoT devices, Procedia CIRP 84 (2019)
393–399.
[39] L. Gonzalez-Manzano, J.M.D. Fuentes, A. Ribagorda, Leveraging user-related internet of things for continuous authentication: a survey, ACM Comput.
Surv. (CSUR) 52 (3) (2019) 1–38.
[40] Z. Guan, J. Li, L. Wu, Y. Zhang, J. Wu, X. Du, Achieving efficient and secure data acquisition for cloud-supported internet of things in smart grid, IEEE
Internet Things J. 4 (6) (2017) 1934–1944.
[41] J. Gubbi, R. Buyya, S. Marusic, M. Palaniswami, Internet of Things (IoT): a vision, architectural elements, and future directions, Fut. Gener. Comput.
Syst. 29 (7) (2013) 1645–1660.
[42] S. Ham, J. Lee, K. Lee, Quicktalk: an association-free communication method for IoT devices in proximity, in: Proceedings of the ACM on Interactive,
Mobile, Wearable and Ubiquitous Technologies, 1, 2017, p. 56.
[43] M. Hasan, M.M. Islam, M.I.I. Zarif, M.M.A. Hashem, Attack and anomaly detection in IoT sensors in IoT sites using machine learning approaches,
Internet Things 7 (2019) 10 0 059.
[44] M.U. Hassan, M.H. Rehmani, J. Chen, Privacy preservation in blockchain based IoT systems: integration issues, prospects, challenges, and future re-
search directions, Fut. Gener. Comput. Syst. 97 (2019) 512–529.
[45] M.S. Hossain, M. Rahman, M.T. Sarker, M.E. Haque, A. Jahid, A smart IoT based system for monitoring and controlling the sub-station equipment,
Internet Things 7 (2019) 10 0 085.
[46] C. Hu, W. Bao, D. Wang, Y. Qian, M. Zheng, S. Wang, sTube+: an IoT communication sharing architecture for smart after-sales maintenance in
buildings, ACM Trans. Sensor Netw. 14 (3–4) (2018) 29.
[47] J. Huang, L. Kong, G. Chen, M.Y. Wu, X. Liu, P. Zeng, Towards secure industrial IoT: blockchain system with credit-based consensus mechanism, IEEE
Trans. Ind. Inform. (2019).
[48] X. Huang, S.H. Hong, Y. Li, Hour-ahead price based energy management scheme for industrial facilities, IEEE Trans. Ind. Inform. 13 (6) (2017)
2886–2898.
[49] M.M. Islam, N. Funabiki, R.W. Sudibyo, K.I. Munene, W.C. Kao, A dynamic access-point transmission power minimization method using PI feedback
control in elastic WLAN system for IoT applications, Internet Things 8 (2019) 10 0 089.
[50] W.A. Jabbar, H.K. Shang, S.N. Hamid, A.A. Almohammedi, R.M. Ramli, M.A. Ali, IoT-BBMS: Internet of Things-based baby monitoring system for smart
cradle, IEEE Access 7 (2019) 93791–93805.
[51] C.M.D. Junior, S.P. da Silva, R.V. da Nóbrega, A.C. Barros, A.K. Sangaiah, P.P. Rebouças Filho, V.H.C. de Albuquerque, A new approach for mobile robot
localization based on an online IoT system, Fut. Gener. Comput. Syst. 100 (2019) 859–881.
[52] A. Kamilaris, A. Pitsillides, Mobile phone computing and the internet of things: a survey, IEEE Internet Things J. 3 (6) (2016) 885–898.
[53] D. Karimanzira, T. Rauschenbach, Enhancing aquaponics management with Iot-based predictive analytics for efficient information utilization, Inf.
Process. Agric. (2019).
[54] M.A. Khan, K. Salah, IoT security: review, blockchain solutions, and open challenges, Fut. Gener. Comput. Syst. 82 (2018) 395–411.
[55] H. Khelifi, S. Luo, B. Nour, A. Sellami, H. Moungla, S.H. Ahmed, M. Guizani, Bringing deep learning at the edge of information-centric Internet of
Things, IEEE Commun. Lett. 23 (1) (2018) 52–55.
[56] I.Y. Ko, H.G. Ko, A.J. Molina, J.H. Kwon, SoIoT: toward a user-centric IoT-based service framework, ACM Trans. Internet Technol. 16 (2) (2016) 8.
[57] C. Kolias, G. Kambourakis, A. Stavrou, J. Voas, DDoS in the IoT: mirai and other botnets, Computer (Long Beach Calif) 50 (7) (2017) 80–84.
[58] A. Konstantinidis, P. Irakleous, Z. Georgiou, D. Zeinalipour-Yazti, P.K. Chrysanthis, IoT data prefetching in indoor navigation SOAs, ACM Trans. Internet
Technol. 19 (1) (2018) 10.
26 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262

[59] T. Kumrai, K. Ota, M. Dong, J. Kishigami, D.K. Sung, Multiobjective optimization in cloud brokering systems for connected Internet of Things, IEEE
Internet Things J. 4 (2) (2016) 404–413.
[60] G. Laput, C. Harrison, Exploring the efficacy of sparse, general-purpose sensor constellations for wide-area activity sensing, in: Proceedings of the
ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 3, 2019, p. 55.
[61] S.Y. Lee, S.R. Wi, E. Seo, J.K. Jung, T.M. Chung, ProFiOt: abnormal behavior profiling (ABP) of IoT devices based on a machine learning approach, in:
2017 27th International Telecommunication Networks and Applications Conference (ITNAC), IEEE, 2017, pp. 1–6.
[62] A. Lenart, in: ERP in the Cloud–Benefits and Challenges. In Euro Symposium on Systems Analysis and Design, Springer, Berlin, Heidelberg, 2011,
pp. 39–50.
[63] H. Li, M. Dong, K. Ota, Radio access network virtualization for the social Internet of Things, IEEE Cloud Comput. 2 (6) (2015) 42–50.
[64] H. Li, K. Ota, M. Dong, Learning IoT in edge: deep learning for the Internet of Things with edge computing, IEEE Netw. 32 (1) (2018) 96–101.
[65] J. Li, M. Siddula, X. Cheng, W. Cheng, Z. Tian, Y. Li, Approximate data aggregation in sensor equipped IoT networks, Tsinghua Sci. Technol. 25 (1)
(2019) 44–55.
[66] W. Liu, L. Zhang, Z. Zhang, C. Gu, C. Wang, M. O’neill, F. Lombardi, XOR-based low-cost Reconfigurable PUFs for IoT Security, ACM Trans. Embedd.
Comput. Syst. 18 (3) (2019) 25.
[67] P. Lv, L. Wang, H. Zhu, W. Deng, L. Gu, An IoT-oriented privacy-preserving publish/subscribe model over blockchains, IEEE Access 7 (2019)
41309–41314.
[68] B. Mahmud, Internet of Things (IoT) for manufacturing logistics on SAP ERP applications, J. Telecommun. Electron. Comput. Eng. 9 (2–6) (2017) 43–47.
[69] M. Marjani, F. Nasaruddin, A. Gani, A. Karim, I.A.T. Hashem, A. Siddiqa, I. Yaqoob, Big IoT data analytics: architecture, opportunities, and open research
challenges, IEEE Access 5 (2017) 5247–5261.
[70] Mary Sumner, Pearson New International Edition, British Library Cataloguing-in-Publication Data, 2014.
[71] M. Meddeb, A. Dhraief, A. Belghith, T. Monteil, K. Drira, H. Mathkour, Least fresh first cache replacement policy for NDN-based IoT networks, Pervas.
Mob. Comput. 52 (2019) 60–70.
[72] F. Metzger, T. Hoßfeld, A. Bauer, S. Kounev, P.E. Heegaard, Modeling of aggregated IoT traffic and its application to an IoT cloud, Proc. IEEE 107 (4)
(2019) 679–694.
[73] M. Mohammadi, A. Al-Fuqaha, S. Sorour, M. Guizani, Deep learning for IoT big data and streaming analytics: a survey, IEEE Commun. Surv. Tut. 20
(4) (2018) 2923–2960.
[74] S. Moin, A. Karim, Z. Safdar, K. Safdar, E. Ahmed, M. Imran, Securing IoTs in distributed blockchain: analysis, requirements and open issues, Fut.
Gener. Comput. Syst. 100 (2019) 325–343.
[75] E. Monk, B. Wagner, Concepts in enterprise resource planning, Cengage Learn. (2012).
[76] S.R. Moosavi, T.N. Gia, A.M. Rahmani, E. Nigussie, S. Virtanen, J. Isoaho, H. Tenhunen, SEA: a secure and efficient authentication and authorization
architecture for IoT-based healthcare using smart gateways, Procedia Comput. Sci. 52 (2015) 452–459.
[77] J. Muñuzuri, L. Onieva, P. Cortés, J. Guadix, Using IoT data and applications to improve port-based intermodal supply chains, Comput. Ind. Eng. (2019).
[78] F. Nawaz, N.K. Janjua, O.K. Hussain, PERCEPTUS: predictive complex Event Processing and reasoning For IoT-enabled supply Chain, Knowl.-Based Syst.
(2019).
[79] O. Novo, Blockchain meets IoT: an architecture for scalable access management in IoT, IEEE Internet Things J. 5 (2) (2018) 1184–1195.
[80] O. Novo, Making constrained things reachable: a secure IP-agnostic NAT traversal approach for IoT, ACM Trans. Internet Technol. 19 (1) (2019) 3.
[81] N.U. Okafor, D. Delaney, Considerations for system design in IoT-based autonomous ecological sensing, Procedia Comput. Sci. 155 (2019) 258–267.
[82] T. Omitola, G. Wills, Towards mapping the security challenges of the Internet of Things (IoT) supply chain, Procedia Comput. Sci. 126 (2018) 441–450.
[83] C. Pan, M. Xie, S. Han, Z.H. Mao, J. Hu, Modeling and optimization for self-powered non-volatile IoT edge devices with ultra-low harvesting power,
ACM Trans. Cyber-Phys. Syst. 3 (3) (2019) 32.
[84] J. Pawlick, Q. Zhu, Strategic trust in cloud-enabled cyber-physical systems with an application to glucose control, IEEE Trans. Inf. Forensics Secur. 12
(12) (2017) 2906–2919.
[85] M.A.L. Peña, I.M. Fernández, SAT-IoT: an architectural model for a high-performance fog/edge/cloud IoT platform, in: 2019 IEEE 5th World Forum on
Internet of Things (WF-IoT), IEEE, 2019, pp. 633–638.
[86] C. Perera, C. Liu, R. Ranjan, L. Wang, A.Y. Zomaya, Privacy-knowledge modeling for the internet of things: A look back, Computer 49 (12) (2016)
60–68.
[87] M. Pourvahab, G. Ekbatanifard, An efficient forensics architecture in software-defined networking-IoT using blockchain technology, IEEE Access 7
(2019) 99573–99588.
[88] R.H. Randhawa, A. Hameed, A.N. Mian, Energy efficient cross-layer approach for object security of CoAP for IoT devices, Ad Hoc Netw. 92 (2019)
101761.
[89] M.A. Rashid, L. Hossain, J.D. Patrick, The evolution of ERP systems: a historical perspective, in: Enterprise Resource Planning: Solutions and Manage-
ment, IGI Global, 2002, pp. 35–50.
[90] G. ‫‏‬Rathee, A. Sharma, R. Kumar, R. Iqbal, A secure communicating things network framework for industrial IoT using blockchain technology, Ad Hoc
Netw. (2019) 101933.
[91] M.M. Rathore, A. Paul, A. Ahmad, M. Anisetti, G. Jeon, Hadoop-based intelligent care system (hics): analytical approach for big data in iot, ACM Trans.
Internet Technol. 18 (1) (2017) 8.
[92] F. Restuccia, S. D’Oro, T. Melodia, Securing the internet of things in the age of machine learning and software-defined networking, IEEE Internet
Things J. 5 (6) (2018) 4829–4842.
[93] O. Rholam, M. Tabaa, Smart device for multi-band industrial IoT communications, Procedia Comput. Sci. 155 (2019) 660–665.
[94] T. Saheb, L. Izadi, Paradigm of IoT Big Data Analytics in Healthcare Industry: A Review of Scientific literature and Mapping of Research Trends,
Telematics and Informatics, 2019.
[95] F. Samie, L. Bauer, J. Henkel, From cloud down to things: an overview of machine learning in Internet of Things, IEEE Internet Things J. (2019).
[96] F.K. Santoso, N.C. Vun, 2015 International Symposium on Consumer Electronics (ISCE), IEEE (2015) 1–2.
[97] N. Saxena, S. Grijalva, N.S. Chaudhari, Authentication protocol for an IoT-enabled LTE network, ACM Trans. Internet Technol. 16 (4) (2016) 25.
[98] K. Sha, T.A. Yang, W. Wei, S. Davari, A survey of edge computing based designs for security, Digit. Commun. Netw. (2019).
[99] S.K. Sharma, X. Wang, Live data analytics with collaborative edge and cloud processing in wireless IoT networks, IEEE Access 5 (2017) 4621–4635.
[100] S. Shen, L. Huang, H. Zhou, S. Yu, E. Fan, Q. Cao, Multistage signaling game-based optimal detection strategies for suppressing malware diffusion in
fog-cloud-based IoT networks, IEEE Internet Things J. 5 (2) (2018) 1043–1054.
[101] S. Shevtsov, D. Weyns, M. Maggio, SimCA∗ : a control-theoretic approach to handle uncertainty in self-adaptive systems with guarantees, ACM Trans.
Autonom. Adapt. Syst. 13 (4) (2019) 17.
[102] C.S. Shih, J.J. Chou, N. Reijers, T.W. Kuo, Designing CPS/IoT applications for smart buildings and cities, IET Cyber-Phys. Syst. 1 (1) (2016) 3–12.
[103] S. Siboni, A. Shabtai, N.O. Tippenhauer, J. Lee, Y. Elovici, Advanced security testbed framework for wearable IoT devices, ACM Trans. Internet Technol.
16 (4) (2016) 26.
[104] F. Siddiqui, J. Beley, S. Zeadally, G. Braught, Secure and lightweight communication in heterogeneous IoT environments, Internet Things (2019) 10 0 093.
[105] G. Siegemund, V. Turau, A self-stabilizing publish/subscribe middleware for iot applications, ACM Trans. Cyber-Phys. Syst. 2 (2) (2018) 12.
[106] J. Siryani, B. Tanju, T.J. Eveleigh, A machine learning decision-support system improves the internet of things’ smart meter operations, IEEE Internet
Things J. 4 (4) (2017) 1056–1066.
[107] S. Soliman, M.A. Jaela, A.M. Abotaleb, Y. Hassan, M.A. Abdelghany, A. Abdel-Hamid, H. T&Mostafa, FPGA implementation of dynamically reconfigurable
IoT security module using algorithm hopping, Integration (2019).
 M. Tavana, V. Hajipour and S. Oveisi / Internet of Things 11 (2020) 100262 27

[108] C. Stergiou, K.E. Psannis, B.G. Kim, B. Gupta, Secure integration of IoT and cloud computing, Fut. Gener. Comput. Syst. 78 (2018) 964–975.
[109] M. Strohbach, H. Ziekow, V. Gazis, N. Akiva, Towards a big data analytics framework for IoT and smart city applications, in: Modeling and Processing
for Next-Generation Big-Data Technologies, Cham, Springer, 2015, pp. 257–282.
[110] A.F. Subahi, Edge-based IoT medical record system: requirements, recommendations and conceptual design, IEEE Access 7 (2019) 94150–94159.
[111] M. Sun, Z. Zhou, J. Wang, C. Du, W. Gaaloul, Energy-efficient IoT service composition for concurrent timed applications, Fut. Gener. Comput. Syst.
(2019).
[112] W.T. Sung, Y.C. Chiang, Improved particle swarm optimization algorithm for android medical care IoT using modified parameters, J. Med. Syst. 36 (6)
(2012) 3755–3763.
[113] F.M. Tabrizi, K. Pattabiraman, Design-level and code-level security analysis of IoT devices, ACM Trans. Embed. Comput. Syst. 18 (3) (2019) 20.
[114] J. Tang, D. Sun, S. Liu, J.L. Gaudiot, Enabling deep learning on IoT devices, Computer (Long Beach Calif) 50 (10) (2017) 92–96.
[115] T. Tekeste, H. Saleh, B. Mohammad, M. Ismail, Ultra-low power QRS detection and ECG compression architecture for IoT healthcare devices, IEEE
Trans. Circ. Syst. I 66 (2) (2018) 669–679.
[116] L. Tello-Oquendo, S.C. Lin, I.F. Akyildiz, V. Pla, Software-defined architecture for QoS-aware IoT deployments in 5G systems, Ad Hoc Netw. (2019)
101911.
[117] M. Tiloca, K. Nikitin, S. Raza, Axiom: DTLS-based secure IoT group communication, ACM Trans. Embedd. Comput. Syst. 16 (3) (2017) 66.
[118] H.L. Truong, S. Dustdar, Principles for engineering IoT cloud systems, IEEE Cloud Comput. 2 (2) (2015) 68–76.
[119] A. Ukil, S. Bandyopadhyay, A. Pal, IoT-privacy: to be private or not to be private, in: 2014 IEEE Conference on Computer Communications Workshops
(INFOCOM WKSHPS), IEEE, 2014, pp. 123–124.
[120] C. Verdouw, H. Sundmaeker, B. Tekinerdogan, D. Conzon, T. Montanaro, Architecture framework of IoT-based food and farm systems: a multiple case
study, Comput. Electron. Agric. 165 (2019) 104939.
[121] M. Vögler, J.M. Schleicher, C. Inzinger, S. Dustdar, A scalable framework for provisioning large-scale IoT deployments, ACM Trans. Internet Technol. 16
(2) (2016) 11.
[122] A. Wang, Y. Guo, Multiple-Terminal Aggregation Algorithm for IoT Service, Procedia Comput. Sci. 147 (2019) 495–498.
[123] Q. Wang, X. Zhu, Y. Ni, L. Gu, H. Zhu, Blockchain for the iot and industrial iot: a review, Internet Things (2019) 10 0 081.
[124] B.W. Wirtz, J.C. Weyerer, F.T. Schichtel, An integrative public IoT framework for smart government, Gov. Inf. Q. 36 (2) (2019) 333–345.
[125] T. Wu, J.M. Redouté, M.R. Yuce, A wireless implantable sensor design with subcutaneous energy harvesting for long-term IoT healthcare applications,
IEEE Access 6 (2018) 35801–35808.
[126] L. Xiao, X. Wan, X. Lu, Y. Zhang, D. Wu, IoT security techniques based on machine learning, 2018. arXiv preprint arXiv:1801.06275.
[127] S. Xu, G. Yang, Y. Mu, X. Liu, A secure IoT cloud storage system with fine-grained access control and decryption key exposure resistance, Fut. Gener.
Comput. Syst. 97 (2019) 284–294.
[128] P. Yang, D. Stankevicius, V. Marozas, Z. Deng, E. Liu, A. Lukosevicius, G. Min, Lifelogging data validation model for internet of things enabled person-
alized healthcare, IEEE Trans. Syst. Man Cybern. 48 (1) (2016) 50–64.
[129] Z. Yang, Y. Yue, Y. Yang, Y. Peng, X. Wang, W. Liu, Study and application on the architecture and key technologies for IoT, in: 2011 International
Conference on Multimedia Technology, IEEE, 2011, July, pp. 747–751.
[130] H. Yao, T. Mai, J. Wang, Z. Ji, C. Jiang, Y. Qian, Resource trading in blockchain-based industrial Internet of Things, IEEE Trans. Ind. Inform. (2019).
[131] G. Yıldırım, Y. Tatar, Simplified agent-based resource sharing approach for WSN-WSN interaction in IoT/CPS projects, IEEE Access 6 (2018)
78077–78091.
[132] J. Yin, Z. Yang, H. Cao, T. Liu, Z. Zhou, C. Wu, A survey on bluetooth 5.0 and mesh: new milestones of IoT, ACM Trans. Sensor Netw. 15 (3) (2019) 28.
[133] A.H. Zadeh, B.A. Akinyemi, A. Jeyaraj, H.M. Zolbanin, Cloud ERP systems for small-and-medium enterprises: a case study in the food industry, J. Cases
Inf. Technol. 20 (4) (2018) 53–70.
[134] X. Zheng, P. Martin, K. Brohman, L. Da Xu, Cloud service negotiation in internet of things environment: a mixed approach, IEEE Trans. Ind. Inform. 10
(2) (2014) 1506–1515.
[135] J. Zhou, Z. Cao, X. Dong, A.V. Vasilakos, Security and privacy for cloud-based IoT: challenges, IEEE Commun. Mag. 55 (1) (2017) 26–33.
[136] L. Zhou, S.H. Kondapalli, K. Aono, S. Chakrabartty, Desynchronization of self-powered FN tunneling timers for trust verification of IoT supply-chain,
IEEE Internet Things J. (2019).
[137] H. Zhu, P. Podesva, X. Liu, H. Zhang, T. Teply, Y. Xu, A. Niculescu, IoT PCR for Pandemic Disease Detection and Its Spread Monitoring, 2019. Available
at SSRN 3352008.
[138] H. Dai, Z. Zheng, Y. Zhang, Blockchain for Internet of Things: A Survey, IEEE Internet Things 6 (5) (2019).
[139] C.M.D. Junior, S.P. da Silva, R.V. da Nobrega, A.C. Barros, A.K. Sangaiah, P.P. Reboucas Filho, V.H.C. de Albuquerque, A new approach for mobile robot
localization based on an online IoT system, Future Gener. Comp. Sy. 100 (2019) 859–881.
[140] A. Saeed, A. Ahmadinia, A. Javed, H. Larijani, Intelligent intrusion detection in low-power IoTs, ACM T. Internet Techn. (TOIT) 16 (4) (2016) 1–25.
[141] L. Calderoni, A. Magnani, D. Maio, IoT Manager: An open-source IoT framework for smart cities, J. Syst. Architect. 98 (2019) 413–423.
[142] A. Celesti, A. Galletta, L. Carnevale, M. Fazio, A. Ĺay-Ekuakille, M. Villari, An IoT cloud system for traffic monitoring and vehicular accidents prevention
based on mobile sensor data processing, IEEE Sens. J. 18 (12) (2017) 4795–4802.
[143] T. Baranwal, P.K. Pateriya, Development of IoT based smart security and monitoring devices for agriculture, in: 2016 6th International Conference–
Cloud System and Big Data Engineering (Confluence), IEEE, 2016, pp. 597–602.
[144] T.R. Bennett, N. Gans, R. Jafari, Data-driven synchronization for Internet-of-Things systems, ACM T. Embed. Comput. S. (TECS) 16 (3) (2017) 1–24.
[145] F.Y. Yavuz, Deep learning in cyber security for internet of things (Doctoral dissertation), 2018.

View publication stats