Content uploaded by Naeem Raza
Author content
All content in this area was uploaded by Naeem Raza on May 22, 2024
Content may be subject to copyright.
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 1
Securing the Internet of Things: A
Comprehensive Review of Security
Challenges and Artificial Intelligence
Solutions
Aized Amin Soofi a, Muhammad Tahir b, *, Naeem Raza a
a Department of Computer Science, NUML Faisalabad Campus, Faisalabad, Pakistan
b Department of Engineering and Computer Science, NUML Faisalabad Campus,
Faisalabad, Pakistan
* Corresponding author: dr.tahir@numl.edu.pk
Abstract:
One of the major needs and challenges of this century is the use of cutting-edge technology
considering the industry 4.0 revolution. The Internet of Things (IoT) falls in the category of a
cutting-edge example of such innovation in the computing and information industry. In IoT
compared to classical networking methods practically; every device we employ is accessible
at any time from any location. Nevertheless, IoT continues to encounter several security
challenges, and the magnitude of cyber-physical security risks is escalating alongside the
widespread use of IoT technologies considering Moore’s laws expected to be 30 billion
devices by 2025. IoT will continue to face vulnerabilities and risks unless there is a
comprehensive understanding and proactive approach towards tackling its security
concerns. To ensure both the cyber and physical security of IoT devices during data
gathering and sharing, it is imperative to evaluate security considerations, identify instances
of cyber-attacks, and implement effective security protocols at multiple layers for making
highly secured IoT. Conventional security measures like data classification, strict access
controls, monitoring privileged account access, encrypting sensitive data, security
awareness training, network segregation, segmentation cloud security, application security,
patch management, and physical security employed in the realm of IoT are inadequate in
light of the current security difficulties posed by the proliferation of sophisticated attacks and
threats. Utilization of artificial intelligence (AI) techniques, especially machine and deep
learning models is becoming a compelling and effective approach to enhance security of the
IoT devices. This research article presents a comprehensive review of the key aspects of IoT
security, including the challenges, potential opportunities, and AI-driven solutions. The
primary goal of this article is to provide technical resources for cybersecurity experts and
researchers working on IoT initiatives.
Keywords: Internet of Things; IoT Security; Artificial Intelligence; Deep Learning;
Machine Learning; Cyber and Physical Security; Industry 4.0.
1. Introduction
IoT is a decentralized network. It connects devices and humans. This
connection is via the internet. IoT makes device connectivity possible. Any
object reachable via IoT is a “thing”. Even home appliances can be
“things”. “Things” can communicate via IoT. They provide useful data.
Sensors and machine learning are IoT subsets [1], [2]. They enable real-
time analysis. Smart devices share collected data. This data helps in daily
tasks. Figure 1 shows the IoT concept. It connects people and objects.
Foundation University
Journal of Engineering and
Applied Sciences
FUJEAS
Vol. 4, Issue 2, 2023.
DOI:10.33897/fujeas.v4i2.779
Article Citation:
Soofi et al. (2023). “Securing
the Internet of Things: A
Comprehensive Review of
Security Challenges and
Artificial Intelligence
Solutions”. Foundation
University Journal of
Engineering and Applied
Sciences
DOI:10.33897/fujeas.v4i2.779
This work is licensed under a
Creative Commons Attribution
4.0 International License,
which permits unrestricted
use, distribution, and
reproduction in any medium,
provided the original work is
properly cited.
Copyright
Copyright © 2023 Soofi et al.
Published by
Foundation University
Islamabad.
Web: https://fui.edu.pk/
Review Article
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 2
Figure 1: IoT definition [20]
There are no time or place limits. “IoT security” is a term used. It refers to IoT system safety [3]. These
systems are internet-dependent. Hence, they are hacker targets. IoT implementation needs network
security. IoT networks are large-scale. They pose new challenges. Data management is one such
challenge. IoT security is crucial. It protects sensitive data. This data is sent by IoT devices [4], [5], [6].
It prevents data theft. It also prevents privacy breaches. Strong security measures are needed. They
prevent cyberattacks. They also prevent breaches. Such incidents can disrupt IoT systems. They can
cause substantial damages. Healthcare has widespread IoT implementations. Transportation is another
such area [7], [8]. Energy grids also use IoT. These systems need reliable security measures.
Organizations must follow security standards. They must also follow laws. This ensures legal
compliance. It also reduces related risks [9], [10], [11]. Inadequate IoT security has severe
consequences. Data breaches are one such consequence. Financial setbacks are another one. It can
harm reputation. It can lead to legal responsibilities. It can also risk public safety [12], [13], [14].
Therefore, it is crucial to establish and give priority to security measures for the IoT to promote the
durability and long-term viability of IoT ecosystems in a world that is becoming more networked and
digital [15].
Some of the potential attacks that need to be addressed for secure IoT systems include; spoofing,
eavesdropping, tampering, jamming, denial of service (DOS), etc. [16]. Traditional methods of handling
security incidents are ineffective because of the recent surge in sophisticated menaces and invasions
and the complexity of these incidents. Therefore, protecting the IoT system requires a powerful security
system utilizing cutting-edge technologies that can handle the challenges. As a key component of the
4.0 industrial revolution; AI provides the most promising avenues for creating smart systems [17]. To
provide a dynamic and up-to-date security solution for the IoT; we can take leverage of artificial
intelligence (AI) knowledge, specifically machine learning (ML) and deep learning (DL), to identify
anomalies or undesirable malicious activities. The security of data is analyzed using ML or DL models,
which offers a collection of regulations, protocols, and complex mathematical functions for transferring
data [18]. Well-known AI techniques like ML and DL models like artificial neural network (ANN), and
convolutional neural network (CNN) can aid IoT devices in learning from experiences represented as
data and adapting their behavior accordingly [6, 19].
Typically, an IoT network or system operates at different layers, Section 2 has covered the three primary
levels of IoT architecture. Different types of cyber and physical threats are associated with each layer
of IoT. Practically multiple AI techniques can be adopted to ensure IoT security like classification,
regression, clustering, rule-based, DL, and hybrid models. In this article, an attempt is made to discuss
the different security threats in IoT environments with their AI-based available solutions in the literature.
The subsequent sections of this research article are structured as follows; IoT architecture has been
discussed in section 2, and characteristics of IoT networks are presented in section 3. In section 4 role
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 3
of ML and DL techniques in IoT security has been discussed. Classification strategies for the security
of IoT have been addressed in section 5 while regression techniques have been discussed in section
6. Section 7 contains the discussion about clustering techniques for IoT security. Section 8 of the
document has covered the application of DL techniques for enhancing security in the IoT. Potential
challenges and opportunities have been addressed in section 9. Table 6 specifically summarizes the AI
techniques with their advantages and disadvantages in IoT environments with security applications.
Table 1 depicts a list of notations used and their definitions used in this research.
Table 1: List of notations used in securing IoT
Notation
Definition
Notation
Definition
IoT
Internet of Things
WSN
Wireless Sensor Network
ANN
Artificial Neural Network
DNN
Deep Neural Network
CNN
Convolutional Neural Network
MLP
Multi-Layer Perception
ML
Machine Learning
NS2
Network Simulator-2
DL
Deep Learning
NIDS
Network Intrusion Detection System
AI
Artificial Intelligence
DSRC
Dedicated Short-Range
Communication
M2M
Machine to Machine
RNN
Recurrent Neural Network
M2G
Machine to Gateway
D2D
Device-to-Device Communication
M2C
Machine to Cloud
KNN
K-Nearest Neighbor
SVM
Support Vector Machine
LR
Logistic Regression
DDoS
Distributed Denial of Service
GMM
Gaussian Mixture Model
DoS
Denial of Service
IDS
Intrusion Detection System
RF
Random forest
BLR
Binary Logistic Regression
RR
Ridge Regression
DT
Decision Tree
2. IoT Architecture
IoT signifies a significant change in the world of information technology. The phrase "Internet of Things,"
often shortened to IoT, combines two critical terms: "Internet" and "Things." In this context, "Things"
refers to intelligent gadgets or objects. Many companies and research organizations explain IoT and
smart environments in various ways and from various angles.
IoT, as described in [21], refers to a combination of physical hardware components and a digital
transmission of data that relies on RFID tags. According to the Institute of Electrical and Electronics
Engineers (IEEE) [22], the IoT is defined as a network of interconnected things equipped with sensors
that are connected to the Internet. Because no universally accepted model for the IoT architecture has
been developed, many models have recently been presented [23]. A three-tier generic architecture for
IoT has been depicted in Figure 2 which contains a perception layer, network layer, and application
layer.
The perception layer is the foundational layer of the architecture of the IoT paradigm and is mostly
called the brain of three-layered architecture, but in real terms, it is a physical layer. This layer in IoT
design is of utmost importance as it acts as the interface connecting the physical and digital domains
[24]. It enables the smooth integration of data from the physical environment into digital systems. The
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 4
perception layer consists of sensors, actuators, and data-collecting devices that allow for the real-time
capture of contextual information about the surrounding environment, objects, and events [25]. The
primary function of this layer is to serve as the sensory component of IoT systems, collecting a wide
range of data including temperature, motion, sound, and light intensity [26, 27].
In the network layer, data is transferred and processed based on what was sensed from objects in the
perception layer. It is the glue that binds the IoT together, linking up computers, servers, and other
smart devices [28]. Machine-to-machine (M2M), machine-to-gateway (M2G), machine-to-cloud (M2C),
and backend data sharing are all facilitated by this layer [29]. The network layer also serves as the
fundamental framework of IoT infrastructure, incorporating a wide array of networking technologies,
protocols, and standards designed to meet the specific needs of IoT ecosystems, such as scalability,
dependability, and energy efficiency [30]. The application layer is the highest-level layer of the IoT
structure [31] responsible for intelligent services at a high level delivered by this layer to meet the
requirements of the customers [32]. This layer functions as the visible interface and coordinator of
capabilities in IoT architecture [33, 34]. It converts raw data into practical insights and provides value-
added services to end-users. Furthermore, it facilitates smooth incorporation with current corporate
procedures and IT systems, unlocking fresh sources of income, improving customer experiences, and
promoting digital transformation in many sectors [35].
Figure 2: Three-Tier IoT architecture
3. Characteristics of IoT Networks
Security and privacy measures that have been used in traditional networks may not be effective on IoT
networks due to the constantly changing and connected nature of IoT poses unique security challenges
because of the following characteristics described here.
3.1. Massive Scale Deployment
There is a belief that the numerous interconnected devices approximately in billions communicating
with each other through the Internet will eventually outstrip the existing Internet's capabilities.
Implementing IoT on a massive scale also presents difficulties, such as creating networking and storage
infrastructure for smart devices, developing effective data communication standards, identifying and
safeguarding against malicious attacks, standardizing technologies, and creating consistent device and
application interfaces [36].
3.2. Heterogeneity
Within an IoT network, many distinct devices possess various capabilities, characteristics, and
communication protocols exists. These devices may employ different communication standards, and
communication paradigms, and may have varying limitations on their hardware resources creating
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 5
problems from smaller scale to larger scale on different layers [37].
3.3. Intelligence
The capability of IoT to make wise judgments quickly and intelligently is one of its most alluring aspects.
To extract meaning from the data generated by IoT devices and take action based on the processed
data, it must be processed in a meaningful way [38].
3.4. Efficient and Affordable Communication
In order to achieve optimal network performance for IoT devices, it is necessary to implement solutions
that have both ultra-low power consumption and cheap cost. These solutions require due to the massive
connectivity involved and that is only possible by designing efficient protocols for routing of data on the
network layer and by designing applications considering web 3.0 development.
3.5. Safety
Alongside other characteristics, ensuring safety is crucial for the effective operation of IoT networks.
Both customers and devices must take safety precautions due to the proliferation of IoT devices, which
might potentially jeopardize the security of personal data transmitted through these devices.
Furthermore, the safety and secrecy of the gadgets themselves are also crucial considerations.
3.6. Dynamic Changes
Efficient management of a vast number of devices is necessary for IoT. These devices operate
dynamically, adjusting to the needs of the application. Factors such as the device's sleep/wake time,
internet usage, and direct communication must also be incorporated into IoT networks.
3.7. Proximal Communication
Another notable characteristic of the IoT is the ability for devices to communicate with one another in
close proximity, without the need for a central authority like base stations. Device-to-device
communication (D2D) makes use of the inherent characteristics of communication from device to
device, which include Dedicated Short-Range Communication (DSRC) and similar innovations. The
conventional architecture of the internet largely emphasizes network-centric interaction. The division of
service providers and networks has made it easier for devices and content to communicate with one
another, expanding the range of services available in the IoT [39].
3.8. Interconnectivity
The term IoT describes the linking of devices and their ability to communicate with each other, much
like a dialogue. As a result, networks connected to the IoT may be accessible whenever and anywhere,
day or night [40].
4. Securing IoT with ML and DL Techniques
The utilization of AI techniques, specifically ML and DL, is widely recognized as a means for IoT devices
to acquire knowledge from data and subsequently adjust their behavior accordingly. Learning models
utilized for this purpose usually consist of a collection of principles, methodologies, or advanced transfer
functions that can be applied to identify significant security incident patterns in IoT data to predict and
detect behavior [41]. Consequently, in the realm of IoT, both ML and DL can function effectively within
ever-changing IoT networks without the need for human intervention or involvement. Figure 3 depicts
how ML and DL techniques have the potential to create a data-focused model for IoT security
intelligence. Various ML techniques can be employed to gain insights from IoT security data, such as
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 6
regression and classification analysis, rule-based methods, clustering, and feature optimization
methods [42, 43].
DL techniques that rely on artificial neural networks (ANN), such as convolutional networks, multi-layer
perceptron networks, and recurrent networks, can also be utilized to secure IoT networks from different
types of attacks [44, 45]. A significant utilization of DL algorithms is mostly for the purpose of anomaly
identification, whereby they instantly detect and stop any breaches or cyber threats by analyzing real-
time network traffic and device behavior [46]. Moreover, these models exhibit outstanding proficiency
in detecting and examining coding patterns and network linkages, therefore bolstering the IoT devices'
security against malicious software. The DL expertise encompasses verification along with access
control, bolstering safety protocols by employing biometric identification, and behavioral evaluation to
strengthen defense against illegal access [47]. DL approaches are utilized to strengthen the security of
communication amongst IoT devices through the implementation of encryption and decryption
strategies. This guarantees the preservation of data confidentiality and integrity.
The subsequent section will address the application of various machine and DL methods in the field of
security solutions inside the IoT framework. Various techniques have been discussed with their primary
aim, dataset, and accuracy. The utilization of ML and DL techniques in IoT applications also presents
novel challenges. These challenges are multifaceted, including the difficulty of creating an appropriate
model to process data from various IoT applications. Likewise, accurately categorizing incoming data
is consequently a tedious operation [48]. Another obstacle is the utilization of a limited amount of labeled
data throughout the learning process. Deploying these models on IoT devices with limited resources
presents additional problems since it is crucial to minimize processing and storage overhead.
5. Classification Techniques in IoT Security
Classification is one of the popular ML approaches in which an object can be placed into one of several
predetermined classes using its attributes [49, 50]. Classification techniques serve as an effective
protective mechanism [51]. Through the analysis of network traffic and device activity, classification
algorithms are capable of identifying unauthorized devices, classifying lawful ones based on their
purpose, and detecting irregularities that indicate potential security risks [52]. This enables the
implementation of focused security protocols, the automatic identification of potential risks, and the
enforcement of restricted access privileges [53]. Classification enhances IoT security by providing it
with the ability to discern between regular and malicious activities, thus protecting the entire network
[54]. A summary of classification techniques used for IoT security has been presented in Table 2. In IoT
security, a classification task typically entails forecasting a defined discrete value or category, such as
normal or anomaly data, and type of attack such as attack-1, attack-2, attack-3, etc. A few commonly
used classification techniques include k-nearest neighbors(KNN) [55], support vector machines (SVM)
[56], naive bayes [57], random forest (RF) [58], and decision trees [59]. These techniques can be
applied to classify security incidents and mitigate various IoT security concerns, such as detecting
intrusions or attacks, analyzing malware, and identifying anomalies or fraudulent behavior within IoT
systems. KMANB algorithm [60] was designed to secure an IoT network from anomalies like trojans,
worms, passwords, backdoors and DDoS attacks. In the proposed approach K-means clustering
algorithm was used to group the anomalies data and the naive bayes algorithm was used to detect the
anomalies. The anomaly detection accuracy of the proposed algorithm on the ToN_IoT dataset was
between 90% to 100%. Statistical results also show the improved speed, accuracy, flexibility and
scalability of the proposed technique.
In [61] a technique named NBC-MAIDS was introduced in which Naive Bayes classification algorithm
was applied in IDSs to overcome the Distributed Denial of Service (DDOS) attacks in IoT networks.
Naive Bayesian distinguisher model was presented in [62] in which the packet loss state was captured
and classify the packet loss type in an IoT network. In this approach, NS2 simulator was used that
showed up to 95% classification accuracy with improved throughput and friendliness in the network.
IoT-based cyber security of drones was ensured in [63] for the prevention of DoS, jamming, and
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 7
spoofing attacks. In this approach, a Naive Bayes algorithm was used with the KDD’99 dataset and
experimental results showed 96.3% accuracy. Although this approach provides 96.3% accuracy but
one of the problems with this approach is that it uses two layers of processing that cause the
independence between information in predicting items.
Figure 3: Potential role of the ML and DL modeling for IoT security intelligence
Digital identification techniques have been introduced in [64] to prevent digital ID spoofing attacks in
IoT devices. In the proposed technique signals were collected from eighteen WiMAX radio devices.
This technique defeats spoofing attacks through feature-reduced RF-DNA fingerprints and an SVM
(Support Vector Machine) classifier with a true verification accuracy of 97.8%. In [65] a model was
proposed that provides detection against DDoS attacks in IoT networks using Naive Bayes and the
KNN classifier. The proposed model was trained on the BoT-IoT dataset in which two data sets were
used, one was a real-time dataset and the second was a class-balanced dataset. The accuracy of Naïve
bayes algorithm on real time dataset and class balanced dataset was 99.4% and 55.1% respectively.
The accuracy of KNN algorithm on real time dataset and class balanced dataset was 99.6% and 92.1%
respectively.
Table 2: Classification techniques in IoT security
Author
AI approach
Security objective
Dataset/data
collection
Accuracy
[60]
Naive Bayes
anomaly detection
ToN_IoT dataset
90-100%
[61]
Naive Bayes
DDOS attack prevention
realtime
-
[62]
Naive Bayesian
Packet loss detection
NS2 simulator
95%
[66]
Naive Bayes
anomaly detection
UNSW-NB15
dataset
92.48%
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 8
[63]
Naive Bayes
Prevent DoS attacks,
jamming and spoofing
Drone KDD’99
dataset
96.3%
[65]
KNN, Naive
Bayes
DDoS attack prevention
BoT-IoT dataset
99.6%
&
99.4%
[64]
SVM
digital ID spoofing attacks
Multiple WiMAX
radio
97.8%
[67]
SVM
Malware attack
14 different malware
98.9 % –
99.8%
[68]
SVM
Abnormal behavior profiling
Mica2Dot sensors
93.7 %
[69]
SVM
Secure data sharing platform
BCWD and HDD
90.35%
and
93.89%
[70]
RF
Intrusion detection
NSL-KDD, UNSW-
NB15, and GPRS
95.5%
[71]
RF
DDOS attack prevention
Real-time traffic by
using Raspberry Pi
v3
99%
[72]
KNN
Detect unauthorized access
Bot IoT
dataset
92.29%
[73]
KNN/SVM/RF
Malware and intrusion
detection
Aposemat IoT-23
89.80%
to
92.96%
[74]
Decision tree
(DT)
Intrusion detection
NSL-KDD
83.14%
To detect a malware attack on Android devices an experimental approach was used [67] in which
fourteen different malware were analyzed by using different ML algorithms. The results show that the
proposed SVM approach provides 99.8% accuracy on DroidKungFu and zitmo malware and 98.9% on
FakeInst malware. Abnormal behavior profiling of IoT devices was performed in [68] by considering four
factors including temperature, humidity, light, and voltage. The data was collected by deploying
Mica2Dot sensors in a real-time environment. The SVM algorithm was used to train normal and
abnormal datasets. The results show 93.7% accuracy in normal datasets and 69.5% accuracy in
abnormal datasets when the malicious user modified data. In [69] SVM training scheme named
secureSVM was proposed to build a secure data-sharing platform for IoT network's homomorphic
cryptosystems. In which the SVM algorithm is applied to two real-world datasets, namely the Heart
Disease Data Set (HDD) and the Breast Cancer Wisconsin Data Set (BCWD). The accuracy of the
proposed algorithm was 90.35% in the case of the BCWD dataset and 93.89 in the case of the HDD
dataset. This proposed technique helped overcome the challenges of data integrity and data privacy in
the transmission of data in IoT networks.
In [70] a parameterized, efficient RF classifier was presented to enhance anomaly detection in IoT
networks. The experiment included three different data sets (NSL-KDD, UNSW-NB15, and GPRS) and
ten different classifiers (each of which was assessed based on the number of trees in its ensemble).
Statistical analysis showed that RF-800 outperformed competing classifiers with 95.5% accuracy.
DDoS detection using an RF classifier was performed in [71]. The premise upon which the selection of
characteristics was based was that consumer IoT devices generate network traffic that is fundamentally
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 9
different from that generated by the well-studied but non-IoT networked devices. An experimental
consumer IoT device network's regular and DoS attack traffic was used to evaluate five different ML
classifiers including RF. The accuracy on the test set was greater than 99% for all five algorithms.
A KNN approach was proposed [72] to detect vulnerabilities in IoT networks. The attack detection
module employs a DL and ML technique, and this approach was evaluated using a bot-IoT dataset with
an accuracy of 92.29%. In [73] three classification techniques were implemented and tested on the
Aposemat IoT-23 dataset. Accuracy levels for intrusion detection attained by the RF, SVM, and KNN
were 92.96%, 86.23%, and 91.48%, while those for malware detection were 92.27%, 83.52%, and
89.80%, respectively. Three decision trees were utilized in a hybrid categorization system [74] in which
results were compared with SVM and KNN. The result showed that the proposed approach perform
better with an accuracy of 83.14% for intrusion detection as compared to the other two approaches but
in [70] RF approach has been applied on the same dataset with 95.5% accuracy.
5.1. Review of classification techniques in IOT security
We can conclude that the performance of KNN, Naive Bayes, and RF-based classification approaches
are best for DDOS attack prevention on both data sets i.e.; real-time and previously available. While
the SVM approach is good for intrusion detection and some specific types of malwares but SVM-based
models are complex and challenging to understand and interpret. Additionally, decision trees and RF
techniques can be used to classify IoT data and predict potential security threats. The building nature
of DT necessitates vast storage facilities. Using only a small number of DTs makes DT-based
approaches straightforward to grasp. These techniques can be used in conjunction with other security
measures, such as encryption and authentication, to enhance the overall security of IoT systems.
6. Regression Techniques in IoT Security
Most of the time, regression analyses are used to make forecasts and predictions, which is a big part
of the field of ML. In some cases, regression analysis can also be used to figure out how the
independent and dependent variables are related to each other [75]. Through the examination of
patterns in sensor data, regression models have the capability to forecast potential device failures,
allowing for preemptive maintenance and the avoidance of security weaknesses [76]. Furthermore, they
contribute to the optimization of resources by efficiently allocating resources such as power and
bandwidth based on device activity. A summary of regression techniques in IoT security has been
presented in Table 3. In [77] a Network Intrusion Detection System (NIDS) is tailored to resource-limited
WSN (Wireless Sensor Network) and IoT nodes. In which the offline training stage involved the creation
of detection modules using Binary Logistic Regression (BLR). These modules were trained using benign
local node activity and malicious behavior from two typical routing attacks. The authors determined that
utilizing training data from a single network topology was enough for identifying assaults in comparable
network topologies, taking into account their size and network density. Accuracy ratings of the proposed
system ranged from 96% to 100% throughout the real-time evaluation phase.
In [78] several different ML methods were used on the data. The dataset was subjected to five rounds
of cross-validation testing with each method. It was proved with experiments that the logistic regression
(LR) approach performed well in the first two-fold testing phases after that its performance became
weak. The average efficiency of linear regression was 98.3%. A Smart Cybersecurity Framework [79]
for IoT-Empowered Drones was presented in which LR and RF techniques were merged to provide
better security on a collected dataset. The accuracy of the proposed framework was 98.58% while the
accuracy of simple LR and RF approaches was 92.23% and 92.36% respectively.
Modeling of DDoS attacks [80] in IoT networks using ML has been performed. Researchers looked at
how well and quickly several ML methods (supervised, unsupervised, and semi-supervised) could spot
DDoS threats in IoT. The DARPA dataset was used for experimental purposes. The result shows
97.93% accuracy in the case of RR and 98.60% accuracy in the case of LR. A LogitRegTrust model
[81] was proposed to ensure authentication and prevent black hole attacks in IoT networks. Reputation
score was used in this approach to compute trust and the indirect trust value of the node was computed
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 10
Table 3: Regression techniques in IoT security
Author
AI approach
Security objective
Dataset/data
collection
Accuracy
[77]
BLR
Intrusion detection
(Blackhole attack)
Run-Time
Monitoring Tool
(RMT)
96% - 100%
[78]
LR
Multiple anomaly
detection
Open-source
dataset
98.3%
[79]
LR & RF
Detection of DoS and
probe attacks
IoT data from
drones, sensors,
and network
information
98.58%
[80]
RR and LR
DDOS attack
prevention
DARPA
RR: 97.93%
LR: 98.60%
[81]
LR
Blackhole attack
detection
COOJA
simulator
-
[82]
Nonlinear regression
Malware detection
(botnet attacks)
Malware analysis
for 1425 files was
conducted.
98.75%
using local trust in the Trust-based RPL while using global trust in the proposed model. In a laboratory
setting, [82] deliberately infected nine commercial IoT devices using two well-known IoT-based botnets,
Mirai and BASHLITE. The projected results demonstrated the recommended strategy's ability to
accurately and swiftly detect the assaults as they were being launched from the compromised IoT
devices that were part of a botnet. The tests demonstrate a remarkable accuracy of 98.75%.
6.1. Review of Regression Techniques in IoT Security
Regression techniques can play a role in IoT security by helping to identify patterns and relationships
in IoT data that can indicate security threats or vulnerabilities. For example, linear and LR can be used
to analyze IoT sensor data and identify unusual patterns or anomalies that may indicate a security
breach. Regression techniques are good for attack detection and mitigation, malware analysis, anomaly
and intrusion detection.
7. Clustering Techniques in IoT Security
Clustering is a process of grouping similar data points into clusters. The goal of clustering is to discover
natural groupings or patterns in the data, without any prior knowledge about the groupings. Clustering
results in data partitioning. Each cluster contains similar data points. Clustering is an unsupervised
learning method. It discovers patterns in unlabeled data. Hidden patterns can be revealed by clustering.
It helps identify IoT anomalies. Table 4 summarizes IoT security clustering. A method was proposed in
[83]. It examines network attack patterns. It suggests an IoT intrusion detection technique. A node
authority management approach based on traffic restriction was suggested to increase the security of
IoT communication and lessen the downsides brought on by algorithm detection failures. A data
intrusion detection technique was developed, which is based on K-means clustering and is highly
efficient.
In [84] three scenarios were used in the experiments employing wireless communication: regular traffic,
attack traffic, and mixed normal-attack traffic. A related dataset was produced for each scenario.
Datasets were then divided into the normal and assault clusters. The clustering outcomes were
generated using the K-Means technique with an efficiency of 99.94%. In [85] the data was divided into
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 11
Table 4: Clustering techniques in IoT security
Author
AI approach
Security objective
Dataset/data
collection
Accuracy
[83]
K-means
Network intrusion detection
sensors
-
[84]
K-means
ping flood attack
pattern recognition
sensors
99.94%
[85]
Fuzzy clustering
Intrusion detection
-
99%
[86]
GMM
DDOS attack detection
CIC-IDS2017
CIC-DDoS2019
94%
[87]
GMM
Anomaly detection
NAB dataset
Self-made dataset
-
[88]
GMM
spoofing attack detection
2D feature vector
extracted from an
estimated channel
state vector
98%
high-risk data and low-risk data, which are correspondingly detected by high frequency and low
frequency. Both the principal component analysis approach and the suppressed fuzzy clustering
algorithm were used simultaneously for the self-adjustment of the detection frequency. It was found that
as data amount increases, intrusion detection algorithm accuracy and efficiency gradually decline. The
suggested method [86] was effective at identifying known DDoS attacks. However, the system
performance suffers greatly when faced with innovative attacks. The proposed technique [87] can
perform well in the health system. The real-time anomaly detection algorithm works because it was
tested on two different types of datasets. But most of this work is about how to find low-dimensional
anomalies. It does not look at how to find high-dimensional or super-high-dimensional anomalies. In
[88] successful usage has been found for a two-dimensional feature vector based on the distance and
correlation between two channel state vectors. However, other features are feasible, and the use of
more than two features should be seriously examined.
7.1. Review of Clustering Techniques in IoT Security
Clustering identifies patterns in IoT data. It also spots anomalies. It helps detect security breaches.
Suspicious devices can be isolated. GMM models IoT devices’ behavior. It detects anomalies and
malicious activity. GMM can analyze sensor data. It spots unusual patterns. Deviations from normal
behavior are detected. These could indicate a security breach.
8. Deep Learning Techniques in IoT Security
DL is a subset of ML techniques that are based on ANNs with multiple layers [89]. These techniques
are used to automatically learn representations of data, such as images, audio, and text, by training a
neural network on a large dataset. DL techniques can learn from IoT security data by passing through
layers, which are known as hierarchical learning methods due to their ability to capture knowledge in
deep architectures. The proposed architecture and model [90] are both fast and accurate, while also
being sensitive to the restricted resources of IoTs. It has also been observed that accuracy starts falling
at some point with an increase in the size of the dataset. Table 5 represents the summary of DL
techniques in the IoT environment.
Using quantitative measurements for the assessment of images, such as peak signal-to-noise ratio
(SNR), structural similarity index, and mean squared error (MSE), the proposed framework [91] by using
CNN proved to be superior to existing methods. The technique was applied to the MRI dataset but this
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 12
Table 5: DL techniques in IoT security
Author
AI approach
Security objective
Dataset/data
collection
Accuracy
[90]
CNN
Malware detection
IoTPoT
95%
[91]
CNN
Medical image security
MRI Dataset
-
[92]
CNN
Malware detection
IoT_Malware
dataset
97.93%
[93]
CNN
Malicious data identification
Kitsune network
attack database
-
[94]
RNN
Intrusion detection
NSL-KDD
97.35%
[95]
RNN
Intrusion detection
DARPA/KDD Cup
'99
98.91%
[96]
DNN
Anomaly detection
IoT-Botnet 2020
99%
[97]
MLP
Botnet attack detection
captured
from 9 IoT devices
99%
technique may perform poorly for complex blurred and color images. In addition to recognizing other
sorts of attack categories, the model [94] demonstrates excellent sensitivity to DoS attacks, which are
one of the most prominent attacks that impede the growth of IoT networks. The outcomes of a proposed
method [95] were superior to those of previously published work, and they were truly excellent. This
study targeted IoT gadgets with limited processing capabilities and manageable data loads. However,
in a scenario, where processing power is high and data amount is vast, this strategy cannot perform
better. Proposed technique [96] provides efficiency of 99% in case of anomaly detection but this
technique cannot provide higher efficiency in multiclass classification scenarios.
8.1. Review of DL Techniques in IOT Security
DL's primary benefit over conventional machine learning is its higher level of accuracy on massive
datasets. DL techniques can be used in IoT security to improve the detection of anomalies and
malicious activity, as well as to protect the privacy of IoT device users. However, DL methodologies
require massive amounts of data, computing resources, and high hardware specifications.
Table 6: Overview of AI techniques with their advantages and disadvantages in IoT
AI Technique
Advantages
Disadvantages
IoT security
applications
Naive Bayes
powerful and efficient
algorithm that can be used
for IoT security to detect and
classify anomalies and
predict potential security
threats [98].
prone to overfitting the
training data when the
number of features is too
large compared to the size
of the dataset [99].
anomaly detection,
classification, and
prediction.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 13
SVM
robust to noise and can
handle data with missing
values, making it suitable for
analyzing large, real-world
datasets generated by IoT
devices [100].
sensitive to outliers, which
can affect the performance
of the algorithm [101].
Outliers in IoT data can be
common and difficult to
detect, making SVM less
suitable in certain
situations.
anomaly detection,
intrusion detection,
malware detection,
and physical attack
detection.
KNN
simple and easy-to-
understand algorithm that
requires minimal training,
making it a suitable option for
IoT security applications
[102].
cannot handle missing
data and requires
complete datasets. In IoT
security applications
where data can be missing
or incomplete, this can be
a limitation [103].
anomaly detection
RF
robust to noisy or incomplete
data and can handle missing
values effectively.
can be less effective when
dealing with small
datasets, which can be a
limitation in IoT security
applications where data is
limited [104].
device
fingerprinting,
authentication,
botnet detection,
vulnerability
detection, anomaly
detection
DT
can handle both categorical
and numerical data, making
them suitable for a wide
range of IoT security
applications.
can be sensitive to noise
or outliers in the data,
which can result in
inaccurate or unreliable
predictions [105].
vulnerability
detection, intrusion
detection, and
anomaly detection
BLR
can handle imbalanced data,
which may be common in IoT
security applications where
some types of threats are
rare [106].
can overfit the training
data, leading to poor
generalization
performance on new data
[77].
spoofing attacks,
DOS attacks,
malware detection,
and physical
attacks
LR
produces probabilistic
predictions, which can be
useful in IoT security
applications where
understanding the
confidence of a prediction is
important [107].
requires careful feature
selection to avoid
overfitting and to ensure
that the selected features
are relevant to the security
threat being detected.
spoofing attacks,
DOS attacks,
malware detection,
and physical
attacks
RR
can be used for both
regression and classification
tasks in IoT security
applications, making it a
versatile algorithm.
assumes that all input
variables are relevant to
the prediction task, which
may not always be the
case in IoT security
applications [80].
anomaly detection,
intrusion detection,
and malware
detection
K-means
can identify anomalous
patterns in IoT data that may
indicate a security threat
[108].
requires the number of
clusters to be specified in
advance, which can be
difficult to determine in IoT
applications.
anomaly detection,
botnet attacks,
intrusion detection,
and malware
detection
Fuzzy
clustering
can adapt to changing
patterns in IoT data, making
it suitable for applications
where the underlying data
may not work well for all
types of IoT data,
particularly if the data is
highly skewed or contains
outliers that cannot be
network intrusion
and anomaly
detection
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 14
structure may evolve over
time [109].
easily modeled using a
fuzzy approach [110].
CNN
can automatically learn and
extract features from IoT
data without requiring
manual feature engineering.
may not generalize well to
new, unseen IoT security
scenarios, particularly if
the distribution of data is
significantly different from
the training data [111].
man-in-the-middle
attacks, virus
detection, breach
detection, and DOS
attack
RNN
can deal with variable length
and size inputs, which is
important in IoT where data
is noisy and unreliable [112].
can be computationally
expensive, particularly
when dealing with large
datasets, which can limit
their practicality in some
IoT applications [113].
forecasting and
mitigation, malware
identification, and
detecting intrusions
DNN
can learn complex patterns
and relationships in data,
making them well-suited for
detecting security threats in
IoT systems [114].
require labeled data for
training, which can be
time-consuming and
expensive to obtain,
particularly in IoT
applications where data
may be noisy or
unlabeled.
botnet detection,
intrusion detection,
malware detection,
and anomaly
detection
MLP
relatively simple and easy to
implement compared to other
neural network architectures,
making them a popular
choice for many IoT
applications [115].
limited capacity to handle
complex patterns and may
not perform as well as
other neural network
architectures in some IoT
applications [116].
DOS attack, DDOS
attack, malware
detection, anomaly
detection
9. Challenges and Opportunities
The primary emphasis of the present study pertaining to provenance security has revolved around the
identification of requirements and the proposition of solutions employing established AI techniques for
safeguarding data in IoT environments. It would be of great interest to investigate whether, similar to
the realm of privacy, the interplay between data and provenance gives rise to novel security challenges
and corresponding remedies. The absence of universally accepted security standards for IoT devices
poses challenges in maintaining consistent security measures across various products and vendors.
The establishment of reliable security standards still needs the attention of the research community.
Moreover, insufficient authentication systems might facilitate unauthorized access by attackers to IoT
gadgets and networks. There is limited literature available that uses AI approaches to overcome the
issue of authentication. More study is required to identify the strong role of AI in the authentication
process for IoT environments.
There exist multiple datasets that are deemed appealing to investigate network intrusion detection. One
example of a widely employed dataset for the examination of network IDSs is KDD 99. Nevertheless, it
is important to note that there is currently a lack of publicly available datasets specifically focused on
pure IoT threats. Certain widely used datasets, like as NSL-KDD, encompass a variety of security
attacks. It is worth noting that a significant proportion of the malicious instances present in the NSL-
KDD dataset are specifically categorized as DoS attacks. Utilization of these methods for various forms
of attacks poses a significant challenge in terms of study and analysis.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 15
10. Concluding Remarks
A comprehensive critical analysis of the existing literature on the topic of IoT security challenges and
solutions based on AI techniques, specifically machine learning and DL has been conducted in this
research. These techniques can identify anomalous activity or variations from typical patterns in the
connectivity of IoT devices. These techniques can also reduce false positive alarms by differentiating
between legal and harmful operations using learned patterns and contextual information. This study
encompasses several aspects such as the IoT paradigm, IoT-based smart environments, the
associated security concerns, and potential solutions that leverage AI. To facilitate the advancement of
the argument presented in this paper, an extensive examination of the current status of research on
security in the IoT was conducted. The effectiveness and efficiency of an IoT security solution that
utilizes ML or DL techniques are greatly influenced by the qualities and features of the data, as well as
the performance of the learning algorithms. To effectively identify and mitigate cyberattacks targeting
IoT devices and systems, it is imperative to conduct a thorough examination of IoT system architecture.
Hence, a concise examination has been conducted to explore the potential use of several machine and
DL algorithms in addressing security challenges inside the IoT environment. An effective security
framework for the IoT should use machine or DL modeling, as deemed suitable based on the attributes
of the data. For the system to facilitate intelligent decision-making, it is imperative to develop a proficient
learning algorithm that is grounded in the acquired IoT security information that pertains to the specific
application at hand.
We can conclude that the performance of KNN, Naive Bayes, and RF-based classification approaches
are best for DDOS attack prevention on both data sets i.e.; real-time and previously available. While
the SVM approach is good for intrusion detection and some specific types of malwares but SVM-based
models are complex and challenging to understand and interpret. Regression techniques are good for
attack detection and mitigation, malware analysis, anomaly and intrusion detection. GMM can be used
to model the normal behavior of IoT devices and detect any anomalies or malicious activity. For
example, GMM can be used to analyze sensor data and detect any unusual patterns or deviations from
the normal behavior of the device, which could indicate a security breach or attack. DL's primary benefit
over conventional machine learning is its higher level of accuracy on massive datasets. IoT security
may utilize DL approaches to improve the detection of anomalies and malicious activity, as well as to
protect the privacy of IoT device users.
Conflict of Interest
There are no conflicts of interest to declare regarding this manuscript.
Data Availability
Data supporting the findings of this study are provided in this manuscript.
Funding Statement
This research study did not receive any funding.
11. References
[1] R. Mahmoud, T. Yousuf, F. Aloul, and I. Zualkernan, "Internet of things (IoT) security: Current status, challenges and
prospective measures," In 2015 10th International Conference for Internet Technology and Secured Transactions
(ICITST), 2015: IEEE, pp. 336-341.
[2] A. Ali, A. Mateen, A. Hanan, and F. Amin, "Advanced Security Framework for Internet of Things (IoT)," Technologies, vol.
10(3), p. 60, 2022.
[3] M. Javaid and I. H. Khan, "Internet of Things (IoT) enabled healthcare helps to take the challenges of COVID-19
Pandemic," Journal of Oral Biology and Craniofacial Research, vol. 11(2), pp. 209-214, 2021.
[4] G. Alqarawi, B. Alkhalifah, N. Alharbi, and S. El Khediri, "Internet-of-Things Security and Vulnerabilities: Case Study,"
Journal of Applied Security Research, pp. 1-17, 2022.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 16
[5] S. Haseeb, M. Khalil Afzal, M. Tahir, M. Raza Jafri, and N. Raza, "Energy‐efficient selection of relay for UWSNs in the
Internet of underwater things," International Journal of Communication Systems, vol. 36(18), p. e5619, 2023.
[6] I. H. Sarker, A. I. Khan, Y. B. Abushark, and F. Alsolami, "Internet of things (IoT) security intelligence: a comprehensive
overview, machine learning solutions and research directions," Mobile Networks and Applications, pp. 1-17, 2022.
[7] F. Alwahedi, A. Aldhaheri, M. A. Ferrag, A. Battah, and N. Tihanyi, "Machine learning techniques for IoT security: Current
research and future vision with generative AI and large language models," Internet of Things and Cyber-Physical
Systems, 2024.
[8] A. A. Soofi and H. Ahmad, "Deep Learning for IoT Security: Applications and Challenges," Emerging Technologies for
Securing the Cloud and IoT, pp. 69-99, 2024.
[9] D. Kolevski and K. Michael, "Edge Computing and IoT Data Breaches: Security, Privacy, Trust, and Regulation," IEEE
Technology and Society Magazine, vol. 43(1), pp. 22-32, 2024.
[10] M. Adil, M. K. Khan, N. Kumar, M. Attique, A. Farouk, M. Guizani and Z. Jin, "Healthcare Internet of Things: Security
Threats, Challenges and Future Research Directions," IEEE Internet of Things Journal, 2024.
[11] G. Nissar, R. A. Khan, S. Mushtaq, S. A. Lone, and A. H. Moon, "IoT in healthcare: a review of services, applications, key
technologies, security concerns, and emerging trends," Multimedia Tools and Applications, pp. 1-62, 2024.
[12] H. Gupta, S. Sharma, and S. Agrawal, "Artificial Intelligence-Based Anomalies Detection Scheme for Identifying Cyber
Threat on IoT-Based Transport Network," IEEE Transactions on Consumer Electronics, 2023.
[13] C. Ni and S. C. Li, "Machine learning enabled Industrial IoT Security: Challenges, Trends and Solutions," Journal of
Industrial Information Integration, p. 100549, 2024.
[14] M. Muhammad, S. U. Bazai, S. Ullah, S. A. A. Shah, S. Aslam, A. Amphawan and T. K. Neo, "A systematic literature
review on the role of big data in IoT security," Journal of Telecommunications and the Digital Economy, vol. 12(1), pp. 39-
64, 2024.
[15] I. H. Sarker, A. I. Khan, Y. B. Abushark, and F. Alsolami, "Internet of things (IoT) security intelligence: a comprehensive
overview, machine learning solutions and research directions," Mobile Networks and Applications, vol. 28(1), pp. 296-
312, 2023.
[16] S. M. Tahsien, H. Karimipour, and P. Spachos, "Machine learning based solutions for security of Internet of Things (IoT):
A survey," Journal of Network and Computer Applications, vol. 161, p. 102630, 2020.
[17] I. H. Sarker, M. H. Furhad, and R. Nowrozy, "Ai-driven cybersecurity: an overview, security intelligence modeling and
research directions," SN Computer Science, vol. 2(3), pp. 1-18, 2021.
[18] I. H. Sarker, A. Kayes, S. Badsha, H. Alqahtani, P. Watters, and A. Ng, "Cybersecurity data science: an overview from
machine learning perspective," Journal of Big data, vol. 7(1), pp. 1-29, 2020.
[19] A. A. Soofi, "Exploring Deep Learning Techniques for Glaucoma Detection: A Comprehensive Review," Journal of
Computing & Biomedical Informatics, vol. 5(02), pp. 220-238, 2023.
[20] J. S. Kumar and D. R. Patel, "A survey on internet of things: Security and privacy issues," International Journal of
Computer Applications, vol. 90(11), 2014.
[21] F. Thiesse and F. Michahelles, "An overview of EPC technology," Sensor Review, 2006.
[22] N. J. Kulkarni and J. Bakal, "E-health: Iot based system and correlation of vital stats in identification of mass disaster
event," In 2018 Fourth International Conference on Computing Communication Control and Automation (ICCUBEA),
2018: IEEE, pp. 1-6.
[23] F. Masoodi, S. Alam, and S. T. Siddiqui, "Security & privacy threats, attacks and countermeasures in Internet of Things,"
International Journal of Network Security & Its Applications (IJNSA), vol. 11, 2019.
[24] E. F. N. Shipena and A. M. Gamundani, "A Review of Internet of Things (IoT) Smart Office Data Security Threats," Digital
Technology and Changing Roles in Managerial and Financial Accounting: Theoretical Knowledge and Practical
Application, pp. 375-381, 2024.
[25] Rishikesh and D. Sinha, "Traditional and Blockchain Based IoT and IIoT Security in the Context of Agriculture: A Survey,"
Wireless Personal Communications, pp. 1-29, 2024.
[26] S. A. Kumar, T. Vealey, and H. Srivastava, "Security in internet of things: Challenges, solutions and future directions," In
2016 49th Hawaii International Conference on System Sciences (HICSS), 2016: IEEE, pp. 5772-5781.
[27] P. Sethi and S. R. Sarangi, "Internet of things: architectures, protocols, and applications," Journal of Electrical and
Computer Engineering, vol. 2017, 2017.
[28] V. Agarwal, P. Mishra, S. Kumar, and E. S. Pilli, "A Review on Attack and Security Tools at Network Layer of IoT," Optical
and Wireless Technologies, pp. 497-506, 2022.
[29] H. Tschofenig, J. Arkko, D. Thaler, and D. McPherson, "Architectural considerations in smart object networking," 2070-
1721, 2015.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 17
[30] S. Lakshminarayana, A. Praseed, and P. S. Thilagam, "Securing the IoT Application Layer from an MQTT Protocol
Perspective: Challenges and Research Prospects," IEEE Communications Surveys & Tutorials, 2024.
[31] C. Stergiou, K. E. Psannis, B.-G. Kim, and B. Gupta, "Secure integration of IoT and cloud computing," Future Generation
Computer Systems, vol. 78, pp. 964-975, 2018.
[32] P. K. Sadhu, V. P. Yanambaka, and A. Abdelgawad, "Internet of Things: Security and Solutions Survey," Sensors, vol.
22(19), p. 7433, 2022.
[33] H. G. Hamid and Z. T. Alisa, "Survey on IoT application layer protocols," Indonesian Journal of Electrical Engineering and
Computer Science, vol. 21(3), pp. 1663-1672, 2021.
[34] N. Raza and M. Tariq, "Effect of Node Density over the performance of DSR, TORA, and OLSR Routing Protocols of
MANET," International Journal of Computer Applications, vol. 177(39), pp. 34-41.
[35] B. Nataraj and P. Duraisamy, "An Investigation on Attacks in Application Layer Protocols and Ransomeware Threats in
Internet of Things," In 2023 9th International Conference on Advanced Computing and Communication Systems
(ICACCS), 2023, vol. 1: IEEE, pp. 668-672.
[36] U. S. Shanthamallu, A. Spanias, C. Tepedelenlioglu, and M. Stanley, "A brief survey of machine learning methods and
their sensor and IoT applications," In 2017 8th International Conference on Information, Intelligence, Systems &
Applications (IISA), 2017: IEEE, pp. 1-8.
[37] R. Kollolu, "A Review on Wide Variety and Heterogeneity of IoT Platforms," The International Journal of Analytical and
Experimental Modal Analysis, Analysis, vol. 12, pp. 3753-3760, 2020.
[38] N. Javaid, A. Sher, H. Nasir, and N. Guizani, "Intelligence in IoT-based 5G networks: Opportunities and challenges,"
IEEE Communications Magazine, vol. 56(10), pp. 94-100, 2018.
[39] B. Panigrahi, H. K. Rath, R. Ramamohan, and A. Simha, "Energy and spectral efficient direct Machine-to-Machine (M2M)
communication for cellular Internet of Things (IoT) networks," In 2016 International Conference on Internet of Things and
Applications (IOTA), 2016: IEEE, pp. 337-342.
[40] S. Li, L. D. Xu, and S. Zhao, "The internet of things: a survey," Information Systems Frontiers, vol. 17, pp. 243-259, 2015.
[41] S. Dua and X. Du, Data mining and machine learning in cybersecurity. CRC Press, 2016.
[42] I. H. Sarker, "Machine learning: Algorithms, real-world applications and research directions," SN computer science, vol.
2(3), p. 160, 2021.
[43] A. A. Soofi and A. Awan, "Classification techniques in machine learning: applications and issues," J. Basic Appl. Sci, vol.
13(1), pp. 459-465, 2017.
[44] I. H. Sarker, "Deep cybersecurity: a comprehensive overview from neural network and deep learning perspective," SN
Computer Science, vol. 2(3), p. 154, 2021.
[45] I. H. Sarker, "Deep learning: a comprehensive overview on techniques, taxonomy, applications and research directions,"
SN Computer Science, vol. 2(6), p. 420, 2021.
[46] H. Liao, M. Z. Murah, M. K. Hasan, A. H. M. Aman, J. Fang, X. Hu and A. U. R. Khan, "A Survey of Deep Learning
Technologies for Intrusion Detection in Internet of Things," IEEE Access, 2024.
[47] R. Kumar, G. Joshi, A. K. S. Chauhan, A. K. Singh, and A. K. Rao, "A Deep Learning and Channel Sounding Based Data
Authentication and QoS Enhancement Mechanism for Massive IoT Networks," Wireless Personal Communications, vol.
130(4), pp. 2495-2514, 2023.
[48] M. Zheng, D. Xu, L. Jiang, C. Gu, R. Tan, and P. Cheng, "Challenges of privacy-preserving machine learning in IoT," In
Proceedings of the First International Workshop on Challenges in Artificial Intelligence and Machine Learning for Internet
of Things, 2019, pp. 1-7.
[49] A. O. Adebayo and M. S. Chaubey, "Data mining classification techniques on the analysis of student’s performance,"
GSJ, vol. 7(4), pp. 45-52, 2019.
[50] M. Tariq, A. A. Sufi, H. ur Rehman, S. K. Ajmal, D. Riaz, M. Amjad, B. Ahmad and M. H. Raza, "Brain Tumor
Classification Using Convolutional Neural Network with Neutrosophy, Super-Resolution and SVM," Webology (ISSN:
1735-188X), vol. 18(1), 2022.
[51] W. B. Demilie, "Plant disease detection and classification techniques: a comparative study of the performances," Journal
of Big Data, vol. 11(1), p. 5, 2024.
[52] M. Aljabri, A. A. Alahmadi, R. M. A. Mohammad, F. Alhaidari, M. Aboulnour, D. M. Alomari and S. Mirza, "Machine
learning-based detection for unauthorized access to IoT devices," Journal of Sensor and Actuator Networks, vol. 12(2),
p.27, 2023.
[53] P. Shukla, C. R. Krishna, and N. V. Patil, "EIoT-DDoS: embedded classification approach for IoT traffic-based DDoS
attacks," Cluster Computing, vol. 27(2), pp. 1471-1490, 2024.
[54] J. H. Kalwar and S. Bhatti, "Deep Learning Approaches for Network Traffic Classification in the Internet of Things (IoT): A
Survey," arXiv preprint arXiv:2402.00920, 2024.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 18
[55] P. Cunningham and S. J. Delany, "k-Nearest neighbour classifiers-A Tutorial," ACM computing surveys (CSUR), vol.
54(6), pp. 1-25, 2021.
[56] W. S. Noble, "What is a support vector machine?," Nature Biotechnology, vol. 24(12), pp. 1565-1567, 2006.
[57] G. I. Webb, E. Keogh, and R. Miikkulainen, "Naïve Bayes," Encyclopedia of machine learning, vol. 15, pp. 713-714, 2010.
[58] S. J. Rigatti, "Random forest," Journal of Insurance Medicine, vol. 47(1), pp. 31-39, 2017.
[59] S. B. Kotsiantis, "Decision trees: a recent overview," Artificial Intelligence Review, vol. 39, pp. 261-283, 2013.
[60] L. Best, E. Foo, and H. Tian, "A Hybrid Approach: Utilising Kmeans Clustering and Naive Bayes for IoT Anomaly
Detection," arXiv preprint arXiv:2205.04005, 2022.
[61] A. Mehmood, M. Mukherjee, S. H. Ahmed, H. Song, and K. M. Malik, "NBC-MAIDS: Naïve Bayesian classification
technique in multi-agent system-enriched IDS for securing IoT against DDoS attacks," The Journal of Supercomputing,
vol. 74(10), pp. 5156-5170, 2018.
[62] Y. Chen, L. Lu, X. Yu, and X. Li, "Adaptive method for packet loss types in IoT: an naive Bayes distinguisher,"
Electronics, vol. 8(2), p. 134, 2019.
[63] R. Majeed, N. A. Abdullah, and M. F. Mushtaq, "IoT-based Cyber-security of Drones using the Naïve Bayes Algorithm,"
International Journal of Advanced Computer Science and Applications, vol. 12(7), 2021.
[64] D. Reising, J. Cancelleri, T. D. Loveless, F. Kandah, and A. Skjellum, "Radio identity verification-based IoT security using
RF-DNA fingerprints and SVM," IEEE Internet of Things Journal, vol. 8(10), pp. 8356-8371, 2020.
[65] S. Pokhrel, R. Abbas, and B. Aryal, "IoT security: botnet detection in IoT using machine learning," arXiv preprint
arXiv:2104.02231, 2021.
[66] S. Manimurugan, "IoT-Fog-Cloud model for anomaly detection using improved Naive Bayes and principal component
analysis," Journal of Ambient Intelligence and Humanized Computing, pp. 1-10, 2021.
[67] H. S. Ham, H. H. Kim, M. S. Kim, and M. J. Choi, "Linear SVM-based android malware detection for reliable IoT
services," Journal of Applied Mathematics, vol. 2014, 2014.
[68] S. Y. Lee, S.-r. Wi, E. Seo, J. K. Jung, and 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), 2017: IEEE, pp. 1-6.
[69] M. Shen, X. Tang, L. Zhu, X. Du, and M. Guizani, "Privacy-preserving support vector machine training over blockchain-
based encrypted IoT data in smart cities," IEEE Internet of Things Journal, vol. 6(5), pp. 7702-7712, 2019.
[70] R. Primartha and B. A. Tama, "Anomaly detection using random forest: A performance revisited," In 2017 International
Conference on Data and Software Engineering (ICoDSE), 2017: IEEE, pp. 1-6.
[71] R. Doshi, N. Apthorpe, and N. Feamster, "Machine learning DDoS detection for consumer internet of things devices," In
2018 IEEE Security and Privacy Workshops (SPW), 2018: IEEE, pp. 29-35.
[72] S. S. S. Sugi and S. R. Ratna, "Investigation of machine learning techniques in intrusion detection system for IoT
network," In 2020 3rd International Conference on Intelligent Sustainable Systems (ICISS), 2020: IEEE, pp. 1164-1167.
[73] S. Strecker, R. Dave, N. Siddiqui, and N. Seliya, "A modern analysis of aging machine learning based IOT cybersecurity
methods," arXiv preprint arXiv:2110.07832, 2021.
[74] S. M. Taghavinejad, M. Taghavinejad, L. Shahmiri, M. Zavvar, and M. H. Zavvar, "Intrusion detection in IoT-based smart
grid using hybrid decision tree," In 2020 6th International Conference on Web Research (ICWR), 2020: IEEE, pp. 152-
156.
[75] Z. Cui, X. Xu, X. U. E. Fei, X. Cai, Y. Cao, W. Zhang and J. Chen, "Personalized recommendation system based on
collaborative filtering for IoT scenarios," IEEE Transactions on Services Computing, vol. 13(4), pp. 685-695, 2020.
[76] F. A. M. Solomon, G. W. Sathianesan, and R. Ramesh, "Logistic Regression Trust-A Trust Model for Internet-of-Things
Using Regression Analysis," Computer Systems Science & Engineering, vol. 44(2), 2023.
[77] C. Ioannou and V. Vassiliou, "An intrusion detection system for constrained WSN and IoT nodes based on binary logistic
regression," In Proceedings of the 21st ACM International Conference on Modeling, Analysis and Simulation of Wireless
and Mobile Systems, 2018, pp. 259-263.
[78] M. Hasan, M. M. Islam, M. I. I. Zarif, and M. Hashem, "Attack and anomaly detection in IoT sensors in IoT sites using
machine learning approaches," Internet of Things, vol. 7, p. 100059, 2019.
[79] W. D. Nanda and F. D. S. Sumadi, "LRDDoS Attack Detection on SD-IoT Using Random Forest with Logistic Regression
Coefficient," Jurnal RESTI (Rekayasa Sistem Dan Teknologi Informasi), vol. 6(2), pp. 220-226, 2022.
[80] P. Machaka, O. Ajayi, H. Maluleke, F. Kahenga, A. Bagula, and K. Kyamakya, "Modelling DDoS Attacks in IoT Networks
using Machine Learning," arXiv preprint arXiv:2112.05477, 2021.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 19
[81] K. Prathapchandran and T. Janani, "A trust-based security model to detect misbehaving nodes in Internet of Things (IoT)
environment using logistic regression," In Journal of Physics: Conference Series, 2021, vol. 1850(1): IOP Publishing, p.
012031.
[82] E. Arul, "Deep nonlinear regression least squares polynomial fit to detect malicious attack on IoT devices," Journal of
Ambient Intelligence and Humanized Computing, vol. 12, pp. 769-779, 2021.
[83] J. Zhu, L. Huo, M. D. Ansari, and M. A. Ikbal, "Research on Data Security Detection Algorithm in IoT Based on K-means,"
Scalable Computing: Practice and Experience, vol. 22(2), pp. 149–159-149–159, 2021.
[84] D. Stiawan, M. E. Suryani, M. Y. Idris, M. N. Aldalaien, N. Alsharif, and R. Budiarto, "Ping Flood Attack Pattern
Recognition Using a K-Means Algorithm in an Internet of Things (IoT) Network," IEEE Access, vol. 9, pp. 116475-
116484, 2021.
[85] L. Liu, B. Xu, X. Zhang, and X. Wu, "An intrusion detection method for internet of things based on suppressed fuzzy
clustering," EURASIP Journal on Wireless Communications and Networking, vol. 2018(1), pp. 1-7, 2018.
[86] C.-S. Shieh, W.-W. Lin, T.-T. Nguyen, C.-H. Chen, M.-F. Horng, and D. Miu, "Detection of unknown DDoS attacks with
deep learning and gaussian mixture model," Applied Sciences, vol. 11(11), p. 5213, 2021.
[87] N. Ding, H. Ma, H. Gao, Y. Ma, and G. Tan, "Real-time anomaly detection based on long short-Term memory and
Gaussian Mixture Model," Computers & Electrical Engineering, vol. 79, p. 106458, 2019.
[88] X. Qiu, T. Jiang, S. Wu, and M. Hayes, "Physical layer authentication enhancement using a Gaussian mixture model,"
IEEE Access, vol. 6, pp. 53583-53592, 2018.
[89] J. Han, M. Kamber, and J. Pei, "Data mining concepts and techniques third edition," University of Illinois at Urbana-
Champaign Micheline Kamber Jian Pei Simon Fraser University, 2012.
[90] A. M. Zaza, S. K. Kharroub, and K. Abualsaud, "Lightweight IoT malware detection solution using cnn classification," in
2020 IEEE 3rd 5G World Forum (5GWF), 2020: IEEE, pp. 212-217.
[91] S. More, J. Singla, S. Verma, U. Ghosh, J. J. Rodrigues, A. S. Hosen and I. H. Ra, "Security assured CNN-based model
for reconstruction of medical images on the internet of healthcare things," IEEE Access, vol. 8, pp. 126333-126346, 2020.
[92] M. Asam, S. H. Khan, A. Akbar, S. Bibi, T. Jamal, A. Khan, U. Ghafoor and M. R. Bhutta, "IoT malware detection
architecture using a novel channel boosted and squeezed CNN," Scientific Reports, vol. 12(1), pp. 1-12, 2022.
[93] Z. Gu, S. Nazir, C. Hong, and S. Khan, "Convolution neural network-based higher accurate intrusion identification system
for the network security and communication," Security and Communication Networks, vol. 2020, 2020.
[94] M. Almiani, A. AbuGhazleh, A. Al-Rahayfeh, S. Atiewi, and A. Razaque, "Deep recurrent neural network for IoT intrusion
detection system," Simulation Modelling Practice and Theory, vol. 101, p. 102031, 2020.
[95] M. K. Putchala, "Deep learning approach for intrusion detection system (IDs) in the internet of things (IoT) network using
gated recurrent neural networks (GRU)," Master's thesis, Wright State University, 2017.
[96] Z. Ahmad, A. S. Khan, K. Nisar, I. Haider, R. Hassan, M. R. Haque, S. Tarmizi and J. J. Rodrigues, "Anomaly detection
using deep neural network for IoT architecture," Applied Sciences, vol. 11(15), p. 7050, 2021.
[97] Y. Javed and N. Rajabi, "Multi-layer perceptron artificial neural network based IoT botnet traffic classification," In
Proceedings of the Future Technologies Conference, 2019: Springer, pp. 973-984.
[98] S. Na, T. Kim, and H. Kim, "A study on the classification of common vulnerabilities and exposures using naïve bayes," In
Advances on Broad-Band Wireless Computing, Communication and Applications: Proceedings of the 11th International
Conference On Broad-Band Wireless Computing, Communication and Applications (BWCCA–2016) November 5–7,
2016, Korea, 2017: Springer, pp. 657-662.
[99] M. Jindal, J. Gupta, and B. Bhushan, "Machine learning methods for IoT and their Future Applications," In 2019
International Conference on Computing, Communication, and Intelligent Systems (ICCCIS), 2019: IEEE, pp. 430-434.
[100] A. Al Shorman, H. Faris, and I. Aljarah, "Unsupervised intelligent system based on one class support vector machine and
Grey Wolf optimization for IoT botnet detection," Journal of Ambient Intelligence and Humanized Computing, vol. 11, pp.
2809-2825, 2020.
[101] M. Hosseinzadeh, A. M. Rahmani, B. Vo, M. Bidaki, M. Masdari, and M. Zangakani, "Improving security using SVM-
based anomaly detection: issues and challenges," Soft Computing, vol. 25, pp. 3195-3223, 2021.
[102] H. Mrabet, S. Belguith, A. Alhomoud, and A. Jemai, "A survey of IoT security based on a layered architecture of sensing
and data analysis," Sensors, vol. 20(13), p. 3625, 2020.
[103] Z. Chen, J. Liu, Y. Shen, M. Simsek, B. Kantarci, H. T. Mouftah and P. Djukic, "Machine learning-enabled IoT security:
Open issues and challenges under advanced persistent threats," ACM Computing Surveys, vol. 55(5), pp. 1-37, 2022.
[104] N. S. Akash, S. Rouf, S. Jahan, A. Chowdhury, and J. Uddin, "Botnet detection in IoT devices using random forest
classifier with independent component analysis," Journal of Information and Communication Technology, vol. 21(2), pp.
201-232, 2022.
Soofi et al. “
Securing the Internet of Things: A Comprehensive Review of Security Challenges and
Artificial Intelligence Solutions
”
Foundation University Journal of Engineering and Applied Sciences, Vol. 4, Issue 2. 20
[105] A. K. Pathak, S. Saguna, K. Mitra, and C. Åhlund, "Anomaly detection using machine learning to discover sensor
tampering in IoT systems," In ICC 2021-IEEE International Conference on Communications, 2021: IEEE, pp. 1-6.
[106] B. K. Mohanta, D. Jena, N. Mohapatra, S. Ramasubbareddy, and B. S. Rawal, "Machine learning based accident
prediction in secure IoT enable transportation system," Journal of Intelligent & Fuzzy Systems, vol. 42(2), pp. 713-725,
2022.
[107] H. Bashir, S. Lee, and K. H. Kim, "Resource allocation through logistic regression and multicriteria decision making
method in IoT fog computing," Transactions on Emerging Telecommunications Technologies, vol. 33(2), p. e3824, 2022.
[108] S. Kumar, V. Kumar-Solanki, S. Kumar Choudhary, A. Selamat, and R. González-Crespo, "Comparative study on ant
colony optimization (ACO) and K-means clustering approaches for jobs scheduling and energy optimization model in
internet of things (IoT)," International Journal of Interactive Multimedia and Artificial Intelligence, vol. 6(1), pp.107-116,
2020.
[109] M. D. Alshehri and F. K. Hussain, "A fuzzy security protocol for trust management in the internet of things (Fuzzy-IoT),"
Computing, vol. 101(7), pp. 791-818, 2019.
[110] D. Palani and K. Venkatalakshmi, "An IoT based predictive modelling for predicting lung cancer using fuzzy cluster based
segmentation and classification," Journal of Medical Systems, vol. 43, pp. 1-12, 2019.
[111] A. Anand, S. Rani, D. Anand, H. M. Aljahdali, and D. Kerr, "An efficient CNN-based deep learning model to detect
malware attacks (CNN-DMA) in 5G-IoT healthcare applications," Sensors, vol. 21(19), p. 6346, 2021.
[112] I. Ullah and Q. H. Mahmoud, "Design and development of RNN anomaly detection model for IoT networks," IEEE
Access, vol. 10, pp. 62722-62750, 2022.
[113] C. Gao, S. Braun, I. Kiselev, J. Anumula, T. Delbruck, and S.-C. Liu, "Real-time speech recognition for IoT purpose using
a delta recurrent neural network accelerator," In 2019 IEEE International Symposium on Circuits and Systems (ISCAS),
2019: IEEE, pp. 1-5.
[114] W. Zhang, D. Yang, H. Peng, W. Wu, W. Quan, H. Zhang and X. Shen, "Deep reinforcement learning based resource
management for DNN inference in industrial IoT," IEEE Transactions on Vehicular Technology, vol. 70(8), pp. 7605-7618,
2021.
[115] I. Garcia-Magarino, R. Muttukrishnan, and J. Lloret, "Human-centric AI for trustworthy IoT systems with explainable
multilayer perceptrons," IEEE Access, vol. 7, pp. 125562-125574, 2019.
[116] M. Roopak, G. Y. Tian, and J. Chambers, "Deep learning models for cyber security in IoT networks," In 2019 IEEE 9th
Annual Computing and Communication Workshop and Conference (CCWC), 2019: IEEE, pp. 0452-0457.
