Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology

ISSN No- 2584-2706

Machine Learning in IoT Security: Current

Issues and Future Prospects
Gabriel Tosin Ayodele
Faculty of Engineering and
Informatics/ University of Bradford
Bradford West Yorkshire UK

Abstract:
The Internet of Things (IoT) connects techniques that can be used to improve IoT
billions of intelligent devices that can security solutions. It describes the many
communicate with each other without possible attack surfaces of IoT systems, the
human intervention. With an estimated 50 potential risks associated with each surface,
billion devices by the end of 2020, it is one and lists unique or newly established IoT
of the fastest growing areas of computing security threats. It then details the ML / DL
history. On the other hand, IoT technology approach to IoT security, highlighting the
is essential to the advancement of various potential, strengths, and weaknesses of each
real-world intelligent applications that can method. Learn about the possibilities and
improve people's quality of life. However, challenges of using machine learning and
the interdisciplinary components involved in deep learning for IoT security. These
networking and deploying IoT systems raise opportunities and challenges may be used as
new security concerns. Encryption, future research paths.
authentication, access control, network Keywords—Deep Learning, Machine
security, and application security solutions Learning, Internet of Things Security,
are worthless when it comes to IoT devices Security based Intelligence, IoT Big Data,
and their inherent shortcomings. Therefore, Attacks, Privacy
existing security measures need to be
updated to properly protect the IoT I. Introduction
environment. Machine learning and deep The Internet of Things (IoT) is defined as a
learning (ML / DL) have evolved network of interconnected and dispersed
dramatically in recent years, and machine embedded systems that interact via wired or
intelligence has evolved from laboratory wireless communication methods (Novo et
curiosity to viable machines. An important al. , 2015). It is also described as an
defense against new or zero-day attacks is environment network of physical objects,
the ability to intelligently monitor IoT having comparatively little computational
devices. ML / DL is a powerful data capacity for data acquisition, processing,
exploration technique for revealing memory, and communication, incorporating
"normal" and "bad" behavior in the context electronics (including sensors and
of IoT components and devices. Therefore, actuators), software programs, and
machine learning and deep learning networking. IoT items are smart appliances
technologies are important not only to like smart light bulbs, smart adapters, smart
enable secure device communication, but meters, smart refrigerators, smart ovens, air
also to transform IoT security into a conditioners, temperature sensors, smoke
security-based intelligence system. The detectors, IP cameras, etc Clients that have
purpose of this study is to provide a updated items such as smartphones will also
comprehensive overview of recent advances comes under this category, Examples of
in machine learning and deep learning modern smart appliances include. This
global scale bring new difficult situations to

IJMSRT24SEP033 www.ijmsrt.com 213

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

control the devices, the volume of generated ones. For successful and safe systems, the
data, storage and communications, and, of IoT systems need to move from just enabling
course, processing and security and secure communication of the devices to
privacy.There is great literature available in security intelligent supported.
the following domains of the IoT:
architecture, communications, protocols, Figure 1
applications, security, and privacy, among Illustration of the potential role of ML/DL
others (Alfuqaha et al., 2015) (Granjal, in IoT security
Monteiro, Silva, 2015).software, and The key research question of this survey is
network connections. Smart appliances such listed as follows:
as smart light bulbs, smart adapters, smart • In earlier studies, what strategies were
meters, smart refrigerators, smart ovens, air employed to solve IoT security issues
conditioners, temperature sensors, smoke using Machine Learning?
detectors, and IP cameras are examples of • What are the different IOT security
IoT items, as are more modern devices such treats?
as smartphones. The large scale of IoT • How might the arrangement of those
networks creates new challenging featured treat be handled utilizing
conditions in terms of device control, sheer machine learning?
volume of data, storage, communication,
processing, and protection and privacy. II. Iot Security Threats
Extensive research has been conducted on IoT devices, on the whole, function in a
several areas of the IoT (architecture, variety of settings to achieve a variety of
communications, protocols, applications, goals. As a result, balancing security
security, privacy, etc. ) (Alfuqahaet al. , requirements with the IoT framework's large
2015) (Granjal, Monteiro, Silva, 2015). This attack surface is difficult. As a result, an
is because the amount of data produced by unauthorized person might get access to
IoT devices is extraordinarily large, and these devices. Because of their limited
conventional data acquisition, storage, and computation and power capabilities, IoT
analysis methods may not operate at this devices are unable to maintain complicated
level. In addition, the proliferation of data by security frameworks (Abomhara and Kien,
IoT paves a new channel that traditional data 2015). Obtaining the IoT framework in this
processing systems cannot exhaust. In this manner is a confusing and challenging
respect, Machine Learning (ML) is effort.
considered one of the most suitable
paradigms of computing for offering local A. Threats in IoT
intelligence at IoT devices (Saeid, Rezvan There are two types of security threats: cyber
and Barekatain, 2018). It can also be defined and physical. Cyber risks are further divided
as an ability of a smart device to modify or into passive and active categories. The
manage a state or perform an action as a dangers are briefly discussed in the next
result of an existing or gained awareness, sections.
which is also believed to be an essential part
of an IoT system. From Figure 1 it is 1) Cyber Threats
observed that the credence of monitoring Passive threats: Eavesdropping through
IoT devices can provide an intelligent level communication channels or the network is a
answer to new or zero raw attacks. Secondly, passive danger. An attacker can use
due to the ability to learn instances of eavesdropping to obtain data from sensors,
unknown new assaults and forecast them, follow sensor owners, or do both. On the
ML/DL algorithms may help anticipate new black market, the collection of valuable
attacks as they are mutations of previous personal information, particularly personal

IJMSRT24SEP033 www.ijmsrt.com 214

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

health data, has become commonplace viewpoints, including correspondent

(Restuccia, Oro and Melodia, 2018). security, data security, protection, technical
Active threats: In active threats, the attacker security, character the executives, malware
is capable of not only listening in on research, and so forth (Ray et al., 2016).C.
channels of communication but actively Holes in the Current Security Solution for
manipulating IoT systems to change IoT Networkse commercial acceptance of
settings, regulate communication, refuse IoT administrations and applications. The
services, and so on. For illustration, active current Internet is a magnet for security
attacks on an IoT system might include attacks ranging from simple hacks to
impersonation (e.g., faking, Sybil, and man- corporate-level all-around constructed
in-the-middle), fraudulent inputs, data security breaches that have negatively
manipulation, and denial of service. impacted a variety of businesses, including
medical services and commerce. The
11) Physical Threats limitations of IoT devices, as well as the
Regardless of how distinct improvements environment in which they operate, provide
are used at the real layer for IoT, the idea of additional challenges for the security of both
actual follows commonly seem like and applications and devices. Until now, security
require social creating approaches. and protection concerns in the IoT area have
Furthermore, in order to launch genuine been widely investigated from many
attacks, aggressors must be in close perspectives, such as correspondent
proximity to the gadgets/equipment, with security, data security, protection, technical
different objectives in mind, such as security, character the executives, malware
completely destroying the equipment, research, and so on (Ray et al., 2016). C.
limiting its lifespan, jeopardizing the Gaps in the Existing Security Solution for
communication component, changing the IoT Networks Fundamental issues of
energy source, and so on. Furthermore, such security and privacy need to be addressed to
changing devices may allow adversaries to facilitate the strategic deployment of IoT
make modifications to the directing tables hence the need to look into its drivers. In
and security keys, affecting communication particular, the word IoT has been thrown out
with upper levels (Nawaz and Loscri, 2015). of previous technologies, hence it becomes
(Yang et al. , 2011). apparent if the security threats in IoT are
new or inherited from previous technology.
B. Security Challenges in IoT Deployment According to (Fernandes and Eykholt,
Security and protection have become two 2017), authors studied and discussed the
important factors affecting the commercial comparative analysis of threats that are
adoption of IoT administrations and relevant for IoT and traditional IT gadgets.
applications. The current Internet is a They also focused mostly on the issues
security attack ground that spans from basic affecting privacy. Software, hardware,
hacking, through hacks to all around network, and applications are the key
business constructed hacks that have had a driving forces in arguing referential
negative effect on several corporations such similarities and differences. These
as the medical and commerce services. categorization show that security threat in
There are additional factors concerning IoT the conventional IT and IoT are essentially
application and device security, stemming similar. But the IoT’s primary concern is
from the device limitations as well as from resource constraints, and the problem of
the context in which IoT runs. So far, applying advanced security solutions on IoT
security and protection issues in the IoT networks becomes challenging.
context have been examined
comprehensively from numerous

IJMSRT24SEP033 www.ijmsrt.com 215

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

III. To provide some context, they split Iii. Iot Security and Machine Learning.
out two particular areas of focus: In this section, we discuss various machine
IOT Security and Machine LearningIn this learning algorithms and their applicability in
section, we briefly describe main machine IoT applications. A. Basic Machine
learning algorithms and how they may be Learning Algorithms The ML algorithms
used in IoT systems.A. Basic of Machine can be categorized as supervised,
Learning Algorithmstical components in the unsupervised, semi-supervised and
commercial acceptance of IoT Reinforcement learning algorithms.
administrations and applications. The Supervised Learning: Supervised learning
current Internet is a magnet for security is utilized when specified objectives are
attacks ranging from simple hacks to stated to be achieved from a given input. The
corporate-level all-around constructed data is labeled initially in this type of
security breaches that have negatively learning.
impacted a variety of businesses, including Unsupervised Learning: The environment
medical services and commerce. The just provides inputs for unsupervised
limitations of IoT devices, as well as the learning, with no specific goals in mind. It
environment in which they operate, provide may investigate similarities among
additional challenges for the security of both unlabeled data and organize it into various
applications and devices. Until now, security categories without requiring labeled data.
and protection concerns in the IoT area have Semi-supervised Learning: In the preceding
been widely investigated from many two kinds, either all of the observations in
perspectives, such as correspondent the dataset have no labels or all of the
security, data security, protection, technical observations have labels. Semi-supervised
security, character the executives, malware learning is somewhere in the middle.
research, and so on (Ray et al., 2016). Reinforcement Learning: No explicit goals
C. Gaps in the Existing Security are established in Reinforcement Learning
Solution for IoT Networks. It is critical to (RL), and the operator learns via feedback
investigate the fundamental causes of after observing the environment. It does
security and privacy concerns in order to some acts and made a decision based on the
successfully use IoT. More specifically, the reward it receives.
word IoT has been hurled out of previous D. Machine Learning Techniques Used
technologies, therefore it's critical to identify in IoT Security
whether the security concerns in IoT are new As indicated in Table I, we will explore
or a rehash of prior technologies' legacy. several machine learning techniques
(Fernandes and Eykholt, 2017) compared concentrating on the underpinning security
and contrasted the security concerns faced and privacy issues in IoT networks.
by IoT and conventional IT devices. They Authorization, threat detection and
also concentrated on concerns about mitigation, Dispersed Denial of Service
privacy. Software, hardware, network, and (DDoS) assaults, anomaly and detection
applications are the key driving forces in techniques, and malware analysis are all
arguing about similarities and differences. things we examine.
The security challenges in the conventional
IT sector and the IoT are fundamentally TABLE I
comparable, according to these Machine Learning Techniques Used In
categorization. However, the IoT's main Iot Security
worry is resource limits, which make it Machine Learning Algorithm Description
difficult to adapt existing advanced security Naivebayes
solutions to IoT networks. It's a classification technique that can be
applied in both binary and multi-class

IJMSRT24SEP033 www.ijmsrt.com 216

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

environments. Instead of computing the most common machine learning techniques

actual values, all of the qualities are on their own are not efficient or scalable
considered to be conditionally independent enough to deal with IoT data which require
(Zhou et al., 2017). drastic modifications (Qiu et al., 2016). In
K-nearest neighbour It is a simple and addition, IoT data is inherently imprecise by
effective supervised learning model that is nature, and it is impossible to mitigate this
used to associate new data points with stochasticity. The following sections discuss
existing comparable data points by some of the most obvious evils of using
exploring the given dataset. (Bekara, 2020). machine learning algorithm in IoT networks.
K-means algorithm The K-means
clustering method, which belongs to the Processing power and energy:
unsupervised category of the ML family, is The Internet of Things devices are generally
the most widely utilized well-known small, and their computing abilities are
approach. K-Means clustering is a technique limited due to energy constraints. Smart IoT
for classifying or grouping devices into K devices require real time data processing for
groups based on features or parameters real time applications, while the normal
(Kolias and Kambourakis, 2017). machine learning solutions are not designed
to handle continuous streams of data in real
Random forest and decision tree(dt) time. Indeed, this frequency resulted in a
It is a form of learning that is supervised. It number of issues, which arise when using
creates a model by putting particular rules traditional procedures designed for a vast
into action based on data attributes. This amount of data. Lastly, with increasing data
model is then used to forecast the value of a complexities, the amount of forecast an
new targeted variable. In classification and algorithm can do also decreases (Heureux &
regression issues, decision trees are utilized. Member, 2017).
These trees are essentially used to divide a
dataset into numerous branches depending Data management and analytics:
on particular principles (Roman, Zhou and Wireless data can originate from other
Lopez, 2013). networked in formation systems, sensing
Support Vector Machines (Svm)Svm is a and communication devices and so on
low-complexity supervised machine among them (Bogale, Wang and Le, 2018).
learning technique for classification and Data generated in IoT networks are of syntax
regression (Bekara, 2020). and semantic diverse form, format, and
Deep learning It's a feed forward Neural semantics leading to syntactic and semantic
Network (NN) in which each neuron is heterogeneity. Syntactic variation is evident
linked to another layer and there are no in data types, file formats, encoding systems
connections between the layers. Deep and data models. In the case of huge
learning describes numerous layers that volumes of data and different types of data
retain various levels of perception, with each sets with different characteristics, such
layer receiving input from the previous layer heterogeneity becomes a problem in terms
and feeding the result to the next one of generalizable consistency.
(Azmoodeh, Dehghantanha and Member,
2019). iii. Machine learning: a solution to iot
security challenges
E. Limitations in Applying Machine From then, machine learning is essentially
Learning in IoT Networks smart ways of optimizing execution rules
It covers the following typical features: huge with the help of model’s insights or previous
and diverse amount of IoT traffic as well as occurrences. In addition, it is used practical
its high and random speed. Unfortunately, applications including Google having

IJMSRT24SEP033 www.ijmsrt.com 217

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

deviced machine learning to analyze threats B. Adding ML and DL algorithms to IoT

to its android point and apps. However, in applications raises new concerns. These
the case of DL, a type of ML, the model issues are complicated. For example, it is
itself can come up with a way to decide trying to cultivate an acceptable pattern of
about the precision of an expectation on its responding to data derived from different
own. another perspective of DL models is IoT applications. Furthermore, identifying
presented as more suitable for arrangement information information is also a process
and expectation undertakings in progressive which consumes much time. One other test
IoT applications due to their Self-organizing is related to the least highlighted aspects of
nature with logical and individual educational experience. Other concerns
assistance. While traditional methodology is emerge when these models are organized on
found prevalent in different strands of IoTs asset-required IoT devices where
for example applications, administrations, minimizing handling and accumulation is
designs, conventions, data gathering, asset even more important higher (Bekara, 2020).
allocation, categorization, and analysis Further the variations arising out of ML or
including security the valley associated with DL computations are too excessive to be
IoT demand very shrewd, reliable and handled by the basic architectural schema
consistent approaches. To this aim, ML and and repetitively implemented. In this regard,
DL are interesting approaches for IoT it is crucial to elaborate the IoT security
networks for several reasons, first of all, measures concerning the preceding, with
because IoT networks produce a large respect to the impact on the ML and DL.
volume of data. In addition, the ML and DL
methodologies allow IoT frameworks to Conclusion:
make intelligent and wise decisions The buying specifications of IoT devices
increasing the value of data created by the add layer of complexity to acquire and
IoT. DL methods may also be used in IoT integrate the different innovations that are
devices to perform complex detect and needed from physical, transport distant
acknowledge operations, making it possible transmission, flexibility design, to cloud. In
to admit new applications and/or recent years machine learning and deep
administrations that take into consideration learning have ushered in a number of strong
constant working association with humans, scientific process that can be applied to
smart devices and environment literals. The enhance IoT security. The following IoT
following are a few actual security-related security threats and attack surfaces are
uses of machine learning: described in this report. It offers a detail
analysis of the envisioned uses of machine
• Face recognition for criminology: learning and deep learning in the protection
appearance, lighting, obstacle (glasses, of IoT.
facial hair), make-up, hairstyle, and so
on. References
• Different penmanship styles for character Abomhara, M. and Køien, G. M. (2015)
recognition for security encryption. ‘Cyber Security and the Internet of Things :
• Proof of malevolent code: identifying Vulnerabilities , Threats , Intruders’, 4, pp.
noxious code in applications and 65–88. doi: 10.13052/jcsm2245-1439.414.
programs Ahmed, N., De, D. and Hussain, I. (2018)
• Distributed Denial of Service ‘Internet of Things (IoT) for Smart Precision
(DDoS)detection:DDoS assaults on Agriculture and Farming in Rural Areas’,
infrastructure can be detected via IEEE Internet of Things Journal, 5(6), pp.
behavior analysis.. 4890–4899. doi:
10.1109/JIOT.2018.2879579.

IJMSRT24SEP033 www.ijmsrt.com 218

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Al-fuqaha, A. et al. (2015) ‘Internet of on resource-constrained IoT devices using

Things : A Survey on Enabling’, 17(4), pp. recurrent neural networks’, pp. 60–67.
2347–2376. Du, H. et al. (2020) ‘The Elements of End-
Altawy, R., Youssef, A. M. R. M. and to-end Deep Face Recognition: A Survey of
Member, S. (2016) ‘Security Tradeoffs in Recent Advances’. Available at:
Cyber Physical Systems : A Case Study http://arxiv.org/abs/2009.13290.
Survey on Implantable Medical Devices’, Fernandes, E. and Eykholt, K. (2018)
IEEE Access. IEEE, 4, pp. 959–979. doi: ‘Internet of Things Security Research ’:, pp.
10.1109/ACCESS.2016.2521727. 1–5.
Ammar, M., Russello, G. and Crispo, B. Fosso, S., Anand, A. and Carter, L. (2013)
(2018) ‘Journal of Information Security and ‘International Journal of Information
Applications Internet of Things : A survey Management A literature review of RFID-
on the security of IoT frameworks’, Journal enabled healthcare applications and issues’,
of Information Security and Applications. International Journal of Information
Elsevier Ltd, 38, pp. 8–27. doi: Management. Elsevier Ltd, 33(5), pp. 875–
10.1016/j.jisa.2017.11.002. 891. doi: 10.1016/j.ijinfomgt.2013.07.005.
An, N. et al. (2017) ‘Behavioral Anomaly From, C. (2009) ‘Securing Wireless
Detection of Malware on Home Routers’. Implantable Devices for Healthcare : Ideas
Arulkumaran, K. et al. (2017) ‘A Brief and Challenges’, (July), pp. 74–80.
Survey of Deep Reinforcement Learning’, Granjal, J., Monteiro, E. and Silva, J. S.
pp. 1–16. (2015) ‘Pr E oo f Pr E oo f’, pp. 1–20.
Azmoodeh, A., Dehghantanha, A. and Heureux, A. L. and Member, G. S. (2017)
Member, S. (2019) ‘Robust Malware ‘Machine Learning With Big Data :
Detection for Internet Of ( Battlefield ) Challenges and Approaches’, IEEE Access.
Things Devices Using Deep Eigenspace IEEE, 5, pp. 7776–7797. doi:
Learning’. 10.1109/ACCESS.2017.2696365.
Banerjee, B. A. et al. (2011) ‘Ensuring Imran, M et al. (2017) ‘The rise of
Safety , Security , and Cyber – Physical ransomware and emerging security
Systems’. challenges in the Internet of Things’,
Bekara, C. (2020) ‘Security Issues and Computer Networks. Elsevier B.V. doi:
Challenges for the IoT-based Smart Grid’, 10.1016/j.comnet.2017.09.003.
Procedia - Procedia Computer Science. Items, R. et al. (2018) ‘This is a repository
Elsevier Masson SAS, 34, pp. 532–537. doi: copy of A Deep Recurrent Neural Network
10.1016/j.procs.2014.07.064. Based Approach for Internet of Things
Bertino, E. (2020) ‘Botnets and Internet’. Malware Threat Hunting . White Rose
Bogale, T. E., Wang, X. and Le, L. B. (2020) Research Online URL for this paper :
‘Machine Intelligence Techniques for Next- Version : Accepted Version Article :
Generation Context-Aware Wireless HaddadPajouh , H ., Dehghantanha , A orcid
Networks’. . org / 0000-0002-9294-7554 , Khayami , R
Bogaz, B. et al. (2017) ‘Journal of Network . et al . ( 1 more author ) ( 2018 ) A Deep
and Computer Applications A survey of Recurrent Neural Network Based Approach
intrusion detection in Internet of Things’, for Internet of Things Malware Threat
Journal of Network and Computer Hunting . Future Generation Computer
Applications. Elsevier, (January), pp. 0–1. Systems , 85 . pp . 88-96 . ISSN 0167-739X’.
doi: 10.1016/j.jnca.2017.02.009. Kolias, C. and Kambourakis, G. (2017)
Chauhan, J. et al. (2018) ‘Institutional ‘DDoS in the IoT ’:, (January). doi:
Knowledge at Singapore Management 10.1109/MC.2017.201.
University Breathing-based authentication

IJMSRT24SEP033 www.ijmsrt.com 219

DOI: https://doi.org/10.5281/zenodo.13846031
Volume2, Issue9, Sep 2024 International Journal of Modern Science and Research Technology
ISSN No- 2584-2706

Lake, B. M. et al. (2012) ‘Building Restuccia, F., Oro, S. D. and Melodia, T.

Machines That Learn and Think Like (2018) ‘Securing the Internet of Things in
People’, (2012), pp. 1–58. the Age of Machine Learning and Software-
Lane, N. D., Georgiev, P. and Qendro, L. defined Networking’, 1(1), pp. 1–14.
(2015) ‘DeepEar : Robust Smartphone Roman, R., Zhou, J. and Lopez, J. (2013)
Audio Sensing in Unconstrained Acoustic ‘On the features and challenges of security
Environments using Deep Learning’. and privacy in distributed internet of things’,
Mahmud, M. et al. (2018) ‘Applications of COMPUTER NETWORKS. Elsevier B.V.
Deep Learning and Reinforcement Learning doi: 10.1016/j.comnet.2012.12.018.
to Biological Data arXiv : 1711 . 03985v2 [ Saeid, M., Rezvan, M. and Barekatain, M.
cs . LG ] 7 Jan 2018’, pp. 1–33. doi: (2018) ‘Machine learning for internet of
10.1109/TNNLS.2018.2790388.c. things data analysis : a survey’, Digital
Mishra, P. et al. (2018) ‘A Detailed Communications and Networks. Elsevier
Investigation and Analysis of using Machine Ltd, 4(3), pp. 161–175. doi:
Learning Techniques for Intrusion 10.1016/j.dcan.2017.10.002.
Detection’, IEEE Communications Surveys Sethi, P. and Sarangi, S. R. (2017) ‘Internet
& Tutorials. IEEE, PP(c), p. 1. doi: of Things : Architectures , Protocols , and
10.1109/COMST.2018.2847722. Applications’, 2017.
Mohammadi, M. et al. (2017) ‘Semi- Wallace, B. Y. T. C. et al. (2016)
supervised Deep Reinforcement Learning in ‘ΒΙΟΕΚΧΥΛΙΣΗ ΟΞΕΙΔΩΜΕΝΩΝ
Support of IoT and Smart City Services’, ΜΕΤΑΛΛΕΥΜΑΤΩΝ ΝΙΚΕΛΙΟΥ ΜΕ ΤΗ
X(X), pp. 1–11. ΧΡΗΣΗ ΕΤΕΡΟΤΡΟΦΩΝ
Nawaz, A. H. and Loscri, V. (2014) ‘The ΜΙΚΡΟΟΡΓΑΝΙΣΜΩΝNo Title’, Bulletin
Internet of Things-A survey of topics and of the Seismological Society of America,
trends’. doi: 10.1007/s10796-014-9489-2. 106(1), pp. 6465–6489. Available at:
Nguyen, N. D. U. Y. et al. (2017) ‘System http://www.bssaonline.org/content/95/6/237
Design Perspective for Human-Level 3%5Cnhttp://www.bssaonline.org/content/9
Agents Using Deep Reinforcement 5/6/2373.short%0Ahttp://www.bssaonline.o
Learning : A Survey’, pp. 27091–27102. rg/cgi/doi/10.1785/0120110286%0Ahttp://g
Novo, O. et al. (2015) ‘Capillary Networks ji.oxfordjournals.org/cgi/doi/10.1093/gji/gg
– Bridging the Cellular and IoT Worlds’, v142%0Ahttp://link.springer.com/10.1007/
(December). doi: 10.1109/WF- s00024-01.
IoT.2015.7389117. Wang, T. et al. (2018) ‘Deep Learning for
Perera, C. et al. (2017) ‘Context Aware Wireless Physical Layer : Opportunities and
Computing for The Internet of Things : A Challenges’, pp. 1–14.
Survey’, X(X), pp. 1–41. Yang, Z. et al. (2011) ‘Study and
Qiu, J. et al. (2016) ‘A survey of machine Application on the Architecture and Key
learning for big data processing’, EURASIP Technologies for IOT’, pp. 747–751.
Journal on Advances in Signal Processing. Zhang, D., Han, X. and Deng, C. (2018)
EURASIP Journal on Advances in Signal ‘Review on the Research and Practice of
Processing. doi: 10.1186/s13634-016-0355- Deep Learning and Reinforcement Learning
x. in Smart Grids’, 4(3), pp. 362–370. doi:
Ray, S. et al. (2016) ‘The Changing 10.17775/CSEEJPES.2018.00520.
Computing Paradigm with Internet of Zhou, J. et al. (2017) ‘Security and Privacy
Things ’:, 2356(c), pp. 1–13. doi: for Cloud-Based IoT : Challenges ,
10.1109/MDAT.2016.2526612. Countermeasures , and Future Directions’,
Razzaque, M. A., Milojevic-jevric, M. and (January), pp. 26–33.
Palade, A. (2015) ‘Middleware for Internet
of Things : a Survey’, 0(0), pp. 1–27.

IJMSRT24SEP033 www.ijmsrt.com 220

DOI: https://doi.org/10.5281/zenodo.13846031