Content uploaded by Muhammad Adnan Khan
Author content
All content in this area was uploaded by Muhammad Adnan Khan on Jul 04, 2021
Content may be subject to copyright.
223
IEEE Network • May/June 2021 0890-8044/21/$25.00 © 2021 IEEE
AbstrAct
Realizing secure and private communications
on the Internet of Things (IoT) is challenging, pri-
marily due to IoT’s projected vast scale and exten-
sive deployment. Recent efforts have explored
the use of blockchain in decentralized protection
and privacy supported. Such solutions, however,
are highly demanding in terms of computation
and time requirements, barring these solutions
from the majority of IoT applications. Specifically,
in this paper, we introduce a resource-efficient,
blockchain-based solution for secure and private
IoT. The solution is made possible through novel
exploitation of computational resources in a typ-
ical IoT environment (e.g., smart homes), along
with the use of an instance of Deep Extreme
Learning Machine (DELM). In this proposed
approach, the Smart Home Architecture based in
Blockchain is protected by carefully evaluating its
reliability in regard to the essential security aims
of privacy, integrity, and accessibility. In addition,
we present simulation results to emphasize that
the overheads created by our method (in terms
of distribution, processing time, and energy con-
sumption) are marginal related to their protection
and privacy benefits.
IntroductIon
A smart home is an Internet of Things (IoT) inte-
grated residence that offers consumers security,
health, comfort, improved standard of living, and
so on. Smart home systems are capable of mak-
ing people’s lives and independent living easier
and better. They provide useful tools like track-
ing behaviors and safety assessments that have
attracted users’ and device developers’ attention.
Although intelligent homes offer huge benefits
to homeowners and interested parties, they are
potentially at risk for malicious cyber-attacks that
may endanger users’ safety and privacy. Such
threats have conventional solutions, which are
highly centralized and vulnerable to fierce attacks.
Consequently, the versatility and scalability
required for proper use in the innovative area of
autonomous smart home applications and facil-
ities are lacking. Several intelligent technologies
render people’s lives simpler [1]. Such programs
produce huge amounts of information. The stor-
ing of this constantly evolving data into reposito-
ries creates safety concerns.
Nonetheless, blockchain has attained outstand-
ing performance as a cornerstone of the cyber-
security infrastructure in a variety of smart home
technologies like remote connectivity and data
transmission. Blockchain technologies and cen-
tralized storage networks may be used to address
these problems. Blockchain was invented in 2008
by Satoshi Nakamoto and included a timestamped
collection of malice-proof documentation con-
trolled by a community of decentralized systems
[2]. Inflexibility, decentralization, and openness
are the keystones of blockchain technology. The
three functions have extended their doors to a
wide range of applications, including the nature
of digital currencies and the feasibility review of
intelligent applications, while blockchain technol-
ogy guarantees security. For instance, the form
of attacks has recently grown more complex, like
most of the attacks controlling vote, Sybil attacks
for false identity creation to monitoring consen-
sus.
Because conventional approaches utilize a sig-
nature-based method to identify unique arrange-
ments, a comprehensive Intrusion Detection
System (IDS) is essential to resolve the underlying
issue. However, one of the latest technologies,
known as the Deep Extreme Learning Machine
(DELM) [1], can be used to evaluate data flow to
spot intrusions and attack patterns. Therefore, it
is important to manage smart blockchain-based
applications by developing powerful and versa-
tile algorithms to process this large amount of
data. Machine learning involves machines for
training, reasoning, and behaving without human
intervention. It is known as one of the Artificial
Intelligence (AI) frameworks. The basic goal of
machine learning is to create an effective algo-
rithm to take data from the input and make a pre-
diction and change the outputs through statistical
analysis. Machine learning may process a signifi-
cant amount of information and make decisions
guided by evidence.
This research will, therefore, employ a Deep
Extreme Learning Machine (DELM) approach to
make intelligent homes safer with IoT-enabled
sensors and enhanced performance. For this
study, the major contributions are to provide a
thorough review of cutting-edge technologies rel-
evant to blockchain-based smart homes empow-
ered with the Deep Extreme Learning Machine,
to offer a new viewpoint on various applications
(e.g., smart home data sharing), which is backed
by the recent stages of technology growth. Deep
Extreme Learning Machine architecture has been
proposed for implementation in blockchain-based
A Machine Learning Approach for Blockchain-Based Smart Home Networks Security
Muhammad Adnan Khan, Sagheer Abbas, Abdur Rehman, Yousaf Saeed, Asim Zeb, M. Irfan Uddin, Nidal Nasser, and Asmaa Ali
ACCEPTED FROM OPEN CALL
Digital Object Identifier:
10.1109/MNET.011.2000514
Muhammad Adnan Khan is with Riphah International University Lahore Campus; Sagheer Abbas and Abdur Rehman are with NCBA&E;
Yousaf Saeed is with the University of Haripur; Asim Zeb is with Abbotabad University of Science and Technology; M. Irfan Uddin is with
Kohat University of Science and Technology; Nidal Nasser is with Alfaisal University; Asmaa Ali is with Queen’s University.
NASSER_LAYOUT.indd 223NASSER_LAYOUT.indd 223 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021
224
smart homes, as demonstrated in Fig. 1. The smart
application here gathers information from vari-
ous information sources along with sensors, smart
devices, and IoT devices. Information obtained
from such smart applications is analyzed and pro-
cessed. The blockchain acts as an important part
of those applications. The Deep Extreme Learning
Machine will then be used to interpret and pre-
dict data (data analytics and real-time analytics)
in such applications. The datasets utilized by the
DELM framework could be processed in a block-
chain network, which eliminates errors in informa-
tion such as repetition, loss of data value, errors,
and disruption. Blockchains rely on the data. Thus
data-related problems in the DELM framework
will be excluded. The DELM framework can be
focused on different chain segments instead of
the entire collection of datasets. This will provide
a unique framework for various applications, such
as detecting fraud and predicting theft detection.
The sections of this article are structured as fol-
lows. The related survey articles are summarized
in the following section. We then explain the fun-
damental framework for blockchain, implementa-
tion of the DELM approach on blockchain-based
smart home, and smart home application frame-
work. Following that we address the DELM
approach’s simulation and tests. The research
conclusions are explored in the fi nal section.
reLAted work
C. Badii et al. [3] stated that in all frameworks,
such as electricity, water, and transport, the
dream of intelligent cities is secure, safe, green,
and effi cient. G. S. Aujla et al. [4] stated that the
smart city is an innovation-focused and better city
linking human beings, data, and urban characteris-
tics through innovative technology, thereby result-
ing in a stable, safe city with effi cient and creative
businesses and an improved quality of life. S. Tan-
war et al. [5] argued that the smart city is a clev-
er society in which various components, such as
people, the environment, mobility, democracy,
and the economy, are installed in a smart system.
G. Li et al. [6] proposed a user-centric block-
chain system to protect the exchange of edge
information in IoT. Z. Zhou et al. [7] introduced
a stable and effective energy sharing system for
vehicles to grids by investigating blockchain tech-
nology, contract theory, and edge computing. J.
Wu et al. [8] proposed a software-defi ned block-
chain platform to understand complex blockchain
architectures, and implemented a consensus-func-
tion solution to virtual machines with the appli-
cation-aware framework which can derive and
handle specifi c consensus tools.
This research intends to support people’s
transition to the Internet of Things, in particular,
in emerging smart cities. According to the Inter-
net of Things European Research Cluster, IoT
is a vibrant, worldwide network infrastructure
that automatically configures physical and virtu-
al objects and communicates them through the
standard and interoperable procedure. IoT would
be the best hope for a viable city [9]. To make
IoT intelligent, many computational technologies
have been implemented into it. Only data mining,
artifi cial intelligence, cloud computing, and neural
networks are among the most important technol-
ogies.
M. Moore et al. [10] expects that 70 percent
of people will live in towns by 2050. The intelligi-
FIGURE 1. A 4-layer application framework of a blockchain-based smart home empowered with DELM.
'
NASSER_LAYOUT.indd 224NASSER_LAYOUT.indd 224 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021 225
bility concept generates the comfort of life. Smart
City is an initiative that utilizes state-of-the-art tech-
nology to improve environmental sustainability.
Intelligent communities often boost environmen-
tal efficiency and offer better facilities to people.
Connectivity and information technology are nec-
essary to promote and turn cities into smart cities.
Because blockchain is becoming increasingly
popular in a smart home, a variety of research
papers have been published. For example, in a
study on the use of blockchain technologies in
the smart world, S. Aggarwal et al. [11] addressed
agreement, access control, home care, payments
integration, market sharing, and numerous smart
city facilities. Nevertheless, the implementation
of blockchain on the technology market tends
to be understudied in data analytics focused on
intelligent home services. Many regard this as the
potential usage of blockchain for smart homes. M.
Andoni et al. [12] identified the comprehensive
analysis of various blockchain implementations
of the Peer-to-Peer (P2P) energy trading network.
The research discusses, in great detail, the appli-
cation of the global agreement process and other
features of different smart home systems, such
as safety problems, the smart grid, Artificial Intel-
ligence (AI), payment processing, and Big Data
analytics. Although the technical details were spe-
cifically studied in their report, the sample omit-
ted many conventional ramifications from a smart
home perspective, such as smart home care and
smart cities financial management.
ProPosed deLM FrAMework
bAckground
In 2008, Satoshi Nakamoto developed blockchain
[2]. A P2P payments network will remove third
parties and double-spent issues by using its simple
blockchain cryptocurrency (e.g., bitcoins). This
is a clustered data system in which every data
block is authenticated by SHA-256 (Secure Hash
Algorithm) with the previous hash block. Block
number, prior block hash, transaction informa-
tion, nonce, and time stamps constituted the fun-
damental structure of the block. The timestamp
comprises a continuous variable, but in the case
of a nonce, the variable is random. The valida-
tor and the miner (computer node) constantly
hash down the static (block) and dynamic (time-
stamp and nonce) data for a value that begins
with many consecutive leading zeroes. This meth-
od is commonly called a cryptography puzzle. In
the blockchain, which miners identify the right
hash value, will first consider the winner who is
permitted to insert the block. Proof of work is the
technique used to validate whether or not a block
is valid. The measures below identify key features
of blockchain. In the case of a smart home, every
node linked with an IoT device interacts with a
memory pool, including miners in a blockchain
system. The memory pool contains all transactions
waiting to be used in the blockchain to create a
new block. All transactions are checked and sum-
marized by the Merkle tree.
If transactions are viable, then extracted trans-
actions are inserted into the block ready for min-
ing by miners throughout the intelligent home
system. By modifying the nonce and time label,
miners create a Hash of Block. The program
afterward tries comparing the hash produced to
the target. The hash is linked to the chain once
a miner has finished mining the block. When the
hash reaches the target amount, the process will
restart. If the hash is smaller than the goal value,
the proof of work is tested for performance and
applied to the chain. This message is, therefore,
sent throughout the network to alert each linked
node to remove completed memo pool trans-
actions. The smart home communication envi-
ronment is increasingly formed by blockchain
technology, as it is flexible and adaptable enough
to integrate with smart home IoT apps easily.
Figure 1 demonstrates the smart home network
based on blockchain. The framework contains
four layers: a layer of IoT data sources, a block-
chain network layer empowered with Deep
Extreme Learning Machine (DELM), an intelligent
home device layer, and a client node.
IntegrAtIon oF deeP e xtreMe LeArnIng MAchIne In
bLockchAIn-bAsed sMAr t hoMe
The IoT information layer collects information
from devices that are important for the assess-
ment of smart homes, environments, and users.
Such devices are grouped into three significant
classifications: sensors, multimedia, and health-
care. Atmospheric conditions are calculated by
sensors. For instance, the thermostat is used for
room temperature calculation and regulation. The
IoT sensor network comprises closed-circuit tele-
vision, wearables, and so on, named a blockchain
layer. Information from these nodes is collected
and stored on a centrally built repository or data-
base, such as a blockchain that establishes the
stack’s first layer.
With blockchain-based applications, DELM
computing technologies can be employed to
make them smarter. Security can be improved by
using DELM for the distributed ledger. DELM can
also be utilized to increase the interval required
to achieve understanding by improved routes for
information sharing. It also constructs the chance
to build improved frameworks utilizing blockchain
technology’s centralized architecture. In block-
chain-enabled intelligent technology, we intro-
duced to the DELM implementation architecture,
as indicated in Figure 1. The smart application
here gathers information from various sources of
information, such as cameras, smart devices, and
IoT systems. Information from such methods was
evaluated as a component of smart applications.
Blockchain works as a key component of these
smart applications. Nonetheless, for interpretation
(data analysis and real-time analysis) and predic-
tion, the DELM framework can be used for such
application data. On a blockchain network, data
sets used by DELM models are processed.
Data errors like repetition, incomplete data
values, defects, and noise are minimized. Infor-
mation is distributed on a blockchain, and thus
information-related problems can be minimized
in the DELM system. Instead of the whole data
set, the DELM frame can be based on specific
chain fragments. This can include unique frame-
works for numerous applications, such as fraud
identification or prevention of identity theft. The
blockchain model is based on the edge of the IoT
architecture and consists of three key elements:
blockchain information architecture, smart con-
NASSER_LAYOUT.indd 225NASSER_LAYOUT.indd 225 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021
226
tract, and Deep Extreme Learning Machine layer.
In the proposed DELM framework, large volumes
of hidden layers, hidden neurons, and many trig-
gering functions have been used for maximizing
the security and safety of smart homes. There are
three diff erent layers in the suggested technique:
the data collection, preprocessing, and evalua-
tion stages. Two sub-layers, namely, prediction
and performance assessment, are included in
the evaluation layer. Valid sensor data and actu-
ators for experimental research are obtained.
The information collected is supplied as an input
for the acquisition layer. Specific data cleaning
and planning processes to eliminate knowledge
inconsistencies in the preprocessing layer have
been introduced. The DELM was utilized to opti-
mize the smart home network by any malicious or
intrusive activity in the application layer.
Hash values bind blocks through cryptography.
A home server machine might be perceived as a
miner for checking new transactions and inserting
new blocks while intelligent contracts obey pre-
defi ned laws to render decentralized transactions
simpler and easier. There are different ways to
incorporate blockchains, such as public, propri-
etary, and federated, but in a sophisticated home
network, the usually privately owned blockchain is
used to minimize overhead costs.
The application layer is built to promote the
convergence of current blockchain networks with
specifi c smart home applications. This framework
covers smart home technologies such as the dig-
ital market, access management, interoperability
in home and healthcare, and the automatic pay-
ment of infrastructure and intelligent community
services. Ultimately, the access layer is at the top
of the hierarchy and enables third parties to take
advantage of blockchain-based smart home tech-
nologies, like microgrids, retailing outlets, suppli-
ers of utilities, careers, and so on.
The advent of smart devices has paved the
way for creative IoT technologies to intelligently
and eff ectively run smart residential systems. The
main components of an enhanced smart home
environment include cameras, CCTV, smart TV,
fi tness devices, smartphones, and actuators. The
main services offered by a smart home with IoT
apps include remote controls, alarm generation,
safety surveillance, and so on. A customized
access control device that allows the home user
to be activated on request must be implemented
to increase the smooth operation of smart home
and detect any malicious activities by hackers.
This access authorization may be specifi ed in a set
of IoT system access records for a specifi c user. In
an unforgeable distributed information network,
such information should be stored in order to
survive malicious attackers. To demonstrate how
blockchain leads to safe access, we take, as an
illustration as follows.
The home user (Admin) who may have per-
sonalized access to the smart home and be using
apps inside. Figure 2 demonstrates how block-
chain leads to safe access. First, the user must
identify the level of access and apply it to the
home service machine. For instance, the home-
owner (Admin) is permitted at the maximum
level, while teens, children, visiting families, and
teenagers require mid-level approval. Neighbors
and outsiders have low-ranked (or zero) control
permits. After obtaining a request from the user,
the home server reviews the directory of access
control. The home server subsequently transmits
this order for the legal authentication of that par-
ticular user to the blockchain layer. A blockchain
policy header holds the access control list for mul-
tiple consumers and devices. The policy header is
a component of the block data used to execute
control policies and tools. The new user’s request
is sent to the administrator who can approve or
deny any request for entry. Once the administra-
tor gives or denies permission, blockchain min-
ers insert the header policy details and carry out
actions. This method can be used to eliminate
malicious attackers.
deeP extreMe LeArnIng MAchIne
The DELM can be used in diff erent fi elds to fore-
cast health problems, predict energy consumption,
transport, and traffic control, and so on. [1]. The
existing Artificial Neural Networks (ANN) algo-
rithms require many adjustments and long learning
cycles, and the learning system can be overwritten.
The knowledge of an extreme machine is defi ned
by G. B. Huang et al. [13]. The DELM can be wide-
ly used in diff erent domains for classifi cation and
regression objectives because DELM learns quickly
and is successful in the rate of procedural convo-
lution. The extreme learning machine is a feed-
forward neural network that suggests data travels
only one way along with a series of layers, though
in this proposed system we used the backpropa-
gation approach during the learning phase, where
information fl ows backward through the network
and in the backpropagation, the neural network
adjusts the weights to achieve high accuracy with
minimum error rate. The weights of the network
are consistent throughout the validation phases, in
which we extract the trained model and forecast
the actual data. The proposed DELM approach
FIGURE 2. Blockchain-based smart home management system empowered with
DELM.
NASSER_LAYOUT.indd 226NASSER_LAYOUT.indd 226 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021 227
consists of three layers: an input layer, hidden lay-
ers, and an output layer, respectively. The DELM
framework consists of input layers, multiple hid-
den layers, and an output layer. In an extreme
learning machine, there is only one hidden layer
and multiple neurons employed in the network to
train the dataset, but in the deep extreme learn-
ing machine, there are multiple hidden layers with
a constant number of neurons to increase the
efficiency of the network. So, by increasing the
hidden layers along with the constant number of
neurons in the network, we have achieved better
results as compared to other machine learning
algorithms as shown in Table 1. In DELM we have
incorporated the backpropagation technique with
a feed-forward technique to adjust the weights in
the network to reduce the error rate. The DELM
framework outperforms the other machine learn-
ing algorithms in terms of accuracy.
Diff erent statistical parameters, given in Figure
3, are observed to optimize smart home security
in the evaluation layer like Accuracy, Miss Rate,
Sensitivity, Specificity, False Positive Value, and
Positive Prediction Value.
The backpropagation method involves weight
configuration, feedforward propagation, back-
ward error propagation, and distinguishability
update. A sigmoid activation function is used in
the hidden layer on each neuron. This helps in
the design of the sigmoid input function and the
DELM hidden layer, which can be measured by
dividing the square sum from the required output
by 2. The adjustment in weight is needed to miti-
gate the common error.
resuLts And dIscussIon
In this paper, the Deep Extreme Learning
Machine (DELM) in the proposed framework was
implemented with input data from [14]. The data
were randomly divided into 85 percent training
(125,973 samples), and 15 percent of data is
used for validation (22,543 samples). Data have
been processed in advance to delete informa-
tion irregularities and mitigate the possibility of
the presence of information resulting from errors.
DELM attempted to identify any malice or intru-
sion in various hidden layers, hidden connections,
and activation functions. Additionally, we evaluat-
ed a certain number of neurons in hidden layers
and various types of active functions. In this anal-
ysis, we evaluated the DELM to properly predict
the effi ciency of this system. To calculate the out-
put with the counterpart algorithms of this DELM
algorithm, we used diff erent statistical measures.
Table 2 shows the proposed blockchain-based
smart home empowered with the DELM system
model prediction of intrusion detection during
the training phase. A total of 125,973 samples
are used during training. These are further divid-
ed into 67,343 and 58,630 samples of normal
and attack, respectively. It is observed that 65,366
samples of a normal class (meaning in which no
attack is found) are correctly predicted, while
1,977 records are incorrectly predicted as an
attack found while there is no actual attack exist-
ing. Similarly, a total of 58,630 samples are taken
in the case of attack found, in which 56,210 sam-
ples are correctly predicted as an attack found,
and 2,420 samples are invalid predict as a normal
found while attack exists there.
Table 3 shows the proposed blockchain-based
smart home empowered with the DELM system
model prediction of intrusion detection during the
validation phase. A total of 22,543 samples are
used during validation. These are further divided
into 9,710 and 12,833 samples of normal and
attack, respectively. It is observed that 9,237
samples of a normal class are correctly predicted
while 473 records are incorrectly predicted as
an attack while no actual attack exists. Similarly,
12,833 samples are taken in the case of attack
found, in which 11,935 samples are correctly pre-
dicted as an attack found and 898 samples are
invalidly predicted as a normal found, when in
fact an attack existed there.
Figure 3 shows the proposed block-
chain-based smart home empowered with the
DELM system model performance in terms of
diff erent statistical measures during the training
and validation phase. It is clearly shown that the
proposed system during training produces 96.51
percent and 3.49 percent accuracy and miss
rates, respectively. During validation, the pro-
posed system produces 93.91 percent and 6.09
percent accuracy and miss rates, respectively. It
also shows system model performance in terms
of sensitivity, specifi city, during the training, and
testing phase. It clearly shows that the proposed
system during training gives 96.43 percent and
TABLE 1. Comparison of the proposed DELM meth-
od with other machine learning algorithms with
diff erent datasets.
Method Accuracy of NSL-
KDD dataset [14]
Accuracy of KDD-
CUP-99 dataset [15]
ANN 81.2 % 90.39%
SVM 69.52% 89.94%
Decision Tree 81.5 % 91.12%
Proposed DELM 93.91% 94.6%
FIGURE 3. Performance evaluation of blockchain-based smart home empowered
with deep extreme learning machine system model during the prediction of
malicious activity or attacks using diff erent statistical measure.
NASSER_LAYOUT.indd 227NASSER_LAYOUT.indd 227 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021
228
96.6 percent sensitivity and specificity, respec-
tively, while it gives 91.14 percent and 96.19
percent sensitivity and specificity, respectively,
during validation. In addition, some more statis-
tical measures are added to predict the values
such as false positive, false negative, likelihood
ratio negative and positive as well as positive
and negative prediction values. All results of
these measures are given in Fig. 3.
concLusIon
Intrusion identification in smart homes remains
an outstanding challenge, especially in terms of
evaluation and prediction. Meanwhile, recent
developments in the areas of blockchain and
machine learning have shown great potential in
achieving such objectives. Given the power and
processing limitation of devices in the majori-
ty of smart home deployments, such solutions
cannot be readily applied. Addressing this void,
this work presented a lightweight yet effective
solution for intrusion identification and predic-
tion. A Deep Extreme Learning Machine (DELM)
blockchain-based architecture was proposed.
Several statistical methods were used to mea-
sure the effectiveness of the suggested solution.
Such estimation results demonstrate that the
DELM approach is far more reliable than those of
other algorithms. The proposed DELM approach
achieved spectacular results, demonstrating
93.91 percent accuracy. The obtained results are
promising, and we are currently exploring exten-
sions through the application of further datasets
and varied architectures.
reFerences
[1] S. Abbas et al., “Modeling, Simulation and Optimization of
Power Plant Energy Sustainability for IoT Enabled Smart Cit-
ies Empowered with Deep Extreme Learning Machine,” IEEE
Access, vol. 8, 2020, pp. 39982–97.
[2] S. Nakamoto et al., “Bitcoin: A Peer-to-Peer Electronic Cash
System,” Manubot, 2019.
[3] C. Badii et al., “Smart City IoT Platform Respecting GDPR
Privacy and Security Aspects,” IEEE Access, vol. 8, 2020, pp.
23601–23.
[4] G. S. Aujla et al., “BlockSDN: Blockchain-as-a-Service for
Software Defined Networking in Smart City Applications,”
IEEE Network, vol. 34, no. 2, 2020, pp. 83–91.
[5] S. Tanwar et al., “Machine Learning Adoption in Block-
chain-Based Smart Applications: The Challenges, and a Way
Forward,” IEEE Access, vol. 8, 2020, pp. 474–88.
[6] G. Li et al., “Preserving Edge Knowledge Sharing Among IoT
Services: A Blockchain-Based Approach,” IEEE Trans. Emerg-
ing Topics in Computational Intelligence, 2020, pp. 1–13.
[7] Z. Zhou et al., “Secure and Efficient Vehicle-to-Grid Energy
Trading in Cyber Physical Systems: Integration of Blockchain
and Edge Computing,” IEEE Trans. Systems, Man, and Cyber-
netics: Systems, vol. 50, no. 1, 2020, pp. 43–57.
[8] J. Wu et al., “Application-Aware Consensus Management for
Software-Defined Intelligent Blockchain in IoT,” IEEE Net-
work, vol. 34, no. 1, 2020, pp. 69–75.
[9] Y. H. Park et al., “Blockchain-Based Model for Trustworthy
Shared Internet of Things Device Management,” Proc. 2020
Int’l. Conf. Electronics, Information, and Commun. (ICEIC),
IEEE, 2020, pp. 1–2.
[10] M. Moore, P. Gould, and B. S. Keary, “Global Urbanization
and Impact on Health,” Int’l. J. Hygiene and Environmental
Health, vol. 206, no. 4-5, 2003, pp. 269–78.
[11] S. Aggarwal et al., “Blockchain for Smart Communities:
Applications, Challenges and Opportunities,” J. Network and
Computer Applications, vol. 144, 2019, pp. 13–48.
[12] M. Andoni et al., “Blockchain Technology in the Energy
Sector: A Systematic Review of Challenges and Opportuni-
ties,” Renewable and Sustainable Energy Reviews, vol. 100,
2019, pp. 143–74.
[13] G. B. Huang, D. H. Wang, and Y. Lan, “Extreme Learning
Machines: A Survey,” Int’l. J. Machine Learning and Cyber-
netics, vol. 2, no. 2, 2011, pp. 107–22.
[14] “NSL-KDD,” Kaggle.com, 2020, available: https://www.kag-
gle.com/hassan06/nslkdd; accessed: 15 Jan 2020.
[15] “Network Intrusion Detection,” Kaggle.com, 2020, avail-
able: https://www.kaggle.com/sampadab17/network-intru-
sion-detection; accessed: 26 Aug. 2020.
bIogrAPhIes
MuhaMMad adnan Khan is currently working as an associate
professor in the Department of Computer Science, Faculty of
Computing, Riphah International University Lahore Campus,
Lahore, Pakistan. He completed his Ph.D. at ISRA University,
Pakistan. Prior to joining the LGU, he worked in various aca-
demic and industrial roles in Pakistan. He has been teaching
graduate and undergraduate students in computer science and
engineering for the past 12 years. Presently he is guiding four
Ph.D. scholars and eight M.Phil. Scholars. He has published
about 160 research articles in international journals as well as
respected international conferences. His research interests pri-
marily include MUD, channel estimation in multi-carrier com-
munication systems, image processing and medical diagnosis
using soft computing, with various publications in journals and
conferences of international repute.
Sagheer a bbaS is currently working as an assistant professor
at the School of Computer Science, NCBA&E, Lahore, Paki-
stan. He completed his Ph.D. from the School of Computer
Science, NCBA&E, Lahore, Pakistan. He completed his M.Phil.
in Computer Science from the School of Computer Science,
NCBA&E, Lahore, Pakistan. He has been teaching graduate
and undergraduate students in computer science and engineer-
ing for the past 10 years. He has published about 80 research
articles in international journals as well as reputed international
conferences. His research interests primarily include cloud com-
puting, IoT, intelligent agents, image processing and cognitive
machines with various publications in international journals and
conferences.
abdur r ehMan is currently working as lecturer at School of
Computer Science, NCBA&E, Lahore, Pakistan, and as a game
developer at the GameObject Lahore, Pakistan. He is working
toward a Ph.D. from the School of Computer Science, NCBA&E,
Lahore, Pakistan. He completed his M.Phil. in computer scienc-
es from the NCBA&E, Lahore, Pakistan. He completed his B.S.
in computer sciences from the Institute of Management Scienc-
es, Lahore, Pakistan. He has published and submitted several
research articles in international journals as well as well-respect-
ed international conferences. His research interests primarily
include cloud computing, IoT, medical diagnosis, intelligent
agents, cognitive machines, smart homes, blockchain, network
security, and smart city, with various publications in international
journals and conferences of international repute.
YouSaf Saeed received the Ph.D. degree in cognitive VANETs
from NCBA&E, Lahore, Pakistan, and the M.S. degree in
broadband and high-speed communication networks from the
University of Westminster, London, U.K., where he achieved
distinction for his individual research thesis on IPv6. He is cur-
TABLE 2. Training of the blockchain-based smart home empowered with deep
extreme learning machine system model during the prediction of malicious
activity or attacks.
Proposed DELM based system model (85% of sample data in training)
Total number of samples (N = 125,973) Output results (O0, O1)
Input
Expected output (T0, T1) O0 (normal) O1 (attack)
T0 = 67,343 normal 65,366 1,977
T1 = 58,630 attack 2,420 56,210
TABLE 3. Validation of the blockchain-based smart home empowered with deep
extreme learning machine system model during the prediction of malicious
activity or attacks.
Proposed DELM based system model (15% of sample data in validation)
Total number of samples (N = 22,543) Output results (O0, O1)
Input
Expected output (T0, T1) O0 (normal) O1 (attack)
T0 = 9,710 normal 9,237 473
T1 = 12,833 attack 898 11,935
NASSER_LAYOUT.indd 228NASSER_LAYOUT.indd 228 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.
IEEE Network • May/June 2021 229
rently an assistant professor with the Department of Informa-
tion Technology, University of Haripur, Pakistan. He acquired
seven research projects of the ICT Research and Development
Higher Education Commission of Pakistan. His achievements
include the publication of monographs, journal articles, and
conference papers. His patent is under review regarding emer-
gency vehicles-based traffic light control system. He received
the International Students Award at the College of North West
London, U.K.
aSiM Z eb received his B.Sc. and M.Sc. degrees in computer
science from the University of Peshawar, Pakistan (UOP) in
2002 and 2005, respectively. He then completed his Ph.D. in
computer science from University Technology Malaysia (2012-
2016) and also served as a research fellow at the Nagoya Insti-
tute of Technology, Japan (2014-2015). He has received the
MJIIT-Malaysia Scholarship (2013–2014) and the JASSO-Japan
Scholarship (2014–2015). He served as an assistant professor
at Qurtuba University of Science and I.T. from February 2016
to April 2019. Currently, he has been serving as an assistant
professor in the Department of Computer Science at Abbotta-
bad University of Science and Technology, Pakistan since May,
2019. His research interest includes self-organized networks,
network architectures and protocols, cognitive radio networks
and supervised machine learning
M. irfan u ddin is actively involved with academia and
research. He worked as a graduate research associate at the
University of Peshawar, University of Amsterdam and University
of Turin. He was a faculty member on the computer science
faculty at Al Yamamah University, Saudi Arabia. He is currently
working at the Institute of Computing, Kohat University of Sci-
ence and Technology, Kohat, Pakistan. He has participated in
different research journals and conferences. He has published
several research papers in well-respected international journals
and conference proceedings. He serves as a reviewer for sev-
eral journals. His research interests include machine learning,
data science, deep learning, convolutional neural networks,
reinforcement learning, computer vision, Big Data and parallel
computing.
nidal naSSer (SM-IEEE) completed his Ph.D. at the School of
Computing, Queen’s University, Kingston, Ontario, Canada, in
2004. He is currently a professor of software engineering at
the College of Engineering, Alfaisal University, Saudi Arabia.
He worked in the School of Computer Science at the Univer-
sity of Guelph, Guelph, Ontario, Canada. He was the Founder
and Director of the Wireless Networking and Mobile Comput-
ing Research Lab @ Guelph. He is currently the Founder and
Director of the Telecommunications Computing Research Lab
@ Alfaisal University. He has authored 180 journal publications,
refereed conference publications and book chapters in the area
of wireless communication networks and systems. He is current-
ly serving as an associate editor for IEEE Wireless Communica-
tions Magazine, Wiley’s International Journal on Communication
Systems, and IEEE Systems Journal. He has been a member of
the technical program and organizing committees of several
international IEEE conferences and workshops. He has received
several outstanding research awards as well as a number of best
paper awards.
aSMaa a li completed her Ph.D. degree from the School of
Computing at Queen’s University-Canada. She received two
M.Sc. degrees with honors in the Electrical Engineering Program
from Kuwait University, State of Kuwait, and in the Engineer-
ing Systems and Computing Program from the University of
Guelph-Canada. She received the Best Paper Award at the IEEE
International Conference on Smart Applications, Communica-
tions and Networking (SmartNets 2019). Her current research
interests are computer vision and pattern recognition, robotics,
control systems, artificial intelligence, machine/deep learning,
data science and wireless sensor networks.
NASSER_LAYOUT.indd 229NASSER_LAYOUT.indd 229 6/3/21 1:39 PM6/3/21 1:39 PM
Authorized licensed use limited to: Gachon University. Downloaded on July 04,2021 at 13:45:48 UTC from IEEE Xplore. Restrictions apply.