Weighted Coefficient Firefly Optimization Algorithm and Support Vector Machine for Trust Model and Link Reliability

International Journal of Computer Networks & Communications (IJCNC) Vol.14, No.5, September 2022
DOI: 10.5121/ijcnc.2022.14508 117
WEIGHTED COEFFICIENT FIREFLY OPTIMIZATION
ALGORITHM AND SUPPORT VECTOR MACHINE FOR
TRUST MODEL AND LINK RELIABILITY
Shalini Sharma and Syed Zeeshan Hussain
Department of Computer Science, Jamia Millia Islamia University New Delhi, India
ABSTRACT
Cloud computing is widely used by organizations and individuals due to its flexibility and reliability. The
trust model is important for cloud computing to detect malicious users and protect user privacy. The
existing research faces the issues of local optima trap and overfitting problems when a training user node
is idle for more time. This research proposed Weighted Coefficient Firefly Optimization Algorithm (WC-
FOA) with Support Vector Machine (SVM) for the trust model calculation and identifying paths with better
Quality of Services (QoS). The weighted coefficient is added to the FOA model to balance the exploration
and exploitation in the search of identifying optimal path based on reliability score. The WC-FOA method
measures the link reliability in the model and SVM detects the malicious users in the model. The WC-FOA
model selects the optimal path for transmission in terms of trust and efficient QoS parameters. The entropy
measure and link reliability are provided as input to the SVM model for the detection of attacks in the
network. The WCFOA-SVM model has 96% malicious user detection, whereas the Random Forest
Hierarchical Ant Colony Optimization (RF-HEACO) has 92 % accuracy.
KEYWORDS
Cloud computing, Entropy Measure, Support Vector Machine, Trust model, Weighted Coefficient Firefly
Optimization Algorithm.
1. INTRODUCTION
Trust Management is considered as a new technique to solve network security problems and
become a hot research topic in recent years. The trust management techniques have
characteristics of multi-dimensional evaluation, flexible application, and deployment, real-time
resistance to internal attacks and dynamic [1]. Cloud’sheterogeneitybehaviour is used by
malicious users to create trust problems and affect other devices' reliability and service. The trust
model is widely required in network security to protect cloud smart services. The Trust
management technique identifies untrusted behavior and separates untrusted objects. The existing
techniques still have many limitations like continuously changing its trust behaviors and
ineffectiveness against a large amount of data [2]. The huge quantity of connected users in the
cloud increases the chance of malicious surface on the Internet. The focus of this paper is on the
challenges and issues related to trust and security in the context of consumer cloud applications
[3]. Traditional security techniques can defend against external threats and lower efficiency in
identifying internal attacks due to the malicious inclusion of nodes into the network [4]. The
perimeter-based security architecture is applied to solve this problem that divides networks into
external networks and internal networks with an Intrusion Prevention System (IPS), Intrusion
Detection System (IDS), and firewall as the border [5].

118
A huge amount of users in the cloud environment results in higher traffic and flow of data to be
slow inside the cloud environment. The higher traffic creates a delay in the network and creates
multiple hops for a data to be transferred from one user to another. If in case a malicious user is
on the route of data transmission, there is a higher chance of rerouting the data in a misguided
path or the loss of data can happen. To avoid these scenario, it is essential for a network to have a
reliable path with neighbors for data transmission and effective identification malicious nodes in
that path. Malicious node identification in modern network infrastructure are considered severe
threats to network security. Traditional network security finds it difficult to distinguish between
the malicious node and large legitimate users accessing a targeted network or a resource [6]. Few
studies involve in collect and analyzingmalicious node characteristics for these attacks. This
research mainly focuses on malicious node detection and to mitigate the attack, which is difficult
for legitimate users’ traffic patterns and various network services [7, 8]. Networks' crucial issues
are reliable data exchange and energy consumption problems. Clouds are a vulnerable target for
many security attacks due to the limited resources of sensor nodes. Trust based technique of
energy awareness is a powerful tool for the classification of nodes' behavior and provides security
solutions in cloud [9, 10].The objectives and contribution of this research are described as
follows:
1. SVM is used to classify the malicious users based on the residual energy, bandwidth,
entropy, Packet Loss Ratio (PLR) and End to End Delay (EED). After classifying the
malicious users, the information about these malicious users are broadcasted to achieve
the reliable transmission.
2. The conventional FOA is transformed into WC-FOA by adding the weighted coefficient
that helps to enhance the exploration and exploitation capabilities during the secure route
discovery. Therefore, the combination SVM based malicious node discovery and WC-
FOA based secure routing are used to enhance the link reliability.
The paper is organized as follows: The trust model literature review is given in Section 2, and the
WC-FOA-SVM model explanation is given in Section 3. The results of the WC-FOA-SVM
model are given in Section 4, and the conclusion of the research paper is given in Section 5.
2. LITERATURE REVIEW
The trust model for the network is important to find the malicious nodes in the network and
increases the privacy of the network. Some recent techniques in the trust model and link
reliability technique were reviewed in this section.
Guesmi, et al. [11] applied a supervised detection technique for cloud environments using Fast
Entropy based on unfair ratings and cloud users’ feedback, named Feedback Fast Entropy-based
Detection Strategy (FFED). The user’s rating was monitored by the provided detection system
and fast entropy algorithm was used to detect unfair rating attacks to permit the scale effectively.
The unfair rating attacks are considered as events that cause changes in rating computation and
help to identify different attacks. These attacks are mitigated and feedback was used to detect
changes. The FFED method has lower adaptability and this model has lower efficiency in
detecting new attacks.
AbdelAzim, et al. [12] developed a hybrid model that was a combination of entropy and
KullbackLeibler (KL)-divergence to consider node trust in routing problems for detecting attacks.
The hybrid method is developed to detect denial of service attacks and routing information was
considered in the nodes for quick identification and exclusion from the nodes. The developed
method allows nodes that have been suspected of participating in an attack if they cease their
malicious behavior. The developed method can detect the second attack while another one is

119
ongoing. The developed method of detecting denial of service attacks has higher performance
than non-hybrid techniques. This binary class classification technique and lower efficiency in
attack classification.
Wu, et al. [13] applied Trustworthiness Assessment with Entropy (TAE) for topological
characteristics analysis in IoT networks. The TAE’s trustworthiness of qualitative and
quantitative were carried out using von Neumann entropy which used to increase the feasibility
and robustness of the model. The multi-layer complex network was used to map the model in
cloud-fog-edge computing.
Kou, et al. [14] proposed a Simhash-based link prediction technique to improve privacy-
preserving. The target user indices were used to determine “probably similar” friends based on
Simhash to develop less sensitive users. The developed method effectively protects users’
information and trust-distrust values are calculated for each user related to the target user. The
Social Balance technique was applied to develop a link of possibility between candidate and
target user using trust-distrust values. The real-world Epinions dataset was applied to test the
developed model’s performance. The developed model has higher performance in terms of
overcoming sparsity problems than the existing technique. The Simhash technique only
considered user rating data and is highly based on similarity measures.
Sankaran, et al. [15] applied Energy Based Random Repeat Trust Computation (EB-RRTC) for
trust nodes encounter with the destination node. The Reliable Fuzzy and Heuristic Concurrent
Ant Colony Optimization (RF-HEACO) were applied to improve Quality of Service (QoS). The
ACO model was applied to identify candidate deposit routes among a pair of source and
destination. The Reliable fitness function, heuristic factors, and QoS metrics are considered for
identifying the candidate routes. Each path is measured using the reliability technique and Fuzzy
logic with packet loss rate, residual energy and metric link stability for high-reliability path
selection. The RF-HEACO QoS routing protocol reduces energy consumption and increases the
packet delivery ratio. The ACO model has limitations of local optima trap and overfitting
problems in optimization.
Alshammariet al. [16] developed a Trust Model System (TMS) to ensure the security over the
cloud storage system. In TMS, the value of interaction trust was computed for malicious attackers
to improve the security. The developed TMS was provided the security against the reputation
attacks. Further, the developed TMS was provided the enough security, even when the cloud
services were processed in dynamic topology. This TMS was failed to consider the reliability of
link while performing the data transmission.
Hassan et al. [17] presented the enhanced QoS-based model to estimate the authorization of the
cloud provider. The Accumulative/Computed Trust Value (ATV)was computed for each cloud
provider for evaluating their own authentication level. Here, the user feedback ratings according
to the covariance mathematical approach was used to compute the cloud resource’s
trustworthiness. Next, the resource power was computed for calculating the trustworthiness of a
cloud resource. The elimination of fake users was used to enhance the transaction success rate in
cloud environment. If the number of connections were increased, it caused failure in certainjobs
in cloud.
Qureshi et al. [18] developed the Software-Defined Network (SDN) based Anomaly Detection
System (ADS) for edge computing-based systems in IoT networks. The behavior of device for
SDN and edge computing networks were discovered by ADS. An edge device’s trust for ensuring
the data forwarding was accomplished by developing a trusted authority for edge computing

120
approach. The data delivery of SDN-ADS was less, when the environment has more number of
malicious switches.
The related work along with its advantages and limitations is provided in the Table 1.
Table 1. Related work
Author Approach Advantages Limitations
Guesmiet al.
[11]
Supervised detection
technique and
Feedback Fast
Entropy-based
Detection Strategy
(FFED)
An effective detection of
unfair rating attacks were
done by monitoring the
user’s rating along with
FFED.
The developed FEED
has less adaptability
while detecting the
unknown attacks.
AbdelAzimet
al. [12]
Hybrid entropy and
KL-divergence
model
The Hybrid entropy and
KL-divergence method was
detected the second attack
while the first one is
ongoing in network.
This Hybrid entropy
and KL-divergence
was offered only a
binary classification.
Wuet al. [13] TAE The TAE method
trustworthiness was
performed using von
Neumann entropy that used
to enhance the feasibility.
The TAE was
mapped using multi-
layer complex
network at cloud-
fog-edge computing.
Kouet al. [14] Simhash-based link
prediction technique
The Simhash based
discovery has enhanced
performance in terms of
overcoming sparsity issues
than the existing method.
The Simhash
technique was only
considered rating
data of user and it
was highly depends
on the similarity
measures.
Sankaran, et
al. [15]
EB-RRTC and RF-
HEACO
The energy consumption
was reduced and packet
delivery ratio was increased
by using RF-HEACO QoS
routing protocol.
The ACO has
limitations of local
optima trap and
overfitting issues in
optimization.
Alshammari et
al. [16]
TMS The TMS was provided the
enough security, even when
the cloud services were
processed in dynamic
topology
This TMS was failed
to consider the
reliability of link
while performing the
data transmission.
Hassan et al.
[17]
Enhanced QoS-
based model
The elimination of fake
users using ATV was used
to enhance the transaction
success rate.
If the number of
connections were
increased, it caused
failure in certain jobs
in cloud.
Qureshi et al.
[18]
SDN-ADS Edge device’s trust for
providing the data
forwarding was
accomplished by developing
a trusted authority for edge
computing approach.
The data delivery of
SDN-ADS was less,
when the
environment has
huge number of
malicious switches.

121
The limitations found from the existing research are mentioned as follows: The cloud computing
environment possess less adaptability to the unknown attacks from inside malicious users,
restricted detection of behaviour between the malicious user and legitimate user is also a
challenging issue on cloud and providing security functionalities in addition requires inadequate
cost parameters and suffers from local optima trap, overfitting and less data delivery. Hence, an
effective secure routing under malicious users is developed using SVM and WC-FOA for
achieving the reliable data broadcasting in cloud environment.
3. PROPOSED METHOD
In this research, the SVM and WC-FOA are used for achieving secure reliable communication
over the cloud environment. Specifically, the SVM is used to find the malicious attackers exist in
cloud followed by the WC-FOA used to find the optimal path for transmission. The distinct cost
metrics used in the WCFOA-SVM are residual energy, bandwidth and entropy. The developed
WCFOA-SVM is used to overcome the issues of local optima trap and overfitting. The overview
of the WCFOA-SVM model is given in Figure 1.
Figure 1. Block diagram of WCFOA-SVM method
The important process of the WCFOA-SVM method is mentioned as follows:
 The cloud environment is initialized with enough resources i.e., energy and bandwidth
during system model initialization. Accordingly, the resources are allocated for each user
of cloud.
 For detecting the malicious attackers exist in the cloud, the SVM is trained with unique
cost metric. Here, the cost is generated based on residual energy, bandwidth, entropy,
PLR and EED.
 Next, the WC-FOA is used to discover the route based on residual energy, bandwidth and
entropy. After discovering the route, the data is transmitted from one user to another user
in the cloud environment.

122
The detailed explanation about SVM based malicious user discovery and WC-FOA based route
discovery are given as follows:
3.1. Malicious user discovery using SVM
The trustworthy neighboruser discovery is supported among end to end user by using the SVM.
After identifying the malicious user, the local routing table of each user is updated in each
simulation time. The cost generation and SVM classification is detailed below:
3.1.1. Cost generation
The cost metric used to discover the malicious attacks are residual energy  
RE , bandwidth
 
BW , entropy, PLR and EED. The cost metric used during the attack userdiscovery is expressed
in equation (1).
Cost RE BW Fastentropy PLR EED
     (1)
The user exist in the cloud environment requires enough energy to transmit the information.
Therefore, the user with huge amount of residual energy is preferred in cloud environment.
Equation (2) shows the computation of residual energy.The available bandwidth information of
the user expressed in equation (3) is used to plan their own or distributed nodes. This used to take
the decision about the destination user and data speed. Feedback count is used in supervising the
evolution of existing practices used in this step. A threshold k of trust manager is set for the sum
of existing feedback counts related to Cloud Service Provider (CSP). During a fixed time period
t if this sum is higher than threshold k , the attack detection algorithm is executed by the system
based on fast entropy which is expressed in equation (4). The PLR expressed in equation (5) is
used to define the loss ratio where the malicious users creates the huge amount of loss during the
data transmission. Hence, this used to identify the malicious users in the cloud. Further, the EED
shows the amount of time taken to transmit the data to the desired user.
RE Availableenergy Consumed energy
  (2)
 
/
idle
BW t t Channelcapacity
  (3)
1
1
1 ( )
0 ( )
n
i
n
i
if counter i
Fast entropy
f counter i




 

 





(4)
Numberof lost packets
PLR
Total numberof transmitted packets
 (5)
Where, idle
t is idle time period; t is overall time period; For a fixed time period t , the number
of connections ( )
n attained by CSP is taken as feedbacks count i that is ( )
counter i .The
derived cost function is used as the input for SVM to detect the malicious attackers.
3.1.2. Support Vector Machine for attack detection
Vapnik (1995) proposed a Support Vector Machine for solving classification and regression
problems [19 – 20]. The SVM is a supervised learning technique for the classification training
process in various fields. SVM model can apply for binary and multi-class classification, also
suitable for linear and non-linear data classification tasks. SVM model creates a hyper-plane for

123
partitioning the data in high-dimensional space and selects the best hyper-plane based on its
capacity to partition the data. The non-linear classifier's margins are estimated using various
kernel functions and commonly applied kernel functions are linear, polynomial, radial basis, and
sigmoid. The researchers have successfully applied SVM for many applications due to its
efficiency in classification, SVM model has been widely used in image processing and pattern
recognition models.
The SVM model used in this study is based on Radial Basis Function (RBF) with the proposed
model. Two numeric vectors are used in Squared Euclidean distance of kernel function is applied
and for optimal partition of input data, input data is mapped to high-dimensional space. The
kernel RBF model is effective to partition a set of shared complex boundaries. This study uses the
multi-class classification problem for attack classification. A two-class problem is divided from a
multi-class problem. This study uses Radial Basis Function (RBF), as given in equation (6).
2
( , ) , 0
x y
K x y e 

 
  (6)
For training samples ( , ), 1,2,..
i i
x y i n
 , where maximum number of samples is given as
 
, 1, 1
i i n
i y and x R
  where the positive class is 1 and the negative class is -1, and the subspace
of labels corresponding to i
x is represented as n
R . The solution to the following problem is
provided when using the SVM model, which shown in equation (7).
min 1
2
, , 1
n
T
w b i
w w C



  (7)
Subject to ( ( ) ) 1
i i i
y wTw x b
 
  
Where, b is the bias term of the SVM. The training vector i
x transforms  to higher dimensional
space. A hyper-plane of SVM has a maximum margin to partition various classes of data.
The SVM model has the benefits of having minimal parameter requirement and the disadvantage
is that the requirements of a Gaussian function in the training set for each instance cause
performance degradation and increases training time on a large dataset in classification case. Soft
margin is used if the model fails to find a hyper-plane. The positive slack variables in soft margin
use , 1,2,..
i i N
  in the constraints, as in equations (8 - 10).
 
. 1 1
i i i
w x b for y

      (8)
( . ) 1 1
i i i
w x b for y

      (9)
0
  (10)
Where i
 must exceed unity if an error occurs. Then, an upper bound is i i

 for training error.
The Lagrange is given in equation (11).
 
2
1
1
( . ) 1
2
n
i i i i i i i i i
i
Lp w C y x w b
    

        
 (11)
Where Lagrange multipliers are denoted as i
 and this is used to provide i
 positive
value.Therefore, the SVM provides the information about the malicious attackers and this
information is broadcasted over the network. Accordingly, the ID of malicious attackers is

124
removed from the neighbouring user routing table. Further, the route from one user to another
user is generated by using the WC-FOA.
3.2. Link Reliability using Weighted Coefficient Firefly Optimization Algorithm (WC-
FOA)
The WC-FOA is developed for discovering the route between one user to another user according
to the fitness values of residual energy, bandwidth, and EED. Xin-She developed the
optimization technique of the Firefly algorithm at the University of Cambridge[21–23]. The
conventional FOA is changed into WC-FOA by considering the weighted coefficient during the
position update which used to improve the position update according to the best solution.Firefly
characteristics related to its activities of locomotion are mimicked in this firefly algorithm.
Firefly algorithm is considered to be an effective technique to find a solution to various
engineering problems due to its high exploration capacity, flashlight capability, and brightness.
The bionics principle is used by the firefly algorithm and the best optimal value is selected based
on the most effective firefly algorithm for non-linear and complex design. The light capability of
flash is used by each firefly based on the arbitrary solution to attract the adjacent firefly.
3.2.1. Purpose of flashing light
1. Commonly, every firefly is unisexual and attracts a partner for mating.
2. The attraction capability is based on its brightness and light capability is used by fireflies
to attract prey for survival.
3. Light flashing is used by fireflies to protect against their other enemies.
3.2.2. Light Intensity variation and Attraction Capability
The initialization considered in the WC-FOA is the possible routes from the transmitter user to
the destination user. These possible routes are given as input to the WC-FOA followed by the
iterative process of position update takes place to find the optimal route. In the firefly algorithm,
the attraction capability and variation of light intensity play a significant role. Light intensity is
used to determine the fitness value of the algorithm. The firefly algorithm can handle highly non-
linear and several multi-optimization problems. The firefly with low or high intensity attracts
neighbouring fireflies with low or high intensity. Consider XY
D is the distance between two
fireflies such as X andY . The light intensity reduces concerning distance from the source and
media absorb light. Equation (12) provides the intensity of the light as per the law of square
inverse.
2
( ) s
XY
I
I D
D
 (12)
The source intensity is denoted as S
I . The light intensity expression I
L varies related to distance
XY
D , as in equation (13).
0
XY
D
I
L I e 

 (13)
The fixed light of absorption coefficient is denoted as  . The initial light intensity is denoted as
0
I .

125
Strong attractive capacity is present in each firefly and this has strong firefly behavior attraction
over neighboring firefly groups. Two firefly distances namely X and Y are used to vary the
attractive capability. Fireflies' attractive capability is directly proportional to the light intensity of
neighboring fireflies. The attractive function expression is given in equation (14).
2
0
XY
D
e 
  
 (14)
The attractive capability at distance 0
XY
D  is denoted as 0
 in equation (15). The calculation of
characteristic length for a fixed light is given in equation (15).
1
1,M
M

      (15)
Two different fireflies X and Y Cartesian distances are represented as X
P and Y
P , respectively.
The two fireflies' Cartesian distance formula is given in equation (16).The Cartesian distance
among two points in Euclidean space is the length of a line segment among the two points.
2
, 1
( )
R
R
X Y X Y Y
X
R
D P P P
P


   
 (16)
Attraction movement from one fly to another fly is X and Y as in equation (17). The firefly
movement determination related to attraction capability is given in equation (17).
2
1
0 ( )
XY
D
R R
X X Y X
P P e P P J

 


    (17)
Where an attractive term is denoted as
2
0 ( )
XY
D
Y X
e P P

 
 and random variable is in term J in
range of [0, 1].
To improve Firefly algorithm performance, the weight coefficient ( )
w is applied and the change
firefly random follows the object into the best individual. The modified position update is
described using state i as shown in equation (18).
 
, 1 , , ,
( )
i iter i iter i iter i iter
i
x w t x D f gbest x

      (18)
Where, i
D is the distance from the firefly to the light source;the current iteration of the optimal
solution is denoted as gbest and adaptively changes ( )
w t related to equation (19).
2
2 max
max
( ) exp( / (2 ))
40
iter
w t w t
 
    
 
(19)
Where w maximum initial value is denoted as max
w which is equal to 0.01. If w value is higher in
an early stage, it benefits exploration and if it was smaller in a later stage, it benefits
exploitation.Therefore, the developed WC-FOA used to discover the optimal secure route which
used to enhance the robustness against the malicious user. The elimination of malicious user
helps to improve the PDR of the cloud environment.

126
4. RESULTS
The WCFOA-SVM technique is applied in cloud computing and evaluated various metrics such
as attack detection accuracy, energy consumption, Packet delivery ratio, and Throughput. The
simulation parameters of WCFOA-SVM are given in the following Table 2.
Table 2. Simulation parameters
Parameter Value
Number of users (Nodes) 50-500
Network area 1000 1000
m m

Propagation model Propagation/TwoRayGround
Mac type Mac/802_11
Channel type Channel/WirelessChannel
Phy type Phy/WirelessPhy
Packet size 20 bits
The WCFOA-SVM method is compared with the existing technique to evaluate the efficiency of
the model.
Metrics: The metrics such as Accuracy, Packet Delivery Ratio and Throughput were measured
from the proposed model and mathematically expressed in equations (20-22).
100
TP TN
Accuracy
TP TN FP FN

 
  
(20)
Numberof packet receive
Packet Delivery Ratio
Numberof packet send



(21)
p s
D P
Throughput
total durationof simulation

 (22)
Where p
D is the number of packets delivered, s
P is the size of a packet, TP is True Positive, TN
is True Negative, FP is False Positive, and FN is False Negative.
Parameter Settings: The number of iterations for the WCFOA method is set as 50 and the
population size is set as 50. The MATLAB tool is used to simulate and evaluate the WCFOA-
SVM model.

127
Table 3. Attack detection accuracy of WCFOA-SVM
Number of
Nodes FFED [11] TAE [13] Simhash [14]
RF-HEACO
[15]
WCFOA-
SVM
0 0 0 0 0 0
50 70 70 85 85 90
100 71 71 85 86 90
150 73 71 87 86 90
200 75 72 88 87 91
250 75 73 88 87 93
300 75 77 88 87 94
350 75 80 88 88 96
400 76 81 88 90 96
450 78 81 90 91 96
500 79 83 90 92 96
Figure 1. Detection Accuracy of WCFOA-SVM for various nodes
The WCFOA-SVM technique is tested for various nodes in terms of detection accuracy, as in
Table 3and Figure 2. The WCFOA technique applies the weighted coefficient technique to
balance the exploration and exploitation. The WCFOA technique helps to find reliable links for
transmission and increases the attack detection performance. The WCFOA model finds the
optimal path for transmission and the SVM model can handle high dimensional data. The
WCFOA method has higher efficiency in attack detection than existing methods. The RF-
HEACO [15] method has a local optima trap and overfitting problem in attack detection. The
Simhash [14] focuses on similar based measures for attack classification in the model. The
WCFOA-SVM model has 96 % accuracy in attack detection and RF-HEACO method has 92 %
accuracy for 500 nodes.

128
Table 4. Energy consumption of WCFOA-SVM
Number of
Nodes FFED [11] TAE [13] Simhash [14]
RF-HEACO
[15] WCFOA-SVM
0 0 0 0 0 0
50 60 51 40 36 21
100 60 56 40 37 21
150 60 57 41 37 22
200 61 57 42 39 23
250 61 57 43 39 23
300 65 58 43 41 24
350 66 59 43 42 25
400 67 61 44 43 25
450 69 65 53 44 27
500 69 65 55 45 27
The energy consumption of the WCFOA-SVM model is measured for various nodes and
compared with existing techniques, as given in Table 4and Figure 3. The weighted coefficient in
the Firefly algorithm helps to balance exploration and exploitation. This helps to reduce energy
consumption in exploration or exploitation. The WCFOA method also finds link reliability and
this helps to provide optimal transmission. The WCFOA-SVM method has less energy
consumption than existing methods. The WCFOA-SVM model has 27 J energy consumption and
the RF-HEACO method has 45 J energy consumption for 500 nodes.
Figure 2. WCFOA-SVM energy consumption for various nodes

129
Table 5. Packet Delivery ratio of WCFOA-SVM
Number of
Nodes FFED [11]
TAE
[13]
Simhash
[14]
RF-HEACO
[15]
WCFOA-
SVM
0 0 0 0 0 0
50 80 74 79 85 91
100 71 76 83 87 96
150 71 76 79 89 94
200 70 78 75 90 92
250 80 70 76 80 95
300 77 74 85 85 93
350 80 78 77 80 95
400 72 72 82 81 94
450 79 72 75 82 94
500 73 74 85 83 96
Figure 3. WCFOA-SVM Packet delivery ratio for various nodes
The WCFOA-SVM model packet delivery ratio is measured and compared with existing
techniques in trust model, as in Table 5and Figure 4. The WCFOA-SVM model uses a weighted
coefficient to balance exploration and exploitation in the search process. Balancing the
exploration and exploitation helps to effectively detect the attack and reduce the packet loss. The
WCFOA-SVM model has a higher packet delivery ratio than existing methods in cloud
computing. The existing methods have a limitation of local optima trap and overfitting problems
in the model. The WCFOA-SVM model has 96 % packet delivery ratio and RF-HEACO method
has 83 % for 500 nodes.

130
Table 6. Throughput of WCFOA-SVM
Number of
Nodes FFED [11] TAE [13]
Simhash
[14]
RF-HEACO
[15]
WCFOA-
SVM
0 0 0 0 0 0
50 1 1 5 7 14
100 2 1 7 10 15
150 3 1 7 11 17
200 3 4 7 11 19
250 4 6 7 11 20
300 4 7 9 12 22
350 5 8 9 13 27
400 6 8 11 15 29
450 8 10 14 15 35
500 8 11 15 17 35
Figure 4. WCFOA-SVM Throughput for various nodes
The WCFOA-SVM throughput is measured for various nodes and compared with existing
techniques, as in Table 6and Figure 5. The WCFOA-SVM model balances exploration and
exploitation using a weighted coefficient. The balancing of exploration and exploitation helps to
increase the transmission capacity of the model. The WCFOA-SVM model has higher throughput
than the existing technique. The RF-HEACO model has a limitation of local optima trap and
overfitting problem. The WCFOA-SVM model has 35 MBPS throughput and RF-HEACO model
has 17 MBPS throughput.
The SVM based malicious attack detection with unique cost metrics is used to overcome the issue
of overfitting. The incorporation of weighted coefficient in WC-FOA is used to enhance the
exploration and exploitation capabilities that used to overcome the issue of local optima trap. The
attacks classified by the SVM is avoided while discovering the route using WC-FOA that used to
enhance the packet delivery of the network.

131
5. CONCLUSION
The existing researches on the trust model for the cloud have the limitations of local optima trap
and overfitting problem. The WCFOA-SVM model is proposed to increase the performance of
the trust model in attack detection and efficiency. The WCFOA model is applied for link
reliability measures and path selection. The SVM model is applied for attack detection using
entropy measure and link reliability. The RF-HEACO model has the limitation of local optima
trap and overfitting problem. The Simhash technique mainly focuses on similarity measures for
attack detection. The WCFOA method has higher efficiency than existing methods in terms of
attack detection, energy consumption, packet delivery ratio, and throughput. The WCFOA-SVM
model has 27 J energy consumption and the RF-HEACO model has 45 J energy consumption.The
developed WCFOA-SVM is identified only the malicious users, so further it is required to be
developed for identifying the network traffic-related attacks. The future work of this method
involves applying multi-class classification for attack classification using the deep learning
technique.
CONFLICT OF INTEREST
The authors declare no conflict of interest.
REFERENCES
[1] Rahman, M.A., Asyhari, A.T., Leong, L.S., Satrya, G.B., Tao, M.H. &Zolkipli, M.F. (2020). Scalable
machine learning-based intrusion detection system for IoT-enabled smart cities. Sustainable Cities
and Society, Vol. 61, pp. 102324.
[2] Ali, M.H., Al Mohammed, B.A.D., Ismail, A. &Zolkipli, M.F. (2018). A new intrusion detection
system based on fast learning network and particle swarm optimization. IEEE Access, 6, pp.20255-
20261.
[3] Manzoor, I. and Kumar, N. (2017). A feature reduced intrusion detection system using ANN
classifier. Expert Systems with Applications, Vol. 88, pp. 249-257.
[4] Haider, W., Hu, J., Slay, J., Turnbull, B.P. and Xie, Y. (2017). Generating realistic intrusion detection
system dataset based on fuzzy qualitative modeling. Journal of Network and Computer Applications,
87, pp.185-192.
[5] Jeyanthi, D.V. and Indrani, B. (2022). An Efficient Intrusion Detection System with Custom Features
Using Fpa-Gradient Boost Machine Learning Algorithm.International Journal of Computer Networks
& Communications (IJCNC), Vol.14, No.1, pp. 99-115.
[6] Lv, L., Wang, W., Zhang, Z. and Liu, X. (2020). A novel intrusion detection system based on an
optimal hybrid kernel extreme learning machine. Knowledge-based systems, Vol. 195, pp. 105648.
[7] Almiani, M., AbuGhazleh, A., Al-Rahayfeh, A., Atiewi, S. and Razaque, A. (2020). Deep recurrent
neural network for IoT intrusion detection system. Simulation Modelling Practice and Theory, Vol.
101, pp. 102031.
[8] Alazzam, H., Sharieh, A., &Sabri, K.E. (2020). A feature selection algorithm for intrusion detection
system based on pigeon inspired optimizer. Expert systems with applications, Vol. 148, pp. 113249.
[9] Le,H. D., Luong, N. T. and Nguyen, T. V. (2022). AODVMO: A Security Routing Protocol Using
One-Time Password Authentication Mechanism Based On Mobile Agent.International Journal of
Computer Networks & Communications (IJCNC), Vol.14, No.3, pp. 17-35.
[10] Eskandari, M., Janjua, Z.H., Vecchio, M. and Antonelli, F.(2020). Passban IDS: an intelligent
anomaly-based intrusion detection system for IoT edge devices. IEEE Internet of Things Journal,
7(8), pp.6882-6897.
[11] Guesmi, H., Kalghoum, A., Ghazel, C. &Saidane, L.A.(2021). FFED: a novel strategy based on fast
entropy to detect attacks against trust computing in cloud. Cluster Computing, Vol. 24, No. 3,
pp.1945-1954.

132
[12] AbdelAzim, N.M., Fahmy, S.F., Sobh, M.A. &Eldin, A.M.B.(2021). A hybrid entropy-based DoS
attacks detection system for software defined networks (SDN): A proposed trust mechanism.
Egyptian Informatics Journal, Vol. 22, No. 1, pp.85-90.
[13] Wu X., Wang, J., Wang, P., Bian, Z., Huang, T., Guo, Y. & Fujita, H. (2021). Trustworthiness
assessment for industrial IoT as multilayer networks with von Neumann entropy. Applied Soft
Computing, Vol. 106, pp.107342.
[14] Kou, H., Liu, H., Duan, Y., Gong, W., Xu, Y., Xu, X. & Qi, L., (2021). Building trust/distrust
relationships on signed social service network through privacy-aware link prediction process. Applied
Soft Computing, 100, p.106942.
[15] Sakthidasan, K., Gao, X.Z., Devabalaji, K.R. &Roopa, Y.M., (2021). Energy based random repeat
trust computation approach and Reliable Fuzzy and Heuristic Ant Colony mechanism for improving
QoS in WSN. Energy Reports, 7, pp.7967-7976.
[16] Alshammari, S.T., Albeshri, A. and Alsubhi, K., 2021. Building a trust model system to avoid cloud
services reputation attacks. Egyptian Informatics Journal, 22(4), pp.493-503.
[17] Hassan, H., El-Desouky, A.I., Ibrahim, A., El-Kenawy, E.S.M. and Arnous, R., 2020. Enhanced QoS-
based model for trust assessment in cloud computing environment. IEEE Access, 8, pp.43752-43763.
[18] Qureshi, K.N., Jeon, G. and Piccialli, F., 2021. Anomaly detection and trust authority in artificial
intelligence and cloud computing. Computer Networks, 184, p.107647.
[19] Zhou, J., Huang, S., Wang, M. &Qiu, Y. (2021). Performance evaluation of hybrid GA–SVM and
GWO–SVM models to predict earthquake-induced liquefaction potential of soil: a multi-dataset
investigation. Engineering with Computers, pp.1-19.
[20] Dong, S., (2021). Multi class SVM algorithm with active learning for network traffic classification.
Expert Systems with Applications, Vol. 176, pp.114885.
[21] Hassan, B.A. (2021). CSCF: a chaotic sine cosine firefly algorithm for practical application problems.
Neural Computing and Applications, Vol. 33, No. 12, pp. 7011-7030.
[22] Altabeeb, A.M., Mohsen, A.M., Abualigah, L. &Ghallab, A. (2021). Solving capacitated vehicle
routing problem using cooperative firefly algorithm. Applied Soft Computing, Vol. 108, pp. 107403.
[23] Ewees, A.A., Abualigah, L., Yousri, D., Algamal, Z.Y., Al-qaness, M.A., Ibrahim, R.A. &AbdElaziz,
M., (2021). Improved Slime Mould Algorithm based on Firefly Algorithm for feature selection: A
case study on QSAR model. Engineering with Computers, pp. 1-15.
AUTHORS
Shalini Sharma pursuing Ph.D. in computer Science from Jamia Millia Islamia University,
New Delhi, India. She acquired Master of Technology in Information Technology from
Guru Gobind Singh University, Delhi in 2011.She worked as Assistant Professor in Sharda
University during 2011- 2015.She has served IT organization Infogain India, IRIS
Software & Accenture 2015 onwards.
Syed Zeeshan Hussain acquired Ph.D. from Jamia Millia Islamia, New Delhi, India, and
Master of Computer Applications (MCA) from IGNOU, New Delhi respectively. He is
currently working as Associate Professor in department of Computer Science Jamia Millia
Islamia, New Delhi, India. His research interests include Computer Networks, Network
Security, Web Technology and Applications, Object-Oriented Computing, Scripting
Languages.

Weighted Coefficient Firefly Optimization Algorithm and Support Vector Machine for Trust Model and Link Reliability

More Related Content

Similar to Weighted Coefficient Firefly Optimization Algorithm and Support Vector Machine for Trust Model and Link Reliability (20)

More from IJCNCJournal (20)

Recently uploaded (20)

Weighted Coefficient Firefly Optimization Algorithm and Support Vector Machine for Trust Model and Link Reliability