Content uploaded by Mohamed S. Abdalzaher
Author content
All content in this area was uploaded by Mohamed S. Abdalzaher on Apr 28, 2022
Content may be subject to copyright.
A Game-Theoretic Approach for Enhancing
Security and Data Trustworthiness in IoT
Applications
Mohamed S. Abdalzaher ,Member, IEEE, and Osamu Muta , Member, IEEE
Abstract—Wireless sensor networks (WSNs)-based Internet of
Things (IoT) are among the fast booming technologies that dras-
tically contribute to different systems’ management and resilience
data accessibility. Designing a robust IoT network imposes some
challenges, such as data trustworthiness (DT) and power man-
agement. This article presents a repeated game model to enhance
clustered WSNs-based IoT security and DT against the selective
forwarding (SF) attack. Besides, the model is capable of detect-
ing the hardware (HW) failure of the cluster members (CMs),
preserving the network stability, and conserving the power con-
sumption due to packet retransmission. The model relies on the
TDMA protocol to facilitate the detection process and to avoid
collision between the delivered packets at the cluster head (CH).
The proposed model aims to keep packets transmitting, isotropic
or nonisotropic transmission, from the CMs to the CH for maxi-
mizing the DT and aims to distinguish between the malicious CM
and the one suffering from the HW failure. Accordingly, it can
manage the consequently lost power due to the malicious attack
effect or HW malfunction. The simulation results indicate the
proposed mechanism improved performance with TDMA over
six different environments against the SF attack that achieves
the Pareto-optimal DT as compared to a noncooperative defense
mechanism.
Index Terms—Game theory, Internet of Things (IoT), power
conservation, threats mitigation, wireless sensor networks
(WSNs).
I. INTRODUCTION
ACCORDING to the dramatic growth of the Internet-
of-Things (IoT) technologies, IoT systems are utilized
in a wide range of applications. In particular, wireless sen-
sor networks (WSNs), which can play a significant role in
serving IoT-based applications, such as smart cities, vehic-
ular networks, environmental and earth monitoring, electri-
cal power lines’ management, renewable energy adaptation,
etc. [1]–[3]. However, WSNs suffer from some weaknesses,
such as limited power, low processing capabilities, and espe-
cially the security and data trustworthiness (DT) aspects.
Manuscript received March 27, 2020; accepted May 19, 2020. Date of
publication May 22, 2020; date of current version November 12, 2020.
(Corresponding author: Mohamed S. Abdalzaher.)
Mohamed S. Abdalzaher is with the Center for Japan-Egypt Cooperation
in Science and Technology, Kyushu University, Fukuoka 819-0395,
Japan, and also with the Seismology Department, National Research
Institute of Astronomy and Geophysics, Cairo 11421, Egypt (e-mail:
msabdalzaher@nriag.sci.eg).
Osamu Muta is with the Center for Japan-Egypt Cooperation in Science
and Technology, Kyushu University, Fukuoka 819-0395, Japan (e-mail:
muta@ieee.org).
Digital Object Identifier 10.1109/JIOT.2020.2996671
Therefore, WSNs-based IoT security and DT are enormous
problems that desire an intelligent and adaptive solution to
optimally confront the day-to-day intellectual threats, such
as selective forwarding (SF), injection, and jamming attacks
[4], [5].
In the literature context, various trust mechanisms have been
proposed to resolve the WSNs security issues. Butun et al. [6]
discussed different intrusion detection systems (IDSs), e.g.,
game theory-based IDS, watchdog-based IDS, cluster-based
IDS, etc., to mitigate the WSNs security problems. More par-
ticularly, in [7], game theory has been deployed to establish
a valid IDS to realize the malware detection infrastructure. In
fact, game theory is a dedicated optimization branch that is uti-
lized to handle the interactions of a set of intelligent rational
players. More particularly, it aims to enhance their individ-
ual payoffs by an intellectual and adaptive manner [8]. More
concretely, game theory has emerged due to the distinguish-
ing features in managing the rational players’ interactions and
mitigating several security threats in WSNs, such as packet
(pkt) dropping, false data injection, and data delivery cor-
ruption [9]–[15]. In [14], an initial study of a game-theoretic
approach targeting the security and DT problems is presented.
However, all proposals presented in [9]–[15] did not consider
the crucial situation of hardware (HW) failure existence in the
presence of the attack impact, which can make the system crit-
ically unstable. Moreover, the wrong action can be intuitively
taken for a node caused by an HW failure.
The security issues and HW failure (fault) are among the
critical causes of packet dropping in WSNs, which have a
severe impact on the DT, network stability, and power con-
sumption. From the security front, dropping packets is a
malicious incentive for saving the transmission power in clus-
tered WSNs-based IoT. Accordingly, the decision taken can
be harmful, which causes a steep degradation for the IoT
networks’ DT. Moreover, the HW failure has several defects.
It can cause packet dropping, packet repeating, or over packet
transmission. The fault types can be categorized as offset fault,
gain fault, stuck-at fault, out of bounds, spike fault, noise
fault, data loss fault, and redundant data transmission [16].
In [17], the transmission round trip delay of the packets to
detect the nodes that suffer from the HW failure has been uti-
lized. Noshad et al. [16] extensively studied the HW failure
classifier methods, such as support vector machine (SVM),
machine learning (ML), and random forest. However, most
of the presented works in the literature context have only
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
considered the HW faults regardless of the security issue effect
and how to distinguish between them. In [18], a Stackelberg
game has been used to mitigate the corrupted delivered reports
in cognitive radio (CR) networks due to the spectrum sensing
data falsification (SSDF) attack in the presence of an HW mal-
function. However, to the best of our knowledge, no further
work has been presented in the literature focusing on classify-
ing between the attack effect and HW failure for packet drop.
Consequently, an intellectual and adaptive solution is desired
to resolve these crucial issues along with the WSNs limited
power source.
In this article, we propose a game-theoretic approach using
a repeated game to enhance WSNs-based IoT security and
DT against the SF attack and pinpoint the nodes that suffer
from the HW failure. The main contributions of this article
are fourfold.
1) The proposed repeated game aims to detect and miti-
gate the malicious cluster members (CMs) in clustered
WSNs-based IoT due to the SF attack impact giving
the adequate incentive to the participant CMs to cooper-
ate.1Furthermore, the model can determine the CMs that
suffer from HW malfunction among the CMs infected
by the SF attack at the equilibrium point. Therefore,
it is simultaneously guaranteed that no wrong action is
taken even for a node caused by an HW failure, which
is not considered in the literature context [9]–[15]. In
addition, unlike [14], this article presents more detailed
discussions such as clarifying the Pareto optimality of
the game.
2) The model-based repeated game solves the prisoner’s
dilemma and attains both Nash equilibrium (NE) and
Pareto optimality in the same state at the optimal DT
by resolving the issue of dropping packets due to the
SF attack effect on both high- and low-priority packets
leading to optimally enhancing all types of traffic. In
other words, the model can prevent the designated mali-
cious CMs from dropping packets by supporting these
CMs the incentive to act benevolently at which their
battery life is conserved. Furthermore, the model can
preserve the lost power of packets over transmission or
retransmission as a result of the HW failure situation.
3) The TDMA protocol is used to preserve the synchro-
nization between the cluster head (CH) and CMs, which
reduces the detection mechanism complexity and avoids
the collision between the delivered packets at the CH.
Besides, the isotropic and nonisotropic packet transmis-
sions have been taken in the model consideration.
4) The proposed model considers realistic conditions for
the WSNs-based IoT system using Tmote Sky mote [19]
with the standard of IEEE 802.15.4 supporting precise
locations and scalability for IoT platforms [20], and sig-
nal irregularity relying on real experiments presented
in [21] to represent a real data set. In addition, the
model effectiveness is verified using simulation over
six different environments: 1) outdoor line of sight
1The work in this article is initially presented in part in IEEE Symposium
on Computers and Communications [14]. Details are explained in Section II.
(OL); 2) outdoor nonline of sight (ON); 3) underground
line of sight (UL); 4) underground nonline of sight
(UN); 5) indoor line of sight (IL); and 6) indoor non-
line of sight (IN). Interestingly, the proposed approach
has proved beneficial as illustrated by the enhanced
performance as compared to the corresponding studies
in [13]–[15].
The remainder of this article is portrayed as follows.
Section II presents the related work. In Section III, the system
model is discussed. The proposed game approach is then
presented in Section IV, while the Pareto optimality and game
formulations are addressed in Section V. Section VI shows
the obtained results. Finally, the conclusion of this article is
revealed in Section VII.
II. RELATED WORK
This section discusses the related works of the different
WSNs HW failure detection methods and the security front
to promote WSNs performance. The security and HW failure
in IoT are prominent problems that can yield to losing data
privacy, DT, and wasting power as well [3]–[5], [22], [23].
On the one hand, most of the works in the literature con-
text concerning the HW failure utilize the conventional
paradigms, e.g., SVM, ML, and random forest by [16] and [17]
and self-diagnosis and cooperative diagnosis by [24]–[27].
Game theory can also be used to model the interactions
between nodes at the offloading services [28]. In addition, the
dynamic Bayesian network can be used for fault detection and
repairing [29].
In [24], the faulty node can be detected using redundant
node measurements. Thus, the faulty one is isolated based on
its obtained reputation by a minimum of three neighbor vot-
ing. An approach that relies on the neighboring nodes was used
to adapt time correlation information between these nodes to
detect faulty node(s) [25]. An ineffective solution can cause
more delay, and cost that relies on using an extra HW (testbed)
to check the faulty node was proposed in [30]. In [26], three
statistical algorithms were studied for fault detection called
time-series analysis, descriptive statistics, and Bayesian statis-
tics. The time-series mechanism is utilized to detect packets’
similarities and to measure the amount of data deviation.
Descriptive statistics use the mean or median of neighboring
nodes to vote for determining the faulty node. Bayesian tech-
niques are employed to determine the likelihood of a faulty
sensor based on Bayes’ theorem. Liu et al. [27] mentioned
that the node could detect the failure by self-diagnosis, such
as the faults caused by battery depletion, which is measured by
the battery current or voltage. Accordingly, if a node dropped
or lost packets, an extra memory should be embedded in the
nodes to retrieve those packets and resend again—they said.
The authors also respited the lost packet to the environment
condition; that is why we study the proposed paradigm over
six different environments.
On the other hand, uncontrolled WSNs security can
deteriorate the DT, data privacy, and power consump-
tion. In this trajectory, game theory has introduced sev-
eral approaches [9]–[15], [31], [32]. Abdalzaher et al. [9]
extensively studied various WSN attacks and suitable game
defense models. In [32], the Markovian chain was utilized
to model the game transitions for preserving data privacy.
In [10] and [11], Stackelberg games have been developed
to mitigate the external attacks manipulations using energy
defense to avoid the delivered data disruption in a clustered
WSN. The Stackelberg game has also been extended to con-
front the false injected data from intelligent attacks in WSNs
to enhance the DT [12]. The Stackelberg game was also uti-
lized to confront the false injected noise power in WSNs-based
CR due to the SSDF attack, where the HW failure problem
was considered [18]. The work in [18] aimed to mitigate
the disrupted observed signal-to-noise ratio due to the SSDF
attack to make the fusion center in CR capable of achiev-
ing accurate decision about the spectrum status. Interestingly,
in [13], a nonzero-sum game was proposed to mitigate the
DoS attack and the ON–OFF attack impact and to detect the
HW failure in WSNs. The SF attack was handled by game
theory in [11]. In [14], a repeated game model has been
utilized to enhance WSNs’ DT against the SF attack. More
particularly, the most effective parameter affecting the util-
ity in that game was the cooperation as a function of the
distance between the communicating nodes, which was criti-
cized by [21]. Furthermore, in [14], the model has only treated
high priority packets. In this regard, the low priority pack-
ets will suffer from a steep degradation due to the untreated
attack impact. Therefore, the overall DT will be degraded,
especially, when the majority of traffic contains low priority
packets, which is involved in most of the systems. Finally, the
realistic conditions such as fading problem, the environment
types that the nodes are deployed in especially the harsh envi-
ronment, the signal irregularity, the packet synchronization,
and the packet transmission type have not been considered
in [14]. Duan et al. [15] proposed an energy-aware trust deriva-
tion scheme utilizing a cooperative game-theoretic approach
to enhance WSNs security and manage overhead and latency
focusing on the trust evaluation process. However, the ade-
quate incentive to the malicious CMs due to the SF attack to
react benevolently and then, detects the CMs suffering from
the HW failure were not achieved. To solve this issue, a more
intelligent game model is needed to be designed along with
a robust detection scheme against the SF attack at the HW
failure existence.
Unlike the previous related works, this article presents
an efficient repeated game-theoretic approach along with the
TDMA protocol to classify between the cause of packet drop-
ping in WSNs-based IoT whether a reason of the malicious
effect due to the SF attack or a result of the HW failure to
manage the consequent power waste and achieve the optimal
DT. The repeated game can be defined as the iterated game,
which consists of some repetitive phases [9]. Each phase has
two players where their actions are considered in the con-
secutive actions. The repeated game guarantees collaborative
interaction between the participant players at the price of
elapsed time until reaching the equilibrium point, which meets
the flexibility of IoT applications’ time constraint. The pun-
ishment in the repeated game can be denoted as a payoff
reduction for the noncooperating player due to its reputation
Fig. 1. Disaster management using the WSNs-based IoT system.
based on the previous behavior [9], which is a necessary ele-
ment of solving the security and DT problems in WSNs-based
IoT.
III. SYSTEM MODEL
Fig. 1 illustrates the system model using IoT-based WSN.
In this model, WSN consists of a set of CMs N, where
the number of CMs in the cluster is given by the cardinal-
ity of the set N, which is represented by |N|. Table I lists
the utilized notations and variables. For managing synchro-
nization and packet transmission between the CMs and CH
in the clustered WSN, the TDMA protocol is employed, as
shown in Fig. 2. In fact, the proposed model is applicable
to work with different MAC protocols such as OFDM-based
protocols. But we utilize TDMA due to its features sup-
porting the power conservation. As extensively studied in
the literature, TDMA is effective for prolonging the network
lifetime [33]. Moreover, dynamic TDMA, which is a MAC
protocol based on the scheduled time (DMAC), is efficient for
WSNs energy conservation, prolonging the network lifetime,
and avoiding overhearing [34]. Consequently, it is an adequate
intensive to employ the TDMA protocol with the proposed
approach.
This article considers a game-theoretic model where each
game player has two-action status. First, every ith CM has
to execute either benevolent action by not dropping packets,
no drop (ND); or malicious action by dropping packets, drop
(D). The Daction is an incentive of the CM to save battery
power. Second, the CH performs Beacon (B) or no Beacon
(NB) action status. The Baction means that permission is given
to the benevolent CM, which does not drop packets, to send
their observed data. In addition, Bis used to permit the ith
benevolent CM to go to the sleep mode, to take a rest of
packet transmission to save its battery lifetime, or to get power
recovery when recharging power is available. Conversely, the
NB action is fundamentally used when the ith CM action is
D. Moreover, a controversial CM performance occurs when it
has a random HW failure at the same time of SF attack exist;
specifically, the ones suffering from dropping packets and over
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
TAB LE I
NOTATIONS OF PARAMETERS AND VARIABLES
packets’ transmissions, which are precisely considered through
the proposed model. In other words, the model can effectively
classify between the infected CMs by the SF attack and the
ones suffering from the HW failure. The TDMA is also utilized
to gradually send Bto the benevolent CM when its turn comes
to transmit its data.
Fig. 3 shows the Beacon (B) action/message distribution
after checking the CM behavior. At the beginning (Round # 0),
all CMs are supposed to be benevolent. Accordingly, the CH
action for all CMs is B. After calculating (Ui)s for all CMs
(at Round # z), the CH can determine which CM has the
right to be supported by the Beacon or not. Therefore, if the
ith CM is benevolent and it is its turn using TDMA, it will
receive the Beacon. On the contrary, if this ith CM is malicious
(drops packets), the Beacon will not be sent to this ith CM.
Accordingly, this malicious CM will keep retransmission the
Fig. 2. Packet transmission sequence using the TDMA protocol.
packets and will not receive acknowledgments from the CH.
Therefore, this approach can deteriorate the battery power of
the CM that keeps on malicious behavior, and hence, this CM
is prone to going to die. To this end, at (Round # Nrd), Nrd
is the given period representing the total number of rounds,
all CMs are benevolent, and no one will be changed after that
unless some of them have the HW failure. Indeed, the CH
communicates with the CMs throughout a star topology, in
which the communication link between each CM and the CH
is dedicated to facilitating the check procedure of the received
packets from every ith CM. In other words, the requirement
to enable our proposed algorithm is that each CM is directly
connected to the CH.2Therefore, it will be a conventional
topology to determine the benevolent CM that does not drop.
To this end, there are four scenarios in order due to the CH
action (Ai
CH) and the ith CM action (Ai). Three scenarios out
of these four represent the one-shot games at which both
the CH and CM perform the same action regardless of the
other player’s action. The last scenario introduces the rational
interaction between the CH and the ith CM as a reward and
punishment defense mechanism according to the action done
by every CM. Based on this scenario, all the malicious ratio-
nal CMs will rebehave benevolently. Finally, the four actions’
status of the two players (i.e., CH and the ith CM) is given as
follows.
1) Ai
CH =NB and Ai=D.
2) Ai
CH =NB and Ai=ND.
3) Ai
CH =Band Ai=D.
4) Ai
CH =Band Ai=ND.
2Therefore, it is clear that the proposed approach is applicable with any
other topologies as long as communications between CM and CH are possible.
Fig. 3. CH Beacon distribution for the CMs with TDMA.
IV. PROPOSED REPEATED GAME APPROACH
The proposed game confirms a cooperative attitude between
the CH (the defender) and the participant CMs. Indeed, the
CMs are assumed to act as rational players. The model is
used to mitigate the potential malicious CMs that drop the
observed packets due to the effect of an intelligent SF attack.
The CH plays the game individually with each CM and simi-
larly aims to maximize the whole network security and DT. In
other words, the CH separately interacts with each CM repre-
senting an attack–defense game. The proposed model provides
an adequate incentive for the CMs to react benevolently and
not drop packets. Conversely, if a CM persists in behaving
maliciously, this CM will not be permitted to go to the sleep
mode leading to shortly exhausting its battery.
Consequently, rational CMs will prefer to react benevolently
because it is more profitable based on the proposed game.
This game leads to the Pareto optimality and NE at which
both players reach their optimal utility. The utility function of
every ith CM per every iteration is computed using three main
parameters, namely, received signal strength indicator (RSSIi),
reliability level (RLi), and punishment parameter (ξi) between
the CH and the ith CM, which is given by
Ui=αRSSIi+βRLi−ξi(1)
α+β=1(2)
where αand βare “application-dependent” weighting factors.
A. Transmission Cost-Based Received Signal Strength
Indicator
First, this parameter concentrates on the foremost effective
factors of the transmission cost that are described as follows:
RSSI =TC −PL −Pn(3)
where TC is the transmission cost, while PL is the path loss,
and Pnis the noise power obtained from the six exploited
environments representing the link quality indicator (LQI),
which the CM is deployed over. Pnand PL parameters’
TAB LE I I
PATH -LOSS PARAMETERS [35]
TABLE III
TMOTE SKY TRANSMISSION LEVELS AND CONSUMED ELECTRIC
CURRENT
measurements are indicated in Table II [35]. Then, the trans-
mission cost is expressed by
TC =TC0+TCA(4)
where TC0is the initial cost that the CM starts up with, while
TCAdenotes the amplification cost for packet transmission.
The initial cost can be given by
TC0=V×I0×T0(5)
where Vis the battery voltage exploited for transmission, I0
represents the current in Amperes when the radio is in the ON
state, and T0is the initial startup time
TCA=V×Ic×L
DR (6)
where Icis the current in Amperes with a transmission power
level c,Lrepresents the packet length, and DR is the trans-
mission data rate. Table III shows the eight transmission levels
along with the corresponding consumed electric current using
the Tmote Sky mote [19]. The adopted path loss between every
ith CM and the CH (i,CH)isgivenby
PL[dB]=PLf[dB]+10nlog 10 diCH
d0×ηθ+σ[dB](7)
where PLfis the free-space path loss at the reference distance
d0of the antenna far field, ndenotes the path-loss expo-
nent, diCH is the distance between the transmitting ith CM
and the CH, σis the standard deviation in dB of the shadow
fading (log-normal distribution) measurements as indicated in
Table II, ηθ∈]0,1] is the path-loss direction coefficient when
nonisotropic radiation is used, and θrepresents the radiation
coefficient angle which is depicted in Fig. 4.
If the radiation is isotropic, ηθ=1. On the other hand,
when the radiation is nonisotropic, ηθis expressed by
ηθ=1,if θ=0◦
R×DOI,if 0◦<θ<360◦(8)
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
Fig. 4. Isotropic and Nonisotropic packet transmission.
where Ris a random value ∈]0,1[ and the experienced val-
ues of degree of irregularity (DOI) are obtained from the
experiments presented in [21].
B. Reliability Level
The reliability level (RL) for the ith CM (RLi) represents
the percentage of transmitted packets from the total num-
ber of packets (TP) per message (i.e., window size) during
a given period. In other words, RL describes the number of
successful packet transmissions to the number of total packets
transmitted, which is given by the CH as
RLi=TP
j=1fpktj
i
TP (9)
fpktj
i=⎧
⎪
⎪
⎨
⎪
⎪
⎩
1,if jth pkt is forwarded from the ith
CM to the CH
0,if jth pkt at the ith CM is dropped
(not forwarded to the CH)
(10)
where pktj
icorresponds to the jth forwarded packet from the
ith CM to the CH, and RLi∈]0,1].
C. Punishment Parameter (ξ)
The punishment parameter for the ith CM ξiin 1 is utilized
to enhance the cooperation between the CH and the participant
CMs, which is calculated by
ξi=⎧
⎪
⎪
⎨
⎪
⎪
⎩
x1,if Ai
CH =NB and Ai=ND
x2,if Ai
CH =Band Ai=D
x3,if Ai
CH =NB and Ai=D
0,if Ai
CH =Band Ai=ND
(11)
where x1, x2, and x3aregivenby
x1=
TP
j=1
f(pktj
i)×L×eb (12)
x2=⎛
⎝TP −
TP
j=1
f(pktj
i)⎞
⎠×L×eb (13)
x3=x1+x2 (14)
where eb denotes the transmission energy per bit, and TP is
the total number of packets to be transmitted to the CH.
In fact, the term ξiin (1) is a loss parameter used to stim-
ulate both CMs and CH to establish a cooperative connection
leading to prolonging the network lifetime where three main
ξicases are used. First, the CH does not send the Beacon
message to the ith CM; however, the CM is benevolent by
keeping the packet’s transmission to the CH. Therefore, this
CM cannot take the rest of packet transmission or go to the
sleep mode for saving the battery lifetime at which ξi=x1
due to the noncooperative behavior of CH. Second, ξiequals
x2 when the CM behaves maliciously by dropping packets;
however, the CH cooperates with it by sending it the Beacon.
Third, the worst case at which both the CH and CM choose to
be noncooperative players for which ξiequals x3=x1+x2.
D. Data Trustworthiness
According to the sections mentioned above, the utility of
CH is given as the DT and calculated by the CH, which is
represented by the average utility function of all CMs during
the given period as
DT =Nrd
rd=1|N|
i=1Urd
i
Nrd
.(15)
It is worth mentioning that the relationship of the packet drop
rate and the DT is inversely proportional.
E. Proposed Model and Corresponding One-Shot Games
The proposed repeated game is utilized to resolve the equiv-
alent prisoner-dilemma [8] using the available four scenarios
based on the players’ actions. The game is developed to ful-
fill the Pareto-optimal point and the NE point in the same
state. The first three scenarios represent a one-shot game of
prisoner-dilemma between the CH and the ith CM in which
the DT does not reach to its optimal value. The last scenario
presents the proposed repeated game versus the other one-shot
games, which are discussed as follows.
1) In the first scenario, all CMs’ actions are to cooper-
ate with the CH by not dropping any packet (ND). In
other words, they behave benevolently with the CH. On
the other hand, the CH action is NB to be sent to the
CMs even they are benevolent. Accordingly, the utility
function is affected by (12) of the punishment parameter.
2) In the second scenario, the CH chooses to exert Baction
for all CMs. Conversely, the CM action is to drop (D)
packets aiming at saving the battery lifetime. Therefore,
the punishment parameter here is represented by (13).
3) The third scenario is the worst one at which the mini-
mal utility is obtained due to the CH and CM actions,
NB and D, respectively. It means that the CH does not
provide the CM by the Beacon, and this CM action is
to drop packets. Consequently, the punishment param-
eter (ξ) is calculated based on (14). Therefore, these
malicious CMs will not be permitted to go to the sleep
mode, and hence, they are not capable of preserving their
battery lifetime.
Algorithm 1: Proposed Repeated Game Algorithm for the
SF Attack
1Input Ai
CH ={B,NB},Ai={D,ND};
2while d≤c|N|do
3Compute RSSIi,RLi,ξi, and Uiby Eqs. (1), (3), (9),
and (11) ∀i∈|N|;
4Determine the malicious CMs or the ones have HW
failure;
5if CMidrops pkt →Aiis D then
6This i-th CM →malicious/HWL;
7else
8This i-th CM is designated as benevolent and No
HW failure;
9end
10 if Aiof this i-th CM is ND and rd =i+i|N|then
11 This i-th CM will receive Bmessage (Ai
CH →B);
12 else
13 This i-th CM will be listed in HWL;
14 end
15 if DT is calculated using Eq. (15) and (Ai
CH ,Ai)→
(A∗
CH,A∗
i)∀ithen
16 DT is optimal and NE exists;
17 else
18 Continue;
19 end
20 end
21 Output Optimal DT is attained and (A∗
CH ,A∗
i), ∀i→NE
exists;
22 if Ai=Dthen
23 This i-th CM suffers from a HW failure;
24 end
4) The fourth scenario represents the proposed repeated
game model for handling the behavior of the ith mali-
cious CM. At the equilibrium point, the CH and all
the ith CM choose to cooperate by exerting Band ND
actions, respectively. More particularly, both RL and
utility functions are maximized. Generally speaking, the
process of permitting or preventing a CM from the
Beacon message until reaching the optimal DT is based
on Algorithm 1. In this algorithm, if a CM drops pack-
ets, it will be designated as malicious or suffers from the
HW failure. In other words, the dropping packets CM
will be designated as malicious or labeled in the HW
failure list (HWL). Then, this ith CM will go through
a punishment cycle till reacting benevolently. At satis-
fying two constraints, the round number rd =i+i|N|
comes and this ith CM reacts benevolently again, this
ith CM will be given the Beacon. For instance, if |N|
equals 10 and CM 3 is malicious, the Beacon will be
supplied to this CM only when it rebehaves benevo-
lently at rd =33. This scheme is recursively executed
Nrd =c|N|times till achieving the optimal DT, where
c∈R+>|N|. Consequently, when the CM action is
ND, its gain is much higher than when its action is Dfor
its utility function. Similarly, when the CH action is B
to the CMs, their utility functions are enhanced, which
yields to the optimal DT. Therefore, this process is called
the punish-and-forgive strategy. Finally, both the CH and
the designated rational malicious CM based on this game
will prefer to cooperate due to the gain they can earn.
In other words, the cooperative interaction between the
CH and CMs leads the Pareto optimality and NE in the
same state by reaching the optimum DT. After that, any
CM drops packets or does over packet transmission is
a result of the HW failure, which is listed in HWL and
quietly detected by the CH.
V. PARETO OPTIMALITY AND GAME FORMULATIONS
The proposed repeated game is between the CH (defender)
and every ith CM, which is denoted by
G={CH,i},A,Ui.(16)
The CH set of actions ACH represents the CH defense
strategy DSCH(ξi,DT )against the designated malicious CM.
Conversely, Aidenotes the ith CM malicious strategy
MSCM(f(pktj
i), DT)at which the jth packet is dropped by the
ith CM.
The best strategy is achieved using the proposed game rep-
resented by the fourth scenario in which the optimal action of
the ith CM A∗
i∈Aiis given by
A∗
i=arg max
ξi=0
(Ui).(17)
More concretely, the optimal action of the ith CM and CH
is given by
A∗
i=ND →fpktj
i=1∀j∈TP (18)
A∗
CH =arg max
fpktj
i=1∀j∈TP ∀i∈N,j=i
(DT). (19)
Taking into consideration the rationality principle of the par-
ticipant players, the CH will choose to cooperate and use B
action. Consequently, A∗
CH can be simplified as
A∗
CH =B→ξi=0∀i.(20)
Consequently, the NE is achieved, which is given here by
UiA∗
CH,A∗
i≥UiAi
CH,A∗
i∀Ai∈Ai∀i∈N.(21)
Then, we prove that both the Pareto optimal and NE point
are achieved in the same state as follows.
Let the Pareto optimality is attained at optimal DT (DT).
It is attained when the actions of CH and CMs are kept on
B and ND, respectively. Therefore, the optimal DT (DT) is
achieved as if any other action is done by each of them; the
utility will be decreased, affecting the DT. Consequently, any
other value of DT is less than this optimal value as
DT ≥DT.(22)
DT is attained using the punish-and-forgive strategy that is
presented in the fourth scenario, which is derived by
DT =min
ACH
max
Ai,∀i∈N
DT (23)
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
min
ACH=Bξi=0∀i∈N(24)
∴Ui=αRSSIi+βRLi∀i∈N(25)
max
Ai,∀i∈N
fpktj
i∀j∈TP ∀i∈N,j= i.(26)
Accordingly, A∗
i=ND →f(pktj
i)=1∀i∈N,j= i,and
A∗
CH =B→ξi=0∀i∈N
∴RLi=1∀i∈N(27)
∴DT >DT.(28)
Consequently, the NE and Pareto optimality are achieved in
the same state when DT exists at (A∗
CH,A∗
i)∀i∈N. More con-
cretely, there not be a better utility that will lead to a different
optimal solution.
VI. SIMULATION RESULTS AND DISCUSSION
This section shows the simulation results. The simu-
lation parameters are depicted in Table IV. We use the
Tmote Sky mote as a realistic node model among the most
usable nodes with the standard of IEEE 802.15.4 support-
ing precise locations and scalability for IoT platforms [20]
that is equipped by the Texas instruments MSP430 micro-
controller with a Chipcon CC2420 radio, and utilize the
parameters from its datasheet in [19]. We also assumed that
20% of CMs suffer from a random HW failure, and all
CMs use power level h=31 as an empirically selected
optimum value. The utilized DOI measurement values are
{0.0055,0.0035,0.004,0.0045,0.006,0.0085}[21]. All CMs
are located at the maximum coverage range between the CH
and every Tmote Sky CM (125 m) to show the effective-
ness of the proposed model. Besides, the DT of the received
signal is affected by Pn, path loss, and shadowing fading (log-
normal distribution with standard deviation σ). We assume six
environment scenarios (i.e., OL, ON, UL, UN, IL, and IN),
where their path-loss parameters are given in Table II [35].
The Rayleigh fading is not taken into consideration. For the
performance evaluation, we have simulated and compared the
proposed repeated game model with the corresponding ones
in [13]–[15].
Fig. 5 illustrates the improved performance using the
proposed game as compared to the works in [13]–[15]. It is
clear that the normalized DT of the proposed game reaches
the equilibrium state much faster than the corresponding game
models in [13]–[15]. Moreover, the obtained normalized DT
of the three one-shot games involved in this article gets worse
as compared to the proposed repeated game and correspond-
ing results in [13]–[15]. It means that the proposed repeated
game is more effective than the works in [13]–[15] for stimu-
lating the CH and CM to cooperate using Band ND actions,
respectively.
Afterward, the enhanced performance and model efficiency
using the number of successfully transmitted packets have
been depicted in Fig. 6 using isotropic transmission over the
ON environment representing the worst constraints as com-
pared to the works in [13]–[15]. The number of transmitted
packets counted at the equilibrium state based on the proposed
repeated game is close to the case of no attack and quite higher
TAB LE I V
SIMULATION PARAMETERS
Fig. 5. Normalized DT comparison between the proposed model and
the corresponding in the literature using nonisotropic transmission over UL
environment, the number of malicious CMs equals 2.
than the other three-game models in [13]–[15]. In this regard,
based on the proposed model along with the TDMA protocol,
the CH can receive a much higher number of trusted packets
from the CMs, leading to the optimum DT in the six environ-
ments and DOIs. In other words, the model effectiveness has
been achieved against the SF attack regardless of the environ-
ment and DOI types and whether the packet transmission is
isotropic or nonisotropic.
Moreover, the comparison between the lost power using
the proposed model, without a detection mechanism, and the
works in [13]–[15] against the SF attack is shown in Fig. 7
over ON environment representing the worst constraints. The
proposed model can save about 12.2%, 17.2%, and 17% of
the lost power as compared to the works in [13]–[15], respec-
tively. If the proposed model is not used, the network will
be prone to lose a valuable amount of power due to the SF
attack and the undiscovered CMs suffering from the HW fail-
ure, leading to battery power deterioration, which increases
with the increase of the number of CMs as shown in Fig. 7.
More particularly, the intensive power degradation is a result
of giving negative acknowledgments by the CH to the mali-
cious CMs. Hence, they recursively transmit those packets that
do not receive acknowledgments. Moreover, these CMs will
not receive the Beacon to take a rest from the transmission or
Fig. 6. Number of transmitted packets without and with attack presence
using isotropic transmission over ON environment.
Fig. 7. Lost power for the selfish CMs due to the SF attack impact using
isotropic transmission over ON environment.
to go to the sleep mode. Therefore, the fraudulent behavior of
the CMs will aggravatingly contribute to exhaust their battery
lifetime.
Fig. 8 shows the relationship between the elapsed time
and the number of CMs over the ON environment repre-
senting the worst constraints. The obtained results compare
between the results of the proposed model (fourth scenario)
and the corresponding works in [13]–[15]. It is clearly shown
that the time consumption using the proposed model is much
less than the works in [13]–[15] regardless of the number of
CMs. Moreover, the proposed model presents an enhanced
performance result regardless of increasing the number of
CMs.
Fig. 9 shows the obtained results of the proposed repeated
game model as compared to the three one-shot games. It is
clear that after the equilibrium point, the model maximizes the
Fig. 8. Elapsed time comparison using isotropic transmission over ON
environment.
Fig. 9. Normalized DT based on the proposed repeated game compared to
the three one-shot games using isotropic and nonisotropic transmission and
over ON environment, when the number of malicious CMs equals 1 and 2.
normalized DT leading to the optimal DT, whether the number
of malicious CMs equals 1 or 2. We succeeded to verify that in
both the isotropic and nonisotropic packet transmission. The
minimum normalized DT is achieved by the third scenario
when the CH and CM actions are NB and D, respectively. This
is because the punishment parameter is considered doubled
using (14). The normalized DT of the first and second one-
shot games (first and second scenarios) gets better performance
than the third one but are still less than the proposed game
results.
Fig. 10 shows a comparison between the obtained nor-
malized utility functions of the isotropic and nonisotropic
packet transmission using the proposed repeated game model
over the six environments (OL, ON, UL, UN, IL, and IN)
with each DOI measurements. Generally speaking, all the
obtained results with any environment are close to their
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
(a) (b) (c)
(d) (e) (f)
Fig. 10. Utility comparison of isotropic and nonisotropic packet transmission based on only the proposed repeated game over the six environments at every
DOI, when the number of malicious CMs equals 2. (a) DOI 1. (b) DOI 2. (c) DOI 3. (d) DOI 4. (e) DOI 5. (f) DOI 6.
(a) (b)
Fig. 11. HW failure based on utility representation of isotropic and nonisotropic throughout (a) UL and (b) ON environments, 20% of CMs suffer from the
HW failure.
corresponding ones regardless of the DOI value. It is clear
from Fig. 10(a)–(d), with the IN environment type, at which
the values of σand nare very close, the change of DOI does
not affect the model performance as both the normalized utility
of the isotropic and nonisotropic transmission are overlapped.
Interestingly, we can observe that the nonisotropic exceeds the
isotropic packet transmission by only 1%–4% that reflects the
efficiency of the proposed model.
Detecting the HW failure is discussed using Fig. 11(a)
and (b). At the equilibrium point, all the rational malicious
attempts are refrained, whether with isotropic or nonisotropic
packet transmission. Afterward, any dropping packet or over
packet transmission will be due to the HW failure, not the
SF attack. Fig. 11(a) shows the normalized utilities of the ten
CMs in which the values that exceed 100% are due to the over
packet transmission of the CMs that suffer from the HW fail-
ure such as, CM 4 and 8. The UL environment has the least
path-loss exponent that reflects the minimum effect on the
isotropic radiation of packets. Therefore, the best normalized
utilities of using the proposed model with the isotropic packet
transmission among the six environments are in the UL envi-
ronment in which the utilities reach 96.8%. Conversely, the
ON environment represents the worse effect on the optimal
isotropic normalized utilities, which reach only 93.5%, as
shown in Fig. 11(b).
VII. CONCLUSION
In this article, we have proposed an efficient repeated game
defense model along with the TDMA protocol to effectively
detect the SF attack and HW failure in clustered WSNs-based
IoT systems. The developed model can stimulate the mali-
cious CMs that drop packets to react benevolently, and hence,
facilitate the HW failure detection to achieve the optimal DT.
Consequently, the model solves a prisoner-dilemma problem
and is capable of attaining both the Pareto optimality and NE at
the same state when the maximum DT exists. Moreover, the
model proves beneficial in security enhancement, maximiz-
ing DT, and managing the consequently lost power leading
to robust IoT networks. The simulation results demonstrate
the effectiveness of the proposed model regardless of the
environment, DOI, and isotropic and nonisotropic transmis-
sions achieving robust WSNs-based IoT by preventing the
packet drop due to the SF attack. Furthermore, the model can
maximize the number of successfully transmitted packets. One
of the future works is to evaluate our approach using testbed.
REFERENCES
[1] T.-H. Kim, C. Ramos, and S. Mohammed, “Smart city and IoT,” Future
Gener. Comput. Syst., vol. 76, pp. 159–162, Nov. 2017.
[2] M. T. Lazarescu, “Design of a WSN platform for long-term environmen-
tal monitoring for IoT applications,” IEEE J. Emerg. Sel. Topics Circuits
Syst., vol. 3, no. 1, pp. 45–54, Mar. 2013.
[3] E. Fernandes, A. Rahmati, K. Eykholt, and A. Prakash, “Internet
of Things security research: A rehash of old ideas or new intellec-
tual challenges?” IEEE Security Privacy, vol. 15, no. 4, pp. 79–84,
Aug. 2017.
[4] M. B. M. Noor and W. H. Hassan, “Current research on Internet of
Things (IoT) security: A survey,” Comput. Netw., vol. 148, pp. 283–294,
Jan. 2019.
[5] R. Rani, S. Kumar, and U. Dohare, “Trust evaluation for light weight
security in sensor enabled Internet of Things: Game theory oriented
approach,” IEEE Internet Things J., vol. 6, no. 5, pp. 8421–8432,
Oct. 2019.
[6] I. Butun, S. D. Morgera, and R. Sankar, “A survey of intrusion detection
systems in wireless sensor networks,” IEEE Commun. Surveys Tuts.,
vol. 16, no. 1, pp. 266–282, 1st Quart., 2013.
[7] S. Shen, L. Huang, H. Zhou, S. Yu, E. Fan, and Q. Cao, “Multistage
signaling game-based optimal detection strategies for suppressing mal-
ware diffusion in fog-cloud-based IoT networks,” IEEE Internet Things
J., vol. 5, no. 2, pp. 1043–1054, Apr. 2018.
[8] Z. Han, Game Theory in Wireless and Communication Networks:
Theory, Models, and Applications. Cambridge, U.K.: Cambridge Univ.
Press, 2012.
[9] M. S. Abdalzaher, K. Seddik, M. Elsabrouty, O. Muta, H. Furukawa,
and A. Abdel-Rahman, “Game theory meets wireless sensor networks
security requirements and threats mitigation: A survey,” Sensors, vol. 16,
no. 7, p. 1003, 2016.
[10] M. S. Abdalzaher, K. Seddik, and O. Muta, “Using Stackelberg game
to enhance cognitive radio sensor networks security,” IET Commun.,
vol. 11, no. 9, pp. 1503–1511, 2017.
[11] M. S. Abdalzaher, K. Seddik, O. Muta, and A. Abdelrahman,
“Using Stackelberg game to enhance node protection in WSNs,” in
Proc. 13th IEEE Annu. Consum. Commun. Netw. Conf. (CCNC), 2016,
pp. 853–856.
[12] M. S. Abdalzaher, K. Seddik, and O. Muta, “An effective Stackelberg
game for high-assurance of data trustworthiness in WSNs,” in Proc.
IEEE Symp. Comput. Commun. (ISCC), 2017, pp. 1257–1262.
[13] M. S. Abdalzaher, L. Samy, and O. Muta, “Non-zero-sum game-based
trust model to enhance wireless sensor networks security for IoT
applications,” IET Wireless Sensor Syst., vol. 9, no. 4, pp. 218–226,
Aug. 2019.
[14] M. S. Abdalzaher, K. Seddik, and O. Muta, “Using repeated game
for maximizing high priority data trustworthiness in wireless sensor
networks,” in Proc. IEEE Symp. Comput. Commun. (ISCC), 2017,
pp. 552–557.
[15] J. Duan, D. Gao, D. Yang, C. H. Foh, and H.-H. Chen, “An energy-
aware trust derivation scheme with game theoretic approach in wireless
sensor networks for IoT applications,” IEEE Internet Things J.,vol.1,
no. 1, pp. 58–69, Feb. 2014.
[16] Z. Noshad et al., “Fault detection in wireless sensor networks through
the random forest classifier,” Sensors, vol. 19, no. 7, p. 1568, 2019.
[17] R. N. Duche and N. P. Sarwade, “Sensor node failure detection based on
round trip delay and paths in WSNs,” IEEE Sensors J., vol. 14, no. 2,
pp. 455–464, Feb. 2014.
[18] M. S. Abdalzaher and O. Muta, “Employing game theory and TDMA
protocol to enhance security and manage power consumption in WSNs-
based cognitive radio,” IEEE Access, vol. 7, pp. 132923–132936,
2019.
[19] “Tmote Sky: Ultra low power IEEE 802.15.4 compliant wireless sensor
module,” Data Sheet, Moteiv Corporation, 2006. [Online]. Available:
https://insense.cs.st-andrews.ac.uk/files/2013/04/tmote-sky-datasheet.pdf
[20] S. Raza, S. Duquennoy, T. Chung, D. Yazar, T. Voigt, and U. Roedig,
“Securing communication in 6LoWPAN with compressed IPsec,” in
Proc. Int. Conf. Distrib. Comput. Sensor Syst. Workshops (DCOSS),
2011, pp. 1–8.
[21] G. Zhou, T. He, S. Krishnamurthy, and J. A. Stankovic, “Impact of radio
irregularity on wireless sensor networks,” in Proc. 2nd Int. Conf. Mobile
Syst. Appl. Services, 2004, pp. 125–138.
[22] Y. Wang, “Trust quantification for networked cyber-physical
systems,” IEEE Internet Things J., vol. 5, no. 3, pp. 2055–2070,
Jun. 2018.
[23] H. Al-Hamadi and R. Chen, “Trust-based decision making for health
IoT systems,” IEEE Internet Things J., vol. 4, no. 5, pp. 1408–1419,
Oct. 2017.
[24] L. Jiang, “Sensor fault detection and isolation using system dynamics
identification techniques,” M.S. thesis, Mech. Eng., Univ. Michigan, Ann
Arbor, MI, USA, 2011.
[25] S. Jia, L. Ma, and D. Qin, “Fault detection modeling and analysis in a
wireless sensor network,” J. Sensors, vol. 2018, p. 9, Oct. 2018.
[26] T. Muhammed and R. A. Shaikh, “An analysis of fault detection strate-
gies in wireless sensor networks,” J. Netw. Comput. Appl., vol. 78,
pp. 267–287, Jan. 2017.
[27] H. Liu, A. Nayak, and I. Stojmenovi´
c, “Fault-tolerant algo-
rithms/protocols in wireless sensor networks,” in Guide to Wireless
Sensor Networks. London, U.K.: Springer, 2009, pp. 261–291.
[28] V. Hassija, V. Chamola, G. Han, J. J. P. C. Rodrigues, and M. Guizani,
“DAGIoV: A framework for vehicle to vehicle communication using
directed acyclic graph and game theory,” IEEE Trans. Veh. Technol.,
vol. 69, no. 4, pp. 4182–4191, Apr. 2020.
[29] H. Zhang, Q. Zhang, J. Liu, and H. Guo, “Fault detection and repair-
ing for intelligent connected vehicles based on dynamic Bayesian
network model,” IEEE Internet Things J., vol. 5, no. 4, pp. 2431–2440,
Aug. 2018.
[30] E. Khalastchi, M. Kalech, and L. Rokach, “Sensor fault detection and
diagnosis for autonomous systems,” in Proc. Int. Conf. Auton. Agents
Multiagent Syst., 2013, pp. 15–22.
[31] S. Salim and S. Moh, “An energy-efficient game-theory-based spectrum
decision scheme for cognitive radio sensor networks,” Sensors, vol. 16,
no. 7, p. 1009, 2016.
[32] A. R. Sfar, Y. Challal, P. Moyal, and E. Natalizio, “A game theoretic
approach for privacy preserving model in IoT-based transportation,”
IEEE Trans. Intell. Transp. Syst., vol. 20, no. 12, pp. 4405–4414,
Dec. 2019.
[33] L. F. van Hoesel, T. Nieberg, H. Kip, and P. J. Havinga, “Advantages
of a TDMA based, energy-efficient, self-organizing MAC protocol for
WSNs,” in Proc. IEEE 59th Veh. Technol. Conf. (VTC-Spring),vol.3,
2004, pp. 1598–1602.
[34] G. Lu, B. Krishnamachari, and C. S. Raghavendra, “An adaptive energy-
efficient and low-latency MAC for data gathering in wireless sensor
networks,” in Proc. IEEE 18th Int. Parallel Distrib. Process. Symp.,
2004, pp. 224–231.
[35] V. C. Gungor, B. Lu, and G. P. Hancke, “Opportunities and challenges
of wireless sensor networks in smart grid,” IEEE Trans. Ind. Electron.,
vol. 57, no. 10, pp. 3557–3564, Oct. 2010.
ABDALZAHER AND MUTA: GAME-THEORETIC APPROACH FOR ENHANCING SECURITY AND DATA TRUSTWORTHINESS
Mohamed S. Abdalzaher (Member, IEEE) received
the B.Sc. degree (Hons.) from Obour High Institute
for Engineering and Technology, Cairo, Egypt,
in 2008, and the M.Sc. degree in electron-
ics and communications engineering from Ain
Shams University, Cairo, in 2012, respectively,
and the Ph.D. degree from the Electronics and
Communications Engineering Department, Egypt-
Japan University of Science and Technology,
Madinet Borg Al Arab, Egypt, in 2016.
He is an Assistant Professor with the National
Research Institute of Astronomy and Geophysics, Cairo. He was a special
research student with Kyushu University, Fukuoka, Japan, from 2015 to 2016.
In April 2019, he joined the Center for Japan-Egypt Cooperation in Science
and Technology, Kyushu University, where he is currently working as a
Postdoctoral Researcher. His research interests include data communication
networks, wireless communications, WSNs security, IoT, and deep learning.
Dr. Abdalzaher is a TPC Member with Vehicular Technology Conference
and International Japan-Africa Conference on Electronics, Communications
and Computers and a reviewer of the IEE E INTERNET OF THINGS JOURNAL,
IEEE SYSTEMS JOURNAL, IEEE ACCESS, and IET Journals.
Osamu Muta (Member, IEEE) received the B.E.
degree from Ehime University, Matsuyama, Japan,
in 1996, the M.E. degree from Kyushu Institute of
Technology, Iizuka, Japan, in 1998, and the Ph.D.
degree from Kyushu University, Fukuoka, Japan, in
2001.
In 2001, he joined the Graduate School of
Information Science and Electrical Engineering,
Kyushu University as an Assistant Professor, where
he has been an Associate Professor with the
Center for Japan-Egypt Cooperation in Science and
Technology since 2010. His current research interests include signal process-
ing techniques for wireless communications and powerline communications,
MIMO, and nonlinear distortion compensation techniques for high-power
amplifiers.
Dr. Muta received the Active Research Award from IEICE Technical
Committee of Radio Communication Systems in 2005, and the Chairman’s
Awards for excellent research from IEICE Technical Committee of
Communication Systems in 2014, 2015, and 2017. He is a Senior Member
of IEICE.
