You are on page 1of 15

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO.

11, NOVEMBER 2021 7011

Impact of Hardware Impairment and Co-Channel


Interference on Security-Reliability Trade-Off
for Wireless Sensor Networks
Bin Li , Yulong Zou , Senior Member, IEEE, Jia Zhu, and Weifeng Cao

Abstract— In this paper, we investigate the security and relia- to the limitations of WSN nodes on size, cost, and power
bility trade-off (SRT) for a wireless sensor network (WSN) which energy, their computing power and transmission capacity
consists of one cluster head (CH), multiple users, and one eaves- are much weaker than other networks, which makes WSN
dropper (E), where the eavesdropper attempts to tap the con-
fidential transmission between multiple users and CH. We first vulnerable to attacks. And, security has become one of the
present two multiuser scheduling schemes, namely the selection main obstacles for large-scale application of WSNs, which
combining (SC) scheme and the switch-and-examine combining has attracted extensive attention from researchers and brought
with post-selection (SECps) scheme to enhance physical-layer about a variety of security techniques [1]–[5]. There are dif-
security (PLS) of the WSN system. For comparison purposes, ferent security mechanisms at different layers of WSNs, such
the round-robin scheduling (RRS) scheme is also considered as
a baseline. Then, we analyze the SRT of the RRS, SECps, and as key management mechanisms, cryptography algorithms,
SC schemes by deriving their closed-form outage and intercept identity authentication, two-way authentication schemes, and
probability expressions under the joint impact of hardware so on [6]. However, these security mechanisms require high
impairment (HI) and co-channel interference (CCI). And, secrecy computing power and energy of WSN nodes.
diversity analysis of the three schemes is also carried out. Numer- Alternatively, physical-layer security (PLS) [7] which
ical results show that the SC and SECps scheme outperform the
RRS scheme in terms of the SRT performance, where the SC is to encrypt symbols at the bottom layer can reduce the
scheme can achieve the best SRT performance at the cost of an complexity of encryption algorithms and achieve a good
increased channel estimation complexity. Moreover, a trade-off defense against illegal base stations and eavesdroppers. After
between the security and reliability can be achieved for each of decades of research, various PLS techniques [8], such as
the RRS, SECps, and SC schemes by adjusting the overall data cooperative jamming [9], [10], cooperative relaying [11]–[13],
rate and signal-to-noise ratio (SNR).
non-orthogonal multiple access (NOMA) [14]–[16], multiple-
Index Terms— Physical-layer security, security-reliability input-multiple-output (MIMO) [17], [18], multiuser
trade-off, multiuser scheduling, hardware impairment, diversity [19]–[23], and so on, have emerged. Among
co-channel interference, wireless sensor network.
them, multiuser scheduling is widely used because it does not
require additional resources. In [19], the PLS of multiuser
I. I NTRODUCTION
wireless networks is reviewed. In [20], multiuser scheduling

T ODAY, wireless sensor networks (WSNs) are widely


used in various fields, such as industry, agriculture,
military, business, environmental monitoring, and so on.
was studied in cognitive radio networks and the intercept
probability and achievable secrecy rate were analyzed for
different multiuser scheduling schemes. Deng et al. [21]
Furthermore, WSN which is responsible for the data acqui- proposed a dual user selection scheme to enhance the PLS of
sition and processing of the perception layer is an important multiuser systems, where one of the two selected users acts
part of Internet of Things (IoT) and plays a vital role in the as a served user and the other one acts as a jamming user.
development and application of IoT technology. However, due In [22], a threshold-based multiuser scheduling scheme was
proposed to reduce the feedback of CSI and exact closed-form
Manuscript received April 26, 2020; revised October 9, 2020 and
January 4, 2021; accepted May 9, 2021. Date of publication May 20, ergodic secrecy rate expressions were derived. Moreover,
2021; date of current version November 11, 2021. This work was sup- NOMA and multiuser diversity were jointly considered
ported in part by the National Natural Science Foundation of China under in [23].
Grant 91738201 and Grant 62071253; in part by the Open Project Funding
of State Key Laboratory of Smart Grid Protection and Operational Control At present, most of the existing studies are carried out
of China (Research on technologies of relay protection service over 5G under ideal conditions, ignoring many factors which can not be
network) under Contract SGNR0000KJJS1907531; and in part by the Nanjing avoided in the practical communication systems and will affect
University of Posts and Telecommunications Science Foundation (NUPTSF)
under Grant NY220120. The associate editor coordinating the review of this the communication quality. Among them, hardware impair-
article and approving it for publication was P. Li. (Corresponding authors: ment (HI) and co-channel interference (CCI) have become
Yulong Zou; Jia Zhu.) research hotspots. On the one hand, HI caused by I/Q imbal-
The authors are with the School of Telecommunication and Information
Engineering, Nanjing University of Posts and Telecommunications, Nan- ance [24], phase noise and amplifier nonlinearities [25] is a
jing 210003, China (e-mail: binli@njupt.edu.cn; yulong.zou@njupt.edu.cn; recently proposed concept of equivalent noise. In [26], ergodic
jiazhu@njupt.edu.cn; cwf8268@126.com). secrecy rate was explored in a particular scenario where both
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TWC.2021.3079902. the base station and legitimate users are affected by HIs,
Digital Object Identifier 10.1109/TWC.2021.3079902 whereas the eavesdropper employs ideal hardware. The impact
1536-1276 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7012 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

of HIs on both decode-and-forward (DF) and amplify-and-


forward (AF) protocols based dual-hop relaying networks was
quantified and both exact and asymptotic outage probabilities
were derived in [27]. The impact of HIs was investigated in
multiple antenna systems [28]. Moreover, in [29], the authors
studied the performance of a land mobile satellite (LMS)
communication system, where a satellite serves as a relay is
affected by both HI and CCI.
On the other hand, due to the shortage of spectrum
resources, spectrum reuse technology is developed to improve
the spectrum utilization, which brings the issue of CCI.
Although spectrum resources can be released through cog-
nitive radio technology [30], which can improve the utiliza-
tion of spectrum, the spectrum resources are still scarce.
In [31], two multiuser scheduling schemes were proposed Fig. 1. A multiuser wireless system with joint impact of HI and CCI.
based on the availability of channel state informations (CSIs)
of CCI links which can be predicted by CSI estimation [32], impact of HI and CCI over Rayleigh fading channels.
and closed-form secrecy outage probability expressions were Furthermore, we carry out the secrecy diversity analysis
derived for the two proposed schemes. Multiuser scheduling of these three presented schemes by characterizing an
was further investigated in the presence of CCI in [33], asymptotic behavior of the outage probability with a
where a threshold based multiuser scheduling scheme was given intercept probability in the high signal-to-noise
proposed to improve PLS and reduce computation overhead ratio (SNR) region.
of the system. Additionally, in [34], both selection coopera- • Numerical results show that the SRT performance of
tion and opportunistic relaying were investigated in the face SECps and SC schemes performs better than that of the
of CCI. Reference [35] inferred that the presence of CCI RRS scheme, where the best SRT performance can be
significantly degrades the outage performance of the system achieved by the SC scheme at the cost of estimating
over Nakagami-m fading channels. Differing from [31]–[35] more wireless channels. Moreover, a trade-off between
where only the effect of CCI on system performance is security and reliability for each of the RRS, SECps, and
considered, the authors of [36] presented a new architecture SC schemes can be achieved by adjusting the overall data
receivers to cancel CCI by using baseband chain and signal rate and SNR.
radio frequency. The rest of this paper is organized as follows. The sys-
Furthermore, a few literatures have studied the joint impact tem model of a multiuser system is proposed in Section II.
of HI and CCI. To be specific, by considering HI suffered at In Section III, we present the RRS, SECps, and SC schemes,
both the transmitter and receiver, [37] proposed a general ana- derive closed-form expressions in terms of their outage and
lytical framework for MIMO space shift keying systems and intercept probabilities under the joint impact of HI and CCI.
derived a closed-form average bit error probability (ABEP) In Section IV, we carry out the secrecy diversity analysis
expression in CCI environments. Moreover, relay selection of the three presented schemes. In Section V, numerical and
was considered under the impact of both HI and CCI, and simulation results are given and discussed. Finally, conclusions
both exact and asymptotic closed-form outage probability are presented in Section VI.
expressions for different relay selection schemes were given
in [38]. However, the joint impact of HI and CCI has not II. S YSTEM M ODEL
been investigated for secrecy scenarios. Motivated by this, Fig. 1 shows a wireless sensor network composed
we explore the security-reliability trade-off (SRT) analysis of a cluster head (CH), multiple users denoted by
of a wireless sensor network which consists of one cluster U = {Um |m = 1, 2, 3, . . . , M }, and an eavesdropper (E).
head (CH), multiple user, and one eavesdropper (E), where As depicted in Fig. 1, one user Um is scheduled to trans-
the eavesdropper attempts to tap the confidential transmission mit its confidential message xm to the CH with power
between multiple users and CH under the joint impact of PU while the eavesdropper attempts to wiretap the trans-
HI and CCI. The main contributions of this paper can be mission between Um and CH under the joint impact
summarized as follows. of HI and CCI. Assume that the CH and eavesdrop-
• We present the switch-and-examine combining with per are interfered by K co-channel interferers denoted by
post-examine selection (SECps) scheme and the selec- Ic = {Ik |k = 1, 2, 3, . . . , K } and L co-channel interferers
tion combining SC) scheme to improve the SRT perfor- denoted by Ie = {Il |l = 1, 2, 3, . . . , L }, respectively. More-
mance of the system. For SRT performance comparison over, all interferers are assumed to transmit their signals
purposes, the traditional round-robin scheduling (RRS) (i.e., sik and sil ) with power PI . In addition, throughout this
scheme is also considered as a baseline. paper, we employ Rayleigh fading channels to model all of
• We analyze the SRT performance of the RRS, SECps, and the main, interference, and wiretap links. Let hum c , hum e ,
SC schemes in terms of deriving their closed-form out- gik , and gil represent the channel gains of Um -CH, Um -E,
age and intercept probability expressions under the joint Ik -CH, and Il -E links, respectively. In general, the co-channel

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7013

interferers are far enough from CH and eavesdropper such and


that, the distance from the interferers to CH and eavesdropper ψ|hum e |2
is assumed to be the same, respectively [39], [40]. Therefore, SINRe = . (6)
2 
L
2
the Rayleigh fading coefficients of CCI links, i.e., gik and gil , μ2 ψ|hum e | + (1 + μ2 )ψ  |gil | + 1
are assumed to be independent identically distributed (i.i.d.) l=1

zero-mean complex Gaussian random variables with variances Accordingly, the channel capacity of main and wiretap
2 2 channel can be obtained as
of σic and σie , respectively. Hence, the received signals at CH
and eavesdropper can be written as
Cc = log2 (1 + SINRc ) , (7)
 K 

yc = PU hum c (xm +ηm )+ PI gik (sik +ηik )+ηc +nc, and
k=1
Ce = log2 (1 + SINRe ) , (8)
(1)
where SINRc and SINRe are given by (5) and (6), respectively.
and As discussed in [13], if the capacity of wiretap channel is
 L 
 larger than the difference between the overall rate (Ro ) and
ye = PU hum e (xm +ηm )+ PI gil (sil +ηil )+ηe +ne , secrecy rate (Rs ), there will be an intercept event, while if the
l=1 capacity of main channel is smaller than Ro , there will be an
(2) outage event. Thus, we can describe the intercept and outage
where ηm , ηik , and ηil are the distortion noises at the sched- probability of our system as
uled user and interferers due to the transmit HIs, ηc , ηe are Pint = Pr (Ce > Ro − Rs ) , (9)
the distortion noises at the CH and eavesdropper due to the
receive HIs, nc and ne stand for the zero-mean additive white and
Gaussian noises (AWGNs) received at CH and eavesdropper
Pout = Pr (Cc < Ro ) . (10)
with variance N0 .
Following [38], the distortion noises ηm , ηik , ηil ,
III. M ULTIUSER S CHEDULING S CHEMES AND
ηc , and ηe can be respectively modeled as zero-mean
S ECURITY-R ELIABILITY T RADE -O FF A NALYSIS
circularly-symmetric complex  Gaussian distribution  with
2 2 K 2 Messages transmitted over the network have different
variance σm , σi2k , σi2l , σc2 |hum c | PU + k=1 |gik | PI , and
 L  requirements of reliability and security. Security is more
2 2
σe2 |hum e | PU + l=1 |gil | PI . Without loss of generality, important for confidential messages, while reliability is
we consider that the cluster head and eavesdropper have more important for other messages. The goal of multiuser
different HI levels, while the multiple users and the co-channel scheduling is to reduce the intercept probability of the
2
interferers have the same HI level, i.e., σm = σi2k = network when the reliability of the network is satisfied, and
2 2
σil = σ1 . Hence, the received signal-to-interference-plus-noise vice versa. In this section, two multiuser scheduling schemes
ratio (SINR) at CH can be formulated as are proposed to improve the SRT of our system, including SC
and SECps schemes. For comparison purposes, the traditional
SINRc
2
RRS scheme is also considered as a benchmark. The closed-
ψ|hum c | form intercept and outage probabilities of the SC, SECps as
= ,
2 
K
2 well as RRS schemes are derived under the joint impact of
(σ12 + σc2 ) ψ|hum c | + |gik | (1 + σ12 + σc2 )ψ  + 1
k=1 HI and CCI.
(3)
A. RRS Scheme
where ψ = PNU0 and ψ  = N PI
0
denote the signal-to-noise The RRS scheme allows all users of the system to com-
ratio (SNR) and interference-to-noise ratio (INR), respectively. municate with the CH in turn. Now, it is Um ’ turn to send
Similarly, the received SINR at eavesdropper can be expressed its confidential message to the CH. Here, we derive the
as closed-form outage and intercept probability expressions of
SINRe RRS scheme.
2 1) Outage Probability: Combining (5) and (7) with (10),
ψ|hum e |
= . the outage probability of the Um -CH transmission can be
2 2 2 
K
2 2 2 
(σ1 + σe ) ψ|hum c | + |gik | (1 + σ1 + σe )ψ + 1 given by
k=1 RRS
(4) Pout,m ⎛ ⎞

Let μ1 = σ12 + σc2 and μ2 = σ12 + σe2 , (3) and (4) can be ⎜ ψ|hum c |2 ⎟
⎜ ⎟
further rewritten as = Pr ⎜ < Δ1 ⎟ ,
⎝ 
K ⎠
2 μ1 ψ|hum c |2 + (1 + μ1 ) ψ  |gik |2 + 1
ψ|hum c | k=1
SINRc = , (5) (11)
2 
K
2
μ1 ψ|hum c | + (1 + μ1 )ψ  |gik | + 1
k=1 where Δ1 = 2Ro − 1.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7014 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

The equation (11) can be further obtained as B. SC Scheme


⎛ ⎞
1 − μ1 Δ1 2 In this section, the SC scheme is presented to improve the
ψ |hum c |
⎜ Δ1 ⎟ SRT performance of our system. A user who maximizes the
⎜ ⎟
RRS
Pout,m = Pr ⎜ K
 ⎟. (12) average channel gain of Um -CH links is selected as the best
⎝  2 ⎠
< (1 + μ1 ) ψ |gik | + 1 user. Hence, the criterion of user selection for SC scheme is
k=1 given by
2
If μ1 ≥ Δ1 −1 , then Pout,m
RRS
= 1, and if μ1 < Δ1 −1 , then Best user = arg max |hum c | . (19)
m∈U
(12) can be rewritten as (see Appendix A)
K The main and wiretap channel capacities of SC scheme are
λic given by
Pout,m = 1 −
RRS
e−λum c , (13)
λic + λum c
Cub c
⎛ ⎞
where λic = (1+μ11)ψ σ2 and λum c = (1−μ1 ΔΔ11)ψσ2 , σu2 m c
um c
ic
is the variance of hum c . ⎜ ψ|hub c |
2 ⎟
⎜ ⎟
The outage probability of RRS scheme can be obtained by = log2 ⎜1 + ⎟,
⎝ 2 
K
2 ⎠
averaging that of all M users, i.e., μ1 ψ|hub c | + (1 + μ1 ) ψ  |gik | + 1
k=1

⎨ 1,
⎪ if μ1 ≥ Δ−1
1
(20)
M
RRS
Pout = 1  RRS (14) and

⎩M Pout,m , if μ1 < Δ−1
1
m=1 Cub e
RRS
⎛ ⎞
where is given by (13).
Pout,m
2) Intercept Probability: Combining (6) and (8) with (9), ⎜ ψ|hub e |2 ⎟
⎜ ⎟
the intercept probability of the Um -CH transmission can be = log2 ⎜1 + ⎟,
⎝ 2 
L
2 ⎠
given by μ2 ψ|hub e | + (1 + μ2 ) ψ  |gik | + 1
l=1
(21)
RRS
Pint,m
⎛ ⎞ where ’ub ’ denotes the best user.
Next, we derive the closed-form outage and intercept prob-
⎜ ψ|hum e |
2 ⎟
⎜ ⎟ ability expressions of SC scheme.
= Pr ⎜ > Δ2 ⎟ ,
⎝ 2 
L
2 ⎠ 1) Outage Probability: Combining (10) with (20) and
μ2 ψ|hum e | + (1 + μ2 ) ψ  |gil | + 1 doing some algebraic manipulations, the outage probability of
l=1
(15) SC scheme can be formulated as
⎛ ⎞
1 − μ1 Δ1 2
where Δ2 = 2Ro −Rs − 1. ⎜ ψmax |hum c | ⎟
Δ1 m∈U
The equation (15) can be further expressed as
SC
Pout = Pr ⎜
⎝ 
K
⎟.
⎠ (22)
 2
⎛ ⎞ < (1 + μ1 ) ψ |gik | + 1
1 − μ2 Δ2 2 k=1
⎜ ψ |hum e | ⎟
Δ2 −1
= 1, and if μ1 < Δ1 −1 , then (22)
= 1 − Pr ⎜ ⎟ . (16) If μ1 ≥ Δ1 , then SC
RRS
Pint,m Pout
⎝ 

L
2 ⎠
< (1 + μ2 ) ψ |gil | + 1 can be rewritten as (see Appendix B)
l=1
M
2 −1
−1 −1 |D w |
If μ2 ≥ Δ2 , then = 0, and if μ2 < Δ2 , then
RRS
Pint,m SC
Pout,I = 1+ (−1)
similar to the derivation of (13), (16) can be rewritten as w=1
L  ⎛ ⎞K
λie  λic
RRS
Pint,m = e−λum e , (17) × exp − λum c ⎝  ⎠ ,
λie + λum e λic + λum c
m∈D w
m∈D w
where λie = (1+μ21)ψ σ2 and λum e = (1−μ2 ΔΔ22)ψσ2 , σu2 m e (23)
ie um e
is the variance of hum e .
where Dw stands for the w-th non-empty subset of {U} and
The intercept probability of RRS scheme can be obtained
|Dw | denotes the set cardinality of Dw .
by averaging that of all M users, i.e.,
Finally, the outage probability of SC scheme can be obtained

⎨ 0,
⎪ if μ2 ≥ Δ−1
2
as
M

RRS
Pint = 1  RRS (18) 1, if μ1 ≥ Δ−1

⎩M Pint,m , if μ2 < Δ−1
2 Pout =
SC 1
(24)
m=1
SC
Pout,I , if μ1 < Δ−1
1

RRS SC
where Pint,m is given by (17). where Pout,I is given by (23).

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7015

2) Intercept Probability: Combining (9) with (21), the inter-


cept probability of SC scheme can be expressed as
⎛ ⎞
⎜ ψ|hub e |
2 ⎟
⎜ ⎟
SC
Pint = Pr ⎜ > Δ2 ⎟ .
⎝ 2 
L
2 ⎠
μ2 ψ|hub e | +(1+μ2 ) ψ  |gil | + 1
l=1
(25)
Employing the law of total probability, we can rewrite (25)
as
SC
Pint
⎛ ⎞
ψ|hur e |2
M ⎜ > Δ2 ,⎟
 ⎜ 2 L
2 ⎟
= ⎜
Pr ⎜ 2
μ ψ|h ur e | +(1+μ2 ) ψ  |g il | + 1 ⎟.

r=1 ⎝ l=1
2 2 ⎠
max |hum c | < |hur c |
m∈{U −Ur }
(26)
If μ2 ≥ Δ2 −1 , then Pint
SC
= 0, and if μ2 < Δ2 −1 , then the
equation (26) can be further rewritten as
SC Fig. 2. Flowchart for the SECps scheme.
Pint,I
M
 
2 2
= Pr max |hum c | < |hur c |
m∈{U −Ur } in Fig. 2, the specific user selection policy of SECps scheme
r=1
⎛ ⎞ works as follows:
⎜ ψ|hur e |2 ⎟ 1) The U1 -CH link is first estimated, and if the received SNR
⎜ ⎟
× Pr ⎜ > Δ2 ⎟ at CH is not less than the given threshold, i.e., δ1 ≥ δT ,
⎝ 
L ⎠
μ2 ψ|hur e |2 + (1 + μ2 ) ψ  |gil |2 + 1 the user U1 will be selected for data transmission.
l=1
L 2) The user Um (m ∈ (2, . . . , M )) is selected to transmit
M
 λie its signals to CH when δq (q ∈ (1, . . . , m − 1)) is less
= e−λur e
r=1
λie + λur e than the given threshold, while δm is not less than the
⎡ ⎛ ⎞−1 ⎤ given threshold, i.e., δ = max(δ1 , . . . , δm−1 ) < δT and
2M
−1
−1  σu2 c
⎢ ⎥ δ = max(δ1 , . . . , δm ) = δm ≥ δT .
(−1)| r | ⎝1 +
Dt r ⎠
×⎣1 + ⎦, (27) 3) All wireless links have been estimated, and all received
t=1 t
σu2 m c
m∈Dr SNRs are smaller than the given threshold, i.e., δ =
max(δ1 , . . . , δM ) < δT , the user who maximizes the
where Drt denotes the t-th non-empty subset of {U − Ur } and
received SNR at CH will be selected.
|Drt | denotes the set cardinality of Drt .
Finally, the intercept probability of SC scheme can be Based on above discussion, we can obtain the cumulative
obtained as distribution function (CDF) of δ as

0, if μ2 ≥ Δ−1
2
Pint =
SC
(28) Fδ (x) = Pr (δ < x)
SC
Pint,I , if μ2 < Δ−1
2
M
SC
where Pint,I is given by (27). = Pr [δ = max (δ1 , δ2 , . . . , δm ) & δ < x]
m=1

⎪ M m−1

C. SECps Scheme ⎪


⎪ Fδ (x) − Fδq (δT )


1
In this subsection, a threshold-based multiuser scheduling ⎨ m=2 q=1
scheme, namely the SECps scheme, is presented to enhance = × (1 − Fδm (x)) , if x ≥ δT (29)


the SRT performance and reduce the computation overhead ⎪
⎪ M


of the system. The SECps scheme can be applied to wireless ⎪
⎩ Fδm (x) , if x < δT
networks with limited computational power and energy due to m=1
cost and node structures, such as the Internet of Things (IoT)
2
and wireless sensor networks (WSNs). Let δm = ψ|hum c | Then, we evaluate the outage and intercept probabilities of
denote the received SNR of Um -CH link. The SECps scheme SECps scheme.
is adopted to select a user whose received SNR is beyond a 1) Outage Probability: Similar to the RRS and SC schemes,
given threshold (δT ) to communicate with the CH. As shown if μ1 ≥ Δ1 −1 , then Pout
SECps
= 1, and if μ1 < Δ1 −1 , then the

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7016 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

outage probability of SECps scheme can be formulated as μ2 ≥ Δ−1 2 , the outage and intercept probabilities of the
⎛ ⎞ system are 1 and 0, respectively, which is no need for analysis.
1 − μ1 Δ1 2
⎜ ψ|hus c | ⎟ According to [41], we derive an asymptotic outage prob-
Δ1
Pout,I = Pr ⎜ ⎟
SECps
⎝ K ⎠ ability expression as a function of the intercept probability
 2
< (1 + μ1 )ψ |gik | + 1 with γ → ∞, then we derive the secrecy diversity gain as a
k=1
 ∞  ratio of the logarithmic asymptotic outage probability to the
Δ1 logarithmic γ, which is described as
= Fδ x fY1 (x) dx, (30)
1 1 − μ1 Δ1
log Pout (γ, Pint )
where ‘us ’ denotes the selected user. ds = − lim . (39)
γ→∞ log γ
Proceeding as in Appendix C, (30) can be further obtained
where Pout (γ, Pint ) denotes an outage probability as a func-
as
⎧ tion of an intercept probability Pint and SNR γ.
⎪ Δ1
⎨ Ξ1 + Ξ2 , if δT ≥
SECps
Pout,I = 1 − μ1 Δ1 (31) A. RRS Scheme
⎪ Δ1
⎩ Ξ3 , if δT < In this subsection, we present the secrecy diversity analysis
1 − μ1 Δ1
for the conventional RRS scheme. Based on the above assump-
where Ξ1 , Ξ2 , and Ξ3 are given by (32), (33), and (34), shown tions, (13) and (17) can be simplified as
at the bottom of the next page, respectively. 
Δ1
Finally, the outage probability of SECps scheme is RRS
Pout,m = 1 − exp − 2
, (40)
expressed as (1 − μ1 Δ1 ) γσuc
 and
1, if μ1 ≥ Δ−1 
SECps
Pout = 1
(35) Δ2
Pout,I , if μ1 < Δ−1
SECps Pint,m = exp −
RRS
2
. (41)
1 (1 − μ2 Δ2 ) γσue
SECps
where Pout,I is given by (31). Combining (41) with Δ2 = 2Ro −Rs − 1, we have
 
2) Intercept Probability: Similarly, if μ2 ≥ Δ2 −1 , then 2
γσue ln Pint,m
RRS
Ro Rs
SECps
Pint = 0, and if μ2 < Δ2 −1 , then the intercept probability 2 =2 2 ln P RRS − 1
+1 . (42)
μ2 γσue int,m
of SECps scheme is formulated as
M When γ → ∞, by using the formula ex − 1 ∼ x (x → 0),

SECps
Pint,I = Fδm (δT )Pint,I
SC
+ (1 − Fδ1 (δT )) Pint,1
RRS (40) can be further obtained as
m=1 Δ1
M m−1
lim P RRS = 2
(43)
  γ→∞ out,m (1 − μ1 Δ1 ) γσuc
+ Fδq (δT ) (1 − Fδm (δT )) Pint,m
RRS
. (36)
m=2 q=1
Substituting (42) and Δ1 = 2Ro − 1 into (43) yields (44),
shown at the bottom of the next page. Therefore, we can
Substituting (17), (27), (C.2), and (C.4) into (36), we can readily have (45), shown at the bottom of the next page, which
express the intercept probability of SECps scheme as (37), shows that the conventional RSS scheme achieves a secrecy
shown at the bottom of the next page. diversity order of one. This means that as the SNR γ increases
Finally, the intercept probability of SECps scheme can be to infinity, the outage probability would approach to zero with
obtained as a required intercept probability constraint.

SECps 0, if μ2 ≥ Δ−12
Pint = (38) B. SC Scheme
Pint,I , if μ2 < Δ−1
SECps
2
In this subsection, we present the secrecy diversity analysis
SECps
where Pint,I is given by (37). for proposed SC scheme. Similarly to (40) and (41), (23) and
(27) can be simplified as
IV. S ECRECY D IVERSITY G AIN A NALYSIS M
Δ1
SC
Pout = 1 − exp − 2
, (46)
In this section, we present the secrecy diversity gain analysis (1 − μ1 Δ1 ) γσuc
for the proposed SECps and SC schemes as well as the and
conventional RRS scheme by characterizing an asymptotic M−1 
behavior of the outage probability with an intercept probability  m 1 Δ2
Pint = M
SC
(−1) exp − 2
,
constraint in the high SNR region. We consider a special m=0
m+1 (1 − μ2 Δ2 ) γσuc
scenario where the effect of CCI is ignored. For the purpose (47)
of mathematical tractability, we assume that the Rayleigh

M−1
m 1
fading coefficients of all main and wiretap channels, i.e., hum c where M (−1) m+1 can be mathematically transformed
and hum e (m = 1, 2, . . . , M ), are i.i.d. zero-mean complex m=0
Gaussian random variables with a variance of σuc 2 2
and σue , to 1, hence, we can obtain 2Ro from (47) as

respectively. Moreover, only the case of μ1 < Δ−1 1 and Ro Rs
2
γσue ln Pint
SC
2 =2 +1 . (48)
μ2 < Δ−1 2 is considered, because when μ1 ≥ Δ−1 1 and 2 ln P SC − 1
μ2 γσue int

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7017

Substituting (48) into (46) yields (49), shown at the bottom C. SECps Scheme
of the page. Therefore, the secrecy diversity gain of proposed
SC scheme is In this subsection, we present the secrecy diversity analysis
for proposed SECps scheme. (31) and (36) can be sim-
log Pout
SC plified as (51), shown at the bottom of the next page,
s = − lim
dSC
γ→∞ log γ
= M, (50) and

which indicates that the proposed SC scheme can achieve the 


secrecy diversity order of M higher than the conventional SECps Δ2
Pint = Φ exp − 2
, (52)
RRS scheme. (1 − μ2 Δ2 ) γσue

 
 ⎛ ⎞ K

M
2 −1  γ K, λic + λum c ξ
γ (K, λic ξ) |D w | ⎝ λic ⎠ m∈D w
Ξ1 = + (−1) exp − λum c  , (32)
(K − 1)! w=1 w
λic + λum c (K − 1)!
m∈D
m∈D w
K  M
Γ (K, λic ξ) Γ (K, (λic + λu1 c ) ξ) λic
Ξ2 = − exp (−λu1 c ) − exp (−λum c )
(K − 1)! (K − 1)! λic + λu1 c m=2
⎡ ⎛ ⎞⎤
2m−1
−1  K
v
|D | δT ⎠⎦ Γ (K, (λic + λum c ) ξ) λic
×⎣1 + (−1) exp ⎝− , (33)
v=1
ψσu2 q c (K − 1)! λic + λum+1 c
q∈D v
⎡ ⎛ ⎞⎤
−K  M 2m−1
−1 
λu1 c ⎣1 +
v
|D | δ T ⎠⎦
Ξ3 = 1 − exp (−λu1 c ) 1 + − (−1) exp ⎝− 2
λic m=2 v=1 v
ψσ uq c
q∈D
−K
λum c
× exp (−λum c ) 1 + . (34)
λic

⎡ M  ⎤
2 −1 
|D w | δT
= ⎣1 + ⎦
SECps
Pint,I (−1) exp −
w=1
ψσu2 m c
m∈D w
⎡ ⎛ ⎞−1 ⎤
M
 2M
−1
−1  σu2 c
⎢ ⎥
(−1)| r | ⎝1 +
Dt r ⎠
× ⎣1 + 2 ⎦
r=1 t=1
σu c
m∈Drt m

L  L
λie δT λie
× exp(−λur e ) + exp − 2 exp(−λu1 e )
λie + λur e ψσu1 c λie + λu1 e
⎡ ⎛ ⎞ ⎤
M 2m−1
−1  δT
⎣ |D v |
+ 1+ (−1) exp ⎝− 2
⎠⎦
m=2 v=1 v
ψσ u c
q∈D q

 L
δT λie
× exp − 2 exp(−λum e ) (37)
ψσum c λie + λum e

⎛⎛ ⎞−1 ⎞−1
   −1
2
⎜⎝ Rs γσue ln Pint,m
RRS
2 ⎟
lim Pout,m
RRS
=⎝ 2 2 ln P RRS −
+1 −1 − μ1 ⎠ γσuc ⎠ . (44)
γ→∞ μ2 γσue int,m 1

log Pout,m
RRS
dRRS
s = − lim = 1, (45)
γ→∞ log γ

⎛⎛ ⎞−1 ⎞−1
   −1
2
⎜⎝ Rs γσue ln Pint,m
SC
2 ⎟
lim Pout
SC
= M⎝ 2 2 ln P SC
+1 −1 − μ1 ⎠ γσuc ⎠ (49)
int,m − 1
γ→∞ μ2 γσue

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7018 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

where Φ is given by (53), shown at the bottom of the page.


Then, we have
⎛   ⎞
2 SECps
γσue ln Pint /Φ
2Ro = 2Rs ⎝   + 1⎠ . (54)
2 ln P SECps /Φ
μ2 γσue − 1
int

When γ → ∞, by using the formula ex − 1 ∼ x (x → 0),


(51) can be rewritten as (55), shown at the bottom of the page.
Substituting (54) into (55), we can obtain the secrecy
diversity gain of proposed SECps scheme as

⎪ Δ1
⎨ M, if δT ≥
dSECps = 1 − μ1 Δ1 (56)
s
⎪ Δ1
⎩ 1, if δT <
1 − μ1 Δ1
which shows that with an arbitrarily low intercept probability,
the overall outage probability of proposed SECps scheme Fig. 3. SRT of the RRS, SECps, and SC schemes for different MERs with
M = 5, SNR ∈ [0, 20 dB], INR = 4 dB, Ro = 2 bit/s/Hz, K = L = 2,
would asymptotically decrease to zero, as the SNR γ increases δT = 8 dB, and μ1 = μ2 = 0.03.
to infinity. Moreover, (56) also shows that when the threshold
is high, the SRT performance of proposed SECps scheme
is similar to that of SC scheme, otherwise, it is similar to 2
is given by 0.1, i.e., σic = σie2
= 0.1. And, a secrecy
RRS scheme. This finding is consistent with the results data rate of Rs = 0.5 is also assumed. Additionally, main-
reflected in Fig. 8. to-eavesdropping ratio (MER) which is defined as the ratio
of the average gain of main channel to that of wiretap
V. N UMERICAL R ESULTS AND D ISCUSSIONS channel is introduced to illustrate the effect of its changes
In this section, some numerical SRT results of the SC on SRT performance of our system. One can see from
and SECps schemes as well as the traditional RRS scheme Figs. 3-10, the analytical outage and intercept probability
in terms of their outage and intercept probabilities are pre- results of the RRS, SECps, and SC schemes match well with
sented. Monte-Carlo simulation results are also provided to their corresponding simulation results, which indicates that our
validate the correctness of the theoretical SRT analysis. In our theoretical SRT analysis stated above is correct.
numerical evaluation, the Rayleigh fading coefficients of all In Fig. 3, we plot the SRT of the RRS, SECps, and SC
users to CH links are assumed to have the same variance, schemes for different MERs with M = 5, SNR ∈ [0, 20 dB],
which is given by 1, i.e., σu2 m c = 1, m ∈ (1, 2, . . . , M ). INR = 4 dB, Ro = 2 bit/s/Hz, K = L = 2, δT = 8 dB,
Similarly, we assume that the Rayleigh fading coefficients and μ1 = μ2 = 0.03. It is observed from Fig. 3 that as
of all CCI links are also with the same variance, which the outage probability decreases, intercept probabilities of

⎧ M

⎪ Δ1 Δ1

⎪ 1 − exp − , if δT ≥

⎪ ⎡
2
(1 − μ1 Δ1 ) γσuc ⎤ 1 − μ1 Δ1



⎪ Δ1

⎪ 1 − exp −
⎨⎢⎢
2
(1 − μ1 Δ1 ) γσuc ⎥

SECps
Pout = ⎢⎢ M  m−1 ⎥
⎥ (51)

⎪ ⎢ δT ⎥ , if δT < Δ1

⎪ ⎢ − 1 − exp − ⎥

⎪ ⎢ m=2 γσ 2
⎥ 1 − μ1 Δ1

⎪ ⎢
uc

⎪ ⎥

⎪ ⎣ Δ1 ⎦
⎩ × exp − 2
(1 − μ1 Δ1 ) γσuc

 M M−1 m 


βT  βT βT
Φ= 1 − exp − 2 + 1 − exp − 2 exp − 2 . (53)
γσuc m=0
γσuc γσuc



⎪ Δ1 Δ1
⎨M 2
, if δT ≥
SECps (1 − μ1 Δ1 ) γσuc 1 − μ1 Δ1
lim Pout = Δ1 1 Δ1 (55)
γ→∞ ⎪

⎩ − δT 2
, if δT <
(1 − μ1 Δ1 ) γσuc 1 − μ1 Δ1

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7019

Fig. 4. SRT of the RRS, SECps, and SC schemes for different HI levels with
M = 5, SNR ∈ [0, 20 dB], INR = 4 dB, Ro = 2 bit/s/Hz, K = L = 2,
δT = 8 dB, and MER = 3 dB. Fig. 5. SRT of the RRS, SECps, and SC schemes for different INRs with
M = 5, SNR ∈ [0, 20 dB], MER = 3 dB, Ro = 2 bit/s/Hz, K = L = 2,
δT = 8 dB, and μ1 = μ2 = 0.03.
the RRS, SECps, and SC schemes increase, and vice versa.
This indicates that the improvement of transmission security
comes at the cost of the decrease of reliability, which means
a trade-off between security and reliability. It is also shown
from Fig. 3 that the SECps and SC schemes always perform
better than the RRS scheme in terms of their SRT performance
for both cases of MER = 2 and 6 dB, where the SC scheme
achieves the best SRT performance under the joint impact of
HI and CCI. Moreover, as MER increases from 2 to 6 dB,
the SRT performance of the RRS, SECps, and SC schemes is
obviously improved.
Fig. 4 demonstrates the SRT of the RRS, SECps, and
SC schemes for different HI levels with M = 5, SNR ∈
[0, 20 dB], INR = 4 dB, Ro = 2 bit/s/Hz, K = L = 2, δT =
8 dB, and MER = 3 dB. It is shown from Fig. 4 that the SRT
performance of the SECps and SC schemes outperforms that of
RRS scheme. It needs to be pointed out that the SECps scheme Fig. 6. SRT of the RRS, SECps, and SC schemes for different values of M
has a close SRT performance to the SC scheme when the with INR = 4 dB, SNR ∈ [0, 20 dB], MER = 3 dB, Ro = 2 bit/s/Hz,
received SNR at CH is less than the given threshold. Moreover, K = L = 2, δT = 8 dB, and μ1 = μ2 = 0.03.
as HI increases from 0.03 to 0.09, the SRT performance of the
three presented schemes all decline slightly. improved significantly, whereas the RRS scheme achieves no
Fig. 5 depicts the SRT of the RRS, SECps, and SC improvement. For the SECps scheme, with an increase of the
schemes for different INRs with M = 5, SNR ∈ [0, 20 dB], number of users, when SNR is less than the threshold, the SRT
MER = 3 dB, Ro = 2 bit/s/Hz, K = L = 2, δT = 8 dB, and performance improvement is obvious. However, when SNR is
μ1 = μ2 = 0.03. We can observe from Fig. 5 that the SRT larger than the threshold, the SECps scheme has little or no
performance of the SECps and SC schemes is better than that SRT performance improvement. This is because that when
of the RRS scheme. In particular, when the received SNR SNR is large enough, after estimating the channels of the
is greater than the given threshold, the SC scheme performs first few users, the acceptable user can be found. As a result,
better than the SECps scheme in terms of the SRT performance increasing the number of users does not lead to significant
at the cost of estimating more wireless channels. Additionally, SRT performance improvement for the SECps scheme.
with an increase of INR, the SRT performance of the RRS, Fig. 7 depicts the SRT of the RRS, SECps, and SC schemes
SECps, and SC all decrease a bit. for different values of the average main and wiretap channel
Fig. 6 shows the SRT of the RRS, SECps, and SC schemes gains with INR = 4 dB, SNR ∈ [0, 20 dB], Ro = 2 bit/s/Hz,
for different number of users with INR = 4 dB, SNR ∈ K = L = 2, δT = 8 dB, and μ1 = μ2 = 0.03.
[0, 20 dB], MER = 3 dB, Ro = 2 bit/s/Hz, K = L = 2, As illustrated in Fig. 7, when the average gain of wiretap
δT = 8 dB, and μ1 = μ2 = 0.03. We can see from Fig. 6 that channel is set to 0.5, the SRT performance of the conventional
the SC scheme has the best SRT performance, followed by RRS, SECps, and SC schemes can be improved as the average
the SECps scheme, while the RRS scheme has the worst gain of main channel increases from 0.5 to 1. On the contrary,
SRT performance. Moreover, as the number of users increases when the average gain of main channel is set to 0.5, as the
from 5 to 10, the SRT performance of SC scheme can be average gain of wiretap channel increases from 0.1 to 0.5,

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7020 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

Fig. 7. SRT of the RRS, SECps, and SC schemes for different values of the Fig. 9. Sum of outage and intercept probability for the RRS, SECps, and
average main and wiretap channel gains with INR = 4 dB, SNR ∈ [0, 20 dB], SC schemes versus SNR for different MERs with INR = 4 dB, M = 15,
Ro = 2 bit/s/Hz, K = L = 2, δT = 8 dB, and μ1 = μ2 = 0.03. Ro = 2 bit/s/Hz, K = L = 2, δT = 8 dB, and μ1 = μ2 = 0.03.

Fig. 8. Sum of outage and intercept probability for the RRS, SECps, and Fig. 10. Sum of outage and intercept probability for the RRS, SECps, and
SC schemes versus δT with SNR = 10 dB, M = 5, Ro = 2 bit/s/Hz, SC schemes versus Ro for different HIs with SNR = 10 dB, INR = 4 dB,
K = L = 2, INR = 4 dB, MER = 8 dB, and μ1 = μ2 = 0.03. M = 15, K = L = 2, δT = 5 dB, and MER = 8 dB.

the SRT performance of the conventional RRS, SECps, and in Fig. 9, all curves of the RRS, SECps, and SC schemes
SC schemes all degrade. regarding the sum of outage and intercept probability are
Fig. 8 illustrates the Sum of outage and intercept probability concave curves, and there is a minimum value on each of
for the RRS, SECps, and SC schemes versus δT with SNR = these curves, indicating that better SRT performance of the
10 dB, M = 5, Ro = 2 bit/s/Hz, K = L = 2, INR = 4 dB, three presented multiuser scheduling schemes can be achieved
MER = 8 dB, and μ1 = μ2 = 0.03. From Fig. 8, we can by adjusting SNR. Moreover, the sum of outage and intercept
see that the sum of outage and intercept probability for the probability for the RRS, SECps, and SC schemes significantly
SECps scheme is close to that for the RRS scheme when decreases as MER increases from 4 to 8 dB.
the threshold is small. With an increase of δT , the sum Fig. 10 depicts the sum of outage and intercept probability
of outage and intercept probability for the SECps scheme for the RRS, SECps, and SC schemes versus Ro for different
constantly decreases until it converges to a floor which is the HIs with SNR = 10 dB, INR = 4 dB, M = 15, K = L = 2,
sum of outage and intercept probability for the SC scheme, δT = 5 dB, and MER = 8 dB. One can observe from
as expected. In addition, it is worth noting that when the given Fig. 10, with an increase of Ro , the sum of outage and
threshold is much higher than the actual received SNR at CH, intercept probability for the RRS, SECps, and SC schemes
the SECps scheme will gradually degenerate into SC scheme. goes down to its lowest and up again. This means that we
Fig. 9 shows the sum of outage and intercept probability for can adjust Ro to get better SRT performance of the RRS,
the RRS, SECps, and SC schemes versus SNR for different SECps, SC schemes. Additionally, as the HI level increases
MERs with INR = 4 dB, M = 15, Ro = 2 bit/s/Hz, from 0.03 to 0.09, the sum of outage and intercept probability
K = L = 2, δT = 8 dB, and μ1 = μ2 = 0.03. As observed for the RRS, SECps, and SC schemes declines in low Ro

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7021

TABLE I
P ROPOSED G OLDEN S EARCH I TERATIVE M ETHOD FOR A CHIEVING THE O PTIMAL SNR

Fig. 11. SRT of the RRS, SECps, and SC schemes with different HI levels
for different devices when M = 5, SNR = 10 dB, INR = 4 dB, Ro = Fig. 12. Sum of outage and intercept probability for the RRS, SECps, and
2 bit/s/Hz, K = L = 2, δT = 8 dB, and MER = 8 dB. SC schemes versus M for different number of interferers with SNR = 10
dB, Ro = 2 bit/s/Hz, INR = 4 dB, δT = 5 dB, MER = 8 dB, and
μ1 = μ2 = 0.03.

region, and increases in high Ro region. This is because the


increase of HI level will have an impact on the capacity of of eavesdropper μ2 is set to 0.08. On the contrary, when the
both main and wiretap channels. However, the degree of this HI level of CH μ1 is set to 0.08, the outage and intercept
impact on security and reliability is different under different probability for the three presented schemes all decrease.
system parameters. Fig. 12 illustrates the sum of outage and intercept prob-
As shown in Figs. 9 and 10, we can see that each curve of ability for the RRS, SECps, and SC schemes versus M for
the presented multiuser scheduling schemes has a minimum different number of interferers with SNR = 10 dB, Ro = 2
value, which means that there is an optimal SNR or Ro for bit/s/Hz, INR = 4 dB, δT = 5 dB, MER = 8 dB, and μ1 =
the proposed multiuser scheduling schemes to achieve the best μ2 = 0.03. As seen from Fig. 12, with an increased number of
SRT performance. As observed in (14), (18), (24), (28), (35), users, the sum of outage and intercept probability for the SC
and (38), it is very difficult to obtain a closed-form solution of scheme decreases obviously, followed by the SECps scheme,
the optimal SNR and Ro . To this end, we propose an iterative while that of the RRS scheme does not change at all. This
algorithm based on the golden search to determine the optimal further demonstrates the benefits of employing the SECps and
SNR and Ro of the proposed SECps, SC, and RRS schemes, SC schemes. Moreover, as the number of interferers increases
which is summarized in Table I. Table I shows the process from 2 to 8, the sum of outage and intercept probability for
of iterative algorithm for seeking the best SNR, which can the RRS and SECps schemes increases, while that for the
be similarly applied to find the optimal Ro . As seen from SC scheme does not increase but decline when M is greater
Figs. 9 and 10, the red hexagram on each curve is the than or equal to 4. This is due to the fact that an increase in
optimal SNR or Ro obtained by the iterative algorithm for the the number of interferers has different effect on security and
proposed SECps, SC, and RRS schemes. reliability of the system.
Fig.11 shows the SRT of the RRS, SECps, and SC schemes
with different HI levels for different devices when M = 5, VI. C ONCLUSION
SNR = 10 dB, INR = 4 dB, Ro = 2 bit/s/Hz, K = L = 2, In this paper, we have examined the security and reliability
δT = 8 dB, and MER = 8 dB. We can see from Fig. 11 that the trade-off for a wireless sensor network under the joint impact
sum of outage and intercept probability for the conventional of hardware impairment and co-channel interference, where a
RRS, SECps, and SC schemes all increase when the HI level sensor user is selected to send its confidential message to CH

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7022 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

in the face of an eavesdropper. We presented the SECps and (B.1) can be expanded as
SC schemes to enhance transmission security and reliability M
  
of the system. Conventional RRS scheme was considered 1 − e−λum c α
for comparison purposes. Specifically, in the SECps scheme, m=1
a user whose SNR is greater than a given threshold would M
2 −1
 
|D w |

be selected to communicate with CH rather than the best = 1+ (−1) exp − λum c α , (B.2)
one, which can reduce the complexity of wireless channel w=1 m∈D w
estimations. We derived closed-form expressions of outage where Dw stands for the w-th non-empty subset of U and
and intercept probability and carried out secrecy diversity |Dw | denotes the set cardinality of Dw .
analysis for the RRS, SECps, and SC schemes over Rayleigh When μ1 < Δ1 −1 , the equation (22) can be further obtained
fading channels. Moreover, numerical results illustrated that as
the SECps and SC schemes outperform the RRS scheme in  ∞
terms of their SRT performance. It was also shown that a
SC
Pout,I = FX2 (α)fY1 (α) dα. (B.3)
1
trade-off between security and reliability can be achieved by
Substituting (B.2) and (A.2) into (B.3), we have
adjusting the overall data rate and SNR for each of the RRS, ⎡  ⎤
SECps, and SC schemes.  ∞ M
2 −1 
w
SC
Pout,I = ⎣1 + (−1)|D | exp − λum c α ⎦
1 w=1 m∈D w
A PPENDIX A
D ERIVATION OF (13) λK
ic K−1
× (α − 1) e−λic (α−1) dα. (B.4)
Denoting X1 = 1−μ1 Δ1
ψ |hum c |2
and Y1 = (K − 1)!
Δ1

K Similar to the derivation of (13), we can arrive at (23) from
(1 + μ1 ) ψ  |gik |2 + 1, the CDF of X1 is given by (B.4) through integral operation.
k=1

FX1 (α) = 1 − exp (−λum c α) , (A.1) A PPENDIX C


D ERIVATION OF (31)
where λum c = (1−μ1 ΔΔ11)ψσ2 . Substituting (29) and (A.2) into (30), we have (C.1), shown
um c
The assumption that the average channel gains of interfer- at the top of next page, where ξ = δT (1−μ Δ1
1 Δ1 )
− 1. The
ers are i.i.d. zero-mean complex Gaussian random variables, CDF of δm is formulated as
leading to the fact that the sum of the squares of the aver- 
x
age channel gain of multiple interference links constitutes a Fδm (x) = 1 − exp − 2 . (C.2)
chi-square distribution. Hence, following [29], the probability ψσum c
density function (PDF) of Y1 can be obtained as !M  
Δ1
Thus, the terms of Fδm 1−μ 1 Δ1
(x + 1) and
λK
ic K−1 −λic (α−1) m=1
fY1 (α) = (α − 1) e (A.2) !
m−1
(K − 1)! Fδq (δT ) can be expanded as
q=1
1
for α > 1, where λic = 2 .
(1+μ1 )ψ  σic M 
 Δ1
When μ1 < Δ1 −1 , the equation (12) can be further obtained Fδm (x + 1)
as m=1
1 − μ1 Δ1
 ∞ M  
2 −1 
RRS
Pout,m = FX1 (α)fY1 (α) dα. (A.3) = 1+ (−1)
|D w |
exp − λum c (x + 1) ,
1
w=1 m∈D w
Substituting (A.1) and (A.2) into (A.3), we have (C.3)
 ∞
  λKic K−1
RRS
Pout,m = 1 − e−λum c α (α − 1) and
⎛ ⎞
1 (K − 1)! m−1 2m−1
 −1 |D | v  δT ⎠
× e−λic (α−1) dα. (A.4) Fδq (δT ) = 1+ (−1) exp ⎝− ,
q=1 v=1
ψσu2 q c
q∈D v
Through variable substitution and using the formulas
∞ v−1 −ux (C.4)
0
x e = u1v Γ (v) [42, eq. (3.381.4)] and Γ (v) =
(v − 1)! [42, eq. (8.339.1)], we can arrive at (13) from (A.4). where D is the v-th subset of {U1 , . . . , Um−1 } and |Dv |
v

represents the set cardinality of Dv .


A PPENDIX B Substituting (C.2) and (C.3) into Ξ1 yields
D ERIVATION OF (23) Ξ1 ⎛
1−μ1 Δ1 2   ⎞
Denoting X2 = Δ1 ψmax |hum c | , the CDF of X2 is ξ
M
2 −1 
m∈U ⎝1 + |D w |
given by = (−1) exp − λum c (x + 1) ⎠
0 w=1 m∈D w
M
   λK
FX2 (α) = 1 − e−λum c α . (B.1) × ic
xK−1 exp (−λic x) dx (C.5)
m=1 (K − 1)!

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7023

⎧ ⎡ ⎡ M ⎤ ⎤

⎪  Δ1

⎪ ⎢  ξ⎢ Fδm (x + 1) ⎥ ⎥

⎪ ⎢ ⎢ m=1 1 − μ1 Δ1 ⎥ ⎥

⎪ ⎢ ⎢ ⎥ ⎥

⎪ ⎢ ⎣ ⎦ ⎥

⎪ ⎢ 0 K
λic ⎥

⎪ ⎢ × xK−1 −λic x
e dx ⎥

⎪ ⎢ (K − 1)! ⎥

⎪ ⎢ " #$ % ⎥

⎪ ⎢ ⎥

⎪ ⎢ Ξ1 ⎥

⎪ ⎢ ⎡  ⎤ ⎥

⎪ ⎢ Δ1 ⎥

⎪ ⎢ ⎥


⎪ ⎢ ⎢ Fδ1 1 − μ1 Δ1 (x + 1) ⎥ ⎥ Δ1

⎪ ⎢ ⎢ ⎥ ⎥ , if δT ≥

⎪ ⎢  ∞⎢ M
  m−1 ⎥ ⎥ 1 − μ1 Δ1

⎪ ⎢ ⎢ ⎥ ⎥

⎪ ⎢ + ⎢− F (δ ) ⎥ ⎥

⎪ ⎢ ⎢ δq T ⎥ ⎥

⎪ ⎢ ξ ⎢ m=2 q=1 ⎥ ⎥

⎪ ⎢ ⎢ ⎥ ⎥

⎪ ⎢ ⎣ Δ1 ⎦ ⎥

⎪ ⎢ × 1 − Fδm (x + 1) ⎥

⎪ ⎢ 1 − μ1 Δ1 ⎥

⎨ ⎢ ⎥
⎢ ⎥
SECps
= ⎢ λK
ic ⎥
Pout,I
⎪ ⎢ × xK−1 e−λic x dx ⎥ (C.1)

⎪ ⎣ (K − 1)! ⎦

⎪ " #$ %



⎪ ⎡ Ξ2 ⎤



⎪ 

⎪ ⎢ ⎡ ⎤ ⎥

⎪ ⎢ Δ1 ⎥





⎢ ⎢ Fδ1 1 − μ1 Δ1 (x + 1) ⎥ ⎥


⎪ ⎢ ⎢ ⎥ ⎥

⎪ ⎢  ∞⎢⎢ M m−1
 ⎥ ⎥

⎪ ⎢ ⎥ ⎥

⎪ ⎢ ⎢ − F (δ ) ⎥ ⎥

⎪ ⎢ 0 ⎢ ⎢ δq T ⎥ ⎥ Δ1

⎪ ⎢ m=2 q=1 ⎥ ⎥ , if δT <

⎪ ⎢ ⎢ ⎥ ⎥

⎪ ⎢ ⎣ Δ ⎦ ⎥ 1 − μ1 Δ1

⎪ ⎢ × 1 − Fδm
1
(x + 1) ⎥

⎪ ⎢ 1 − μ1 Δ1 ⎥

⎪ ⎢ ⎥

⎪ ⎢ K ⎥

⎪ ⎢ λ ic ⎥

⎪ ⎢× xK−1 e−λic x dx ⎥

⎪ ⎣ (K − 1)! ⎦

⎩ " #$ %
Ξ3

⎛ ⎛ ⎛ ⎞⎞ ⎞
 ∞ M
 2m−1
−1 
⎝1− ⎝1 + |D v | δT ⎠⎠
Ξ2 = (−1) exp ⎝− exp (−λum c (x + 1)) − exp (−λu1 c (x + 1))⎠
ξ m=2 v=1
ψσu2 q c
q∈D v

λK
ic
× xK−1 exp (−λic x) dx. (C.6)
(K − 1)!

⎛ ⎛ ⎛ ⎞⎞ ⎞
 ∞ M
 2m−1
−1 
⎝1− ⎝1 + |D |v δT ⎠⎠
Ξ3 = (−1) exp ⎝− exp (−λum c (x + 1)) − exp (−λu1 c (x + 1))⎠
0 m=2 v=1
ψσu2 q c
q∈D v

λK
ic
× xK−1 exp (−λic x) dx (C.7)
(K − 1)!

μ n
By using the formula 0 xm e−βx dx = γ(n,βμ )
n

nβ v v = R EFERENCES
m+1
n [42, eq.3.381.8], (C.5) can be further obtained as (32).
[1] Y. Liu, M. Dong, K. Ota, and A. Liu, “ActiveTrust: Secure and trustable
Substituting (C.2) and (C.4) into Ξ2 yields (C.6), shown at routing in wireless sensor networks,” IEEE Trans. Inf. Forensics Security,
the top of the page. n
vol. 11, no. 9, pp. 2013–2027, Sep. 2016.

By employing the formula μ xm e−βx = Γ(v,βμ )
n

nβ v v= [2] D. He, C. Chen, S. Chan, J. Bu, and L. T. Yang, “Security analysis and
m+1 improvement of a secure and distributed reprogramming protocol for
n [42, eq.3.381.9], we can arrive at (33) from (C.6). wireless sensor networks,” IEEE Trans. Ind. Electron., vol. 60, no. 11,
Substituting (C.2) and (C.4) into Ξ3 yields (C.7), shown at pp. 5348–5354, Nov. 2013.

the top of the page. By utilizing the formulas 0 xv−1 e−ux = [3] J. Duan, D. Gao, D. Yang, C. H. Foh, and H.-H. Chen,
1 “An energy-aware trust derivation scheme with game theo-
uv Γ (v) [42, eq. (3.381.4)] and Γ (v) = (v − 1)! retic approach in wireless sensor networks for IoT applica-
[42, eq. (8.339.1)] again, Ξ3 can be further obtained tions,” IEEE Internet Things J., vol. 1, no. 1, pp. 58–69,
as (34). Feb. 2014.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
7024 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 20, NO. 11, NOVEMBER 2021

[4] A. Rasheed and R. N. Mahapatra, “The three-tier security scheme [26] J. Zhu, D. W. K. Ng, N. Wang, R. Schober, and V. K. Bhargava,
in wireless sensor networks with mobile sinks,” IEEE Trans. Parallel “Analysis and design of secure massive MIMO systems in the presence
Distrib. Syst., vol. 23, no. 5, pp. 958–965, May 2012. of hardware impairments,” IEEE Trans. Wireless Commun., vol. 16,
[5] J. Shen, T. Zhou, J. Lai, P. Li, and S. Moh, “Secure and efficient data no. 3, pp. 2001–2016, Mar. 2017.
sharing in dynamic vehicular networks,” IEEE Internet Things J., vol. 7, [27] E. Bjornson, M. Matthaiou, and M. Debbah, “A new look at dual-hop
no. 9, pp. 8208–8217, Sep. 2020. relaying: Performance limits with hardware impairments,” IEEE Trans.
[6] I. Tomic and J. A. McCann, “A survey of potential security issues in Commun., vol. 61, no. 11, pp. 4512–4525, Nov. 2013.
existing wireless sensor network protocols,” IEEE Internet Things J., [28] S. Javed, O. Amin, S. S. Ikki, and M.-S. Alouini, “Multiple antenna
vol. 4, no. 6, pp. 1910–1923, Dec. 2017. systems with hardware impairments: New performance limits,” IEEE
[7] S. Leung-Yan-Cheong and M. Hellman, “The Gaussian wire-tap chan- Trans. Veh. Technol., vol. 68, no. 2, pp. 1593–1606, Feb. 2019.
nel,” IEEE Trans. Inf. Theory, vol. IT-24, no. 4, pp. 451–456, Jul. 1978. [29] K. Guo, M. Lin, B. Zhang, W.-P. Zhu, J.-B. Wang, and T. A. Tsiftsis, “On
[8] Y. Zou, J. Zhu, X. Wang, and L. Hanzo, “A survey on wireless security: the performance of LMS communication with hardware impairments
Technical challenges, recent advances, and future trends,” Proc. IEEE, and interference,” IEEE Trans. Commun., vol. 67, no. 2, pp. 1490–1505,
vol. 104, no. 9, pp. 1727–1765, Sep. 2016. Feb. 2019.
[9] L. Hu et al., “Cooperative jamming for physical layer security enhance- [30] L. Yu et al., “Spectrum availability prediction for cognitive radio
ment in Internet of Things,” IEEE Internet Things J., vol. 5, no. 1, communications: A DCG approach,” IEEE Trans. Cognit. Commun.
pp. 219–228, Feb. 2018. Netw., vol. 6, no. 2, pp. 476–485, Jun. 2020.
[10] B. Li, Y. Zou, J. Zhou, F. Wang, W. Cao, and Y.-D. Yao, “Secrecy [31] B. Li, J. Zhou, Y. Zou, and F. Wang, “Closed-form secrecy outage
outage probability analysis of friendly jammer selection aided multiuser analysis of cellular downlink systems in the presence of co-channel inter-
scheduling for wireless networks,” IEEE Trans. Commun., vol. 67, no. 5, ference,” IEEE Trans. Wireless Commun., vol. 17, no. 7, pp. 4721–4734,
pp. 3482–3495, May 2019. Jul. 2018.
[11] Y. Zou, X. Wang, W. Shen, and L. Hanzo, “Security versus reliability [32] C. Luo, J. Ji, Q. Wang, X. Chen, and P. Li, “Channel state information
analysis of opportunistic relaying,” IEEE Trans. Veh. Technol., vol. 63, prediction for 5G wireless communications: A deep learning approach,”
no. 6, pp. 2653–2661, Jul. 2014. IEEE Trans. Netw. Sci. Eng., vol. 7, no. 1, pp. 227–236, Jan. 2020.
[12] K. Lee, J.-P. Hong, H.-H. Choi, and M. Levorato, “Adaptive wireless- [33] B. Li, J. Zhou, Y. Zou, F. Wang, and W. Cao, “Secrecy versus
powered relaying schemes with cooperative jamming for two-hop secure computation overhead for wireless networks in the face of co-channel
communication,” IEEE Internet Things J., vol. 5, no. 4, pp. 2793–2803, interference,” IEEE Trans. Commun., vol. 67, no. 3, pp. 2306–2320,
Aug. 2018. Mar. 2019.
[13] Y. Zou, J. Zhu, X. Li, and L. Hanzo, “Relay selection for wireless [34] M. Ju and K.-S. Hwang, “Outage equivalence of opportunistic relaying
communications against eavesdropping: A security-reliability trade-off and selection cooperation in presence of co-channel interference,” IEEE
perspective,” IEEE Netw., vol. 30, no. 5, pp. 74–79, Sep. 2016. Trans. Wireless Commun., vol. 14, no. 6, pp. 2981–2991, Jun. 2015.
[35] J. A. Hussein, S. Boussakta, and S. S. Ikki, “Performance study of a
[14] L. Dai, B. Wang, Y. Yuan, S. Han, I. Chih-lin, and Z. Wang, “Non-
UCRN over Nakagami-m fading channels in the presence of CCI,” IEEE
orthogonal multiple access for 5G: Solutions, challenges, opportunities,
Trans. Cogn. Commun. Netw., vol. 3, no. 4, pp. 752–765, Dec. 2017.
and future research trends,” IEEE Commun. Mag., vol. 53, no. 9,
[36] S. W. Kim, Y. J. Chun, and S. Kim, “Co-channel interference cancel-
pp. 74–81, Sep. 2015.
lation using single radio frequency and baseband chain,” IEEE Trans.
[15] B. Di, L. Song, and Y. Li, “Sub-channel assignment, power allo-
Commun., vol. 58, no. 7, pp. 2169–2175, Jul. 2010.
cation, and user scheduling for non-orthogonal multiple access net-
[37] A. Afana and S. Ikki, “Analytical framework for space shift keying
works,” IEEE Trans. Wireless Commun., vol. 15, no. 11, pp. 7686–7698,
MIMO systems with hardware impairments and co-channel interfer-
Nov. 2016.
ence,” IEEE Commun. Lett., vol. 21, no. 3, pp. 488–491, Mar. 2017.
[16] J.-B. Kim and I.-H. Lee, “Non-orthogonal multiple access in coordinated [38] T. T. Duy, T. Q. Duong, D. B. da Costa, V. N. Q. Bao, and M. Elkashlan,
direct and relay transmission,” IEEE Commun. Lett., vol. 19, no. 11, “Proactive relay selection with joint impact of hardware impairment
pp. 2037–2040, Nov. 2015. and co-channel interference,” IEEE Trans. Commun., vol. 63, no. 5,
[17] D. B. Rawat, T. White, M. S. Parwez, C. Bajracharya, and M. pp. 1594–1606, May 2015.
Song, “Evaluating secrecy outage of physical layer security in [39] J. A. Hussein, S. S. Ikki, S. Boussakta, and C. C. Tsimenidis, “Per-
large-scale MIMO wireless communications for cyber-physical sys- formance analysis of opportunistic scheduling in dual-hop multiuser
tems,” IEEE Internet Things J., vol. 4, no. 6, pp. 1987–1993, underlay cognitive network in the presence of cochannel interference,”
Dec. 2017. IEEE Trans. Veh. Technol., vol. 65, no. 10, pp. 8163–8176, Oct. 2016.
[18] D. Kapetanovic, G. Zheng, and F. Rusek, “Physical layer security [40] H. Yu, I.-H. Lee, and G. L. Stuber, “Outage probability of decode-
for massive MIMO: An overview on passive eavesdropping and and-forward cooperative relaying systems with co-channel interference,”
active attacks,” IEEE Commun. Mag., vol. 53, no. 6, pp. 21–27, IEEE Trans. Wireless Commun., vol. 11, no. 1, pp. 266–274, Jan. 2012.
Jun. 2015. [41] Y. Zou, “Intelligent interference exploitation for heterogeneous cellular
[19] A. Mukherjee, S. A. A. Fakoorian, J. Huang, and A. L. Swindlehurst, networks against eavesdropping,” IEEE J. Sel. Areas Commun., vol. 36,
“Principles of physical layer security in multiuser wireless networks: no. 7, pp. 1453–1464, Jul. 2018.
A survey,” IEEE Commun. Surveys Tuts., vol. 16, no. 3, pp. 1550–1573, [42] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
3rd Quart., 2014. Products, 7th ed. San Diego, CA, USA: Academic, 2007.
[20] Y. Zou, X. Wang, and W. Shen, “Physical-layer security with multiuser
scheduling in cognitive radio networks,” IEEE Trans. Commun., vol. 61,
no. 12, pp. 5103–5113, Dec. 2013.
[21] H. Deng, H.-M. Wang, W. Wang, and M. H. Lee, “Dual user selection
for security enhancement in uplink multiuser systems,” IEEE Commun.
Lett., vol. 20, no. 9, pp. 1900–1903, Sep. 2016.
[22] M. Yang, D. Guo, Y. Huang, T. Q. Duong, and B. Zhang, “Physi-
cal layer security with threshold-based multiuser scheduling in multi-
antenna wireless networks,” IEEE Trans. Commun., vol. 64, no. 12,
pp. 5189–5202, Dec. 2016.
[23] L. Bai, L. Zhu, T. Li, J. Choi, and W. Zhuang, “An efficient hybrid Bin Li received the B.E. degree in power engi-
transmission method: Using nonorthogonal multiple access and multi- neering from Southeast University (SEU), Nanjing,
user diversity,” IEEE Trans. Veh. Technol., vol. 67, no. 3, pp. 2276–2288, China, in 2004, the M.S. degree in information
Mar. 2018. and telecommunications engineering from the Nan-
[24] M. Mokhtar, A.-A.-A. Boulogeorgos, G. K. Karagiannidis, and jing University of Posts and Telecommunications
N. Al-Dhahir, “OFDM opportunistic relaying under joint trans- (NUPT), Nanjing, in 2009, and the Ph.D. degree
mit/receive I/Q imbalance,” IEEE Trans. Commun., vol. 62, no. 5, in communication and information system from the
pp. 1458–1468, May 2014. Nanjing University of Aeronautics and Astronau-
[25] E. Costa and S. Pupolin, “M-QAM-OFDM system performance in tics (NUAA), Nanjing, in 2019. He is currently a
the presence of a nonlinear amplifier and phase noise,” IEEE Trans. Lecturer with NUPT. His research interests include
Commun., vol. 50, no. 3, pp. 462–472, Mar. 2002. wireless security and cooperative communications.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.
LI et al.: IMPACT OF HI AND CCI ON SRT FOR WSNs 7025

Yulong Zou (Senior Member, IEEE) received the Weifeng Cao received the B.E. degree in informa-
B.Eng. degree in information engineering from tion engineering from Soochow University, Suzhou,
the Nanjing University of Posts and Telecommu- China, in 2004, and the M.S. degree in pattern
nications (NUPT), Nanjing, China, in July 2006, recognition from the Nanjing University of Posts
the Ph.D. degree in electrical engineering from and Telecommunications, Nanjing, China, in 2008,
the Stevens Institute of Technology, NJ, USA, where she is currently pursuing the Ph.D. degree.
in May 2012, and the Ph.D. degree in signal and Her research interests include satellite communi-
information processing from NUPT in July 2012. cations, cooperative communications, and physical
He is currently a Full Professor and a Doctoral layer security.
Supervisor with NUPT. He has acted as a TPC
Member for various IEEE sponsored conferences,
such as IEEE ICC, IEEE GLOBECOM, IEEE WCNC, IEEE VTC, and IEEE
ICCC. He was awarded the Ninth IEEE Communications Society Asia–Pacific
Best Young Researcher in 2014 and a co-receipt of the Best Paper Award at
the 80th IEEE Vehicular Technology Conference in 2014. He has served as an
Editor for the IEEE C OMMUNICATIONS S URVEYS AND T UTORIALS , IEEE
C OMMUNICATIONS L ETTERS , EURASIP Journal on Advances in Signal
Processing, IET Communications, and China Communications.

Jia Zhu received the B.E. degree in computer


science and technology from Hohai University, Nan-
jing, China, in July 2005, and the Ph.D. degree in
signal and information processing from the Nan-
jing University of Posts and Telecommunications
(NUPT), Nanjing, in April 2010. From June 2010 to
June 2012, she was a Post-Doctoral Research Fellow
with the Stevens Institute of Technology, NJ, USA.
Since November 2012, she has been a full-time
Faculty Member with the School of Telecommuni-
cation and Information, NUPT, where she is cur-
rently a Professor. Her general research interests include the cognitive radio,
physical-layer security, and communications theory.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 08,2022 at 11:32:21 UTC from IEEE Xplore. Restrictions apply.

You might also like