Professional Documents
Culture Documents
Physical-Layer Security of SIMO Communication
Physical-Layer Security of SIMO Communication
Physical-Layer Security of SIMO Communication
Abstract—The present work investigates the physical layer security of wireless communication systems over non-homogeneous
fading environments, i.e., h-m and -m fading models, which are typically encountered in realistic wireless transmission scenarios in the
context of conventional and emerging communication systems. This study considers a single-input multiple-output system that consists
of a single-antenna transmitter, a multi-antenna legitimate receiver, and an active multi-antenna eavesdropper. To this end, novel exact
analytical expressions are derived for the corresponding average secrecy capacity and secrecy outage probability, which are
corroborated by respective results from computer simulations. Capitalizing on the offered results, the physical layer security is
quantified in terms of different parameters, which leads to useful insights on the impact of non-homogeneous fading environment and
the number of employed antennas on the achieved physical layer security levels of the underlying system configuration. The offered
results and insights are useful for the design of such systems as well as for the computational requirements and sustainability relating
to such systems, since emerging communications are largely characterized by stringent quality of service and complexity
requirements.
Index Terms—Multipath fading, secrecy capacity, secrecy outage probability, physical layer security, multi-antenna communications,
computational complexity
1 INTRODUCTION
2377-3782 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See ht_tps://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
106 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 107
fg ðxÞ ¼ pffiffiffi 1
secrecy capacity CS averaged over the instantaneous SNRs
GðmÞ hðh 1Þm2 g mþ2
1
2 (1)
ð1þhÞ2 mx ðh 1Þmx 1. Henceforth, the subscript B is associated with Bob’s measures,
e 2hg Im1 ; while the subscript E refers to Eve’s measures.
2 2hg 2. The subscript “inc” is used in order to differentiate the instanta-
neous SNR g i from the lower incomplete Gamma function g inc ð; Þ.
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
108 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 109
!LB mB
2p ðhB þ 1Þ2
J h1 ¼
lnð2ÞGðLB mB ÞGðLE mE Þ 4hB
2k 2j !LE mE þ2j
1 X
X 1 2L m þ2k
mB B B ð2LE mE þ 2j 1Þ! 1 h2B 1 h2E 4hE
k¼0 j¼0
k!j!22LE mE þ2j1 GðLB mB þ k þ 0:5ÞGðLE mE þ j þ 0:5Þ 4hB 4hE ðhE þ 1Þ2 (16)
" 2LBX 2LE mX
mB þ2k E þ2j1 2LB mX
B þ2kþn
1
ð2LB mB þ 2k 1Þ!eAB Am
B Gðm 2LB mB 2k; AB Þ
m¼1 n¼0 i¼1
n!
Gð2LB mB þ 2k þ nÞAnE ðAB þ AE Þi eAB þAE Gð2LB mB 2k n þ i; AB þ AE Þ :
terms, which yields a relative error due to truncation [27] of To this effect and after some long but basic algebraic
less than 0.002 percent. Also, the algebraic representation of manipulations, one obtains (21), at the top of the next page
the derived series is tractable since it consists of well-known (see p. 6).Likewise, an expression for J 2 can be deduced from
elementary and special functions, which render it conve- (21) after replacing LB by LE , mB by mE , HB by HE , hB by hE ,
nient to handle both analytically and numerically. and gB by gE , and vice-versa, whereas J 3 can be expressed as
pffiffiffi
3.2 Exact ASC over -m Fading Channels 2 p
J 3 ¼
Capitalizing on the derived analytical expression for the ASC lnð2ÞGðLE mE Þð1 2 ÞLE mE
over h-m fading channels, we can also derive a similar
X
1 2LE mE þ2k
expression for the ASC for the case of -m fading channels. E mE
2k Gð2LE mE þ 2kÞ
2 2km GðL m þ k þ 0:5Þ
To this end, by substituting (4) and (6) into (8) and following k¼0 k!ð1 E Þ E E
the same process as in the previous section, it follows that
2mE
2p mE þ2k G m 2LE mE 2k; g ð12 Þ
2LEX
J 1 ¼ E E
lnð2Þð1 2B ÞLB mB GðLB mB ÞGðLE mE Þ 2m : (22)
2L m þ2km gE ð12 Þ
E
X1 X 1
2k 2j 2 LE mE
B E ð1 E Þ
m¼1
g E E E e E
2 2k 2LB mB þ2k
k¼0 j¼0 k!j!ð1 B Þ g B
2L m þ2k
212LE mE 2j mB B B Therefore, by substituting (21) and (22) into (8) and after
(18)
GðLB mB þ k þ 0:5ÞGðLE mE þ 2j þ 0:5Þ some algebraic manipulations, the corresponding analytical
Z 1 2mB g B expression for the CS over m fading channels is deduced.
2L m þ2k1 gB ð12 Þ
lnð1 þ g B Þg B B B e B
0
3.3 Upper Bounds for the Truncation Errors
2mE g B As already mentioned, the derived expression for the ASC is
g inc 2LE mE þ 2j; dg B ;
ð1 2E Þ
gE given in terms of infinite series. Depending on the value of
and
(Z the involved parameters, this series requires different num-
1 2mB g B
lnð1 þ g B Þ
g B ð12B Þ ber of terms to ensure acceptable truncation that leads to
I 1 ¼ Gð2LE mE þ 2jÞ 12L m 2k
e dg B
0 gB B B accurate results. Unlike error rate measures that typically
2LE mX
E þ2j1
n require several decimal digits accuracy, the derived ASC
1 2mE
(19) measure only requires, at most, a two-decimal digit accu-
n! g E ð1 E Þ
2
n¼0
) racy. Nevertheless, even though the involved series achieves
Z 1 2mB g B 2mE g B
lnð1 þ g B Þ
g B ð12 Þ g E ð12 Þ
a sufficient accuracy for a relatively low number of terms,
12LB mB 2kn
e B E dg B ; deriving a tight closed-form upper bound for the exact trun-
0 gB
cation error of this series will allow the determination of the
which yields exact accuracy for a specific number of terms at given scenar-
" ios. Based on this, we derive a tight and tractable closed-
2mB
form upper bound for the truncation error of the derived
¼ Gð2LE mE þ 2jÞ ð2LB mB þ 2k 1Þ!egB ð1B Þ
2
I 1
series representation in (8). In this case, the truncation error
mB þ2k
2LBX m of (8) can be bounded by deriving closed-form bounds for
g B ð1 2B Þ J h1 , J h2 and J h3 , for the case of h-m fading.
m¼1
2mB To this end, the truncation of J h1 after p 1 terms results to
2mB the truncation error in (23), at the top of the next page, i.e.,
G m 2LB mB 2k;
gB ð1 2B Þ p. 6. It is evident that (23) can be upper bounded by (24),
2mB (20) shown at the top of the next page (see p. 6).To this effect, we
E þ2j1 2LB mX
2LE mX B þ2kþn
Gð2LB mB þ 2k þ nÞ egB ð1B Þ
2
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
110 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
1 X
X 1 2j 2k
m2k
B ð2LE mE þ 2j 1Þ!ð1 hE Þ 1 h2B
J h1 ¼
k¼p j¼p k!j!22j GðLB mB þ k þ 12ÞGðLE mE þ j þ 12Þð1 þ hE Þ2j 4hB
" 2LBX
mB þ2k
Gð2LB mB 2k þ m; AB Þ
ð2LB mB þ 2k 1Þ!eAB (23)
m¼1
Am
B
2LE mX
#
E þ2j1 2LB mX
B þ2kþn
Gð2LB mB þ 2k þ nÞAnE
Gð2LB mB 2k n þ i; AB þ AE Þ :
n¼0 i¼1 n!ðAB þ AE Þi eAB AE
X
1 X 1
2j
B mB
2k E Gð2LE mE þ 2jÞ
2k
J 1 < 2k 2j GðL m þ j þ 1Þ
k¼0 k!ð1 B Þ GðLB mB þ k þ 2Þ j¼0
1 j!2 E E 2
" 2mB 2LBXmB þ2p m
g B ð1 B Þ 2
2mB
Gð2LB mB þ 2pÞegB ð1B Þ
2
G m 2LB mB 2p;
m¼1
2mB g B ð1 2B Þ
n 2mB (24)
2LE mXE þ2j1 2LB mX
B þ2pþn 2mE
Gð2LB mB þ 2p þ nÞ 2mE þ
gB ð12 Þ gE ð12 Þ
e B E
n¼0 i¼1
n! gE ð1 2E Þ
i #
2mB 2mE 2mB 2mE
þ G 2LB mB 2p n þ i; þ :
gB ð1 2B Þ g E ð1 2E Þ gB ð1 2B Þ gE ð1 2E Þ
n o
0 8
91
o 1 > >
L m 2 2
n oh k = j = p G 2 LEB mBE n < 1h B m2B = n o n o
J1
B LB mB 4hB C L m ð1h Þ2 =ð1þh Þ2
J2 < F
0 1 @ ; LE mE þ ;
A 1 F0 LEB mBE ; ; ð1hE Þ2 =ð1þhE Þ2
Tr G LB mB þ 2 G LE mE þ 2
1 1 2 >: 1hE m2 >
2 2
; B B
4hE E
2 n o
n o2 LB mB
þ2p m n
6 n o
AB LE mE
X o n o
6 L m AE A1 L m A
6 2 LBE mBE þ 2p 1 !e B
A1
G 2 LBE mBE 2p þ m; 2 ABE
4 m¼1 E (25)
n o n o 3
L m
2 LE mE
L m
þ2j1 2 LB mB þ2pþn
n o
LB mB
B B
X E E
X Gð2 LE mE þ 2p þ nÞ An eAB þAE n o
7
L m 7
E
G 2 LBE mBE 2p n þ i; AB þ AE 7:
n¼0 i¼1
n! ðAB þ AE Þi 5
a more compact form as in (25), at the top of the page Having derived closed-form upper bounds for the
(see p. 6), where the superscript “Tr” denotes truncate. truncation error of the infinite series representation
Similarly, a closed-form upper bound for the truncation for the ASC over h-m fading channels, we can readily
error of J h3 can be derived as deduce a similar expression for the truncation error of
2 ! the ASC for the case of -m fading channels. To this
2Lm 1h2E
mE E Gð2LmE Þ 1 F0 LmE ; ; 2mE 4h end, by following the same procedure and after long but
E
J Tr;h < basic algebraic manipulations, the closed-form bounds
3
G LmE þ 12 g2LmE þ2p eAE (26) in (27) (top of the next page, i.e., p.7) and (28) are
2LmE þ2p
X g m hE
2 deduced
Gð2LmE 2p þ m; AE Þ:
m¼1 mm
E ðhE þ 1Þ2
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 111
n o
n o G 2
LE mE n o 1 2 m2 =ð12 Þ2 n o 2
J1 k¼j¼p LB mB LB mB LE mE
J2 < F
0 1 ; LE mE þ ; B B B
F
1 0 LB mB ; ; 2
E
Tr G LB mB þ 12 G LE mE þ 12 2 2E m2E =ð12E Þ2 B
2 nL m o
2 2 LB mB þ2p m n
6 n o
2mB =ðg B ð1B2 ÞÞ X
E E o
6 L m 2m g
=ð ð1 ÞÞ g B ð12B Þ=ð2mB Þ LB mB g B ð12B ÞÞ
mB =ð
6G 2 LBE mBE þ 2p e E E E G 2 LE mE 2p þ m; 2
4 m¼1
g E ð1E Þ=ð2mE Þ
2 g E ð1E ÞÞ
mE =ð 2
(27)
n o n o
LE mE
þ2j1 2
LB mB
þ2pþn
n o
þ 2p þ nÞ
n i
2 LB mB
LB mB
X LE mE
X Gð2 LE mE
2mE 2mB 2mE
þ
n¼0 i¼1
n! gE ð1 2E Þ gB ð1 2B Þ gE ð1 2E Þ
2mB 2mE n o #
þ LB mB 2mE 2mE
g B ð12 Þ gE ð12 Þ
e B E G 2
LE mE 2p n þ i; þ :
gB ð1 2B Þ gE ð1 2E Þ
2p
Pout ðRS Þ ¼
GðLE mE ÞGðLB mB Þð1 2E ÞLE mE
X1 X 1 12LB mB 2j 2LE mE þ2k 2j 2k
2 mE B ð1 2B ÞLB mB Gð2LB mB þ 2jÞGð2LE mE þ 2kÞ E
2L m þ2k
k¼0 j¼0 k!j!GðLE mE þ k þ 0:5ÞGðLB mB þ j þ 0:5Þ gE E E 1 2E
" 2LE mE þ2k (31)
ð1Þmi 2RS i m
2mB 2LB mXB þ2j1 Xm
g E ð1 2E Þ ð12RS Þ
gB ð12 Þ
e B
2mE m¼0 i¼0
m! i
m
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
112 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
1;h
X1
p
2L m 2L m 2L m 1 2L m þL m
g E B B g B B B Gð2LB mB þ 2LE mE þ 2mÞmB B B ðhB þ 1Þ2LB mB 22LB mB RS 2mþ1 hE B B E E
Pout ðRS Þ ¼ L m 2L m
: (37)
m¼0 m!LB GðLB mB ÞGðLE mE ÞG LB mB þ 12 G LE mE þ m þ 12 hBB B mE B B ðhE þ 1Þ4LB mB þ2LE mE
1
4.3 Upper Bounds for the Truncation Errors where Pout ðRS Þ will be determined subsequently for both
In the previous section, we derived closed-form upper h-m and -m fading conditions. First, note that (29) can be
bounds for the truncation error of the infinite series repre- approximated using the expression of the lower bound5
sentations for the ASC over h-m and -m fading channels. Z 1
These bounds are tractable and they can determine the 1
Pout ðRS Þ ¼ fg E ðg E ÞFg B ð2RS g E Þdg E : (35)
required number of terms for certain target accuracy. Based 0
on this, similar closed-form upper bounds can be also
Next, an approximation for the CDF Fg B ðg E Þ will be
derived for the truncation error of the derived SOP infinite
derived since it is dependent on g B . To this end, as the PDF
series representations for the case of h-m and -m fading
in (3) is dominated by the term corresponding to k ¼ 0 in
conditions. To this end and given that the algebraic repre-
the summation, for gB ! 1, such expression can be used to
sentation of the series representations for the considered
obtain the corresponding CDF, i.e.,
ASC and SOP scenarios are algebraically similar, we follow
the same procedure in bounding the series which ultimately pffiffiffi 2LB mB 1 2L m
pm B ð1 þ hB Þ2LB mB g E B B
leads to the following simple upper bound Fg B ðg E Þ 2L mB
: (36)
LB GðLB mB ÞGðLB mB þ 0:5Þð4hB ÞLB mB g B B
Gð2LB mB ÞGð2LE mE Þ
Tr;h
Pout < Finally, substituting (3) and (36) into the aforementioned
G LB mB þ 12 G LE mE þ 12 approximated SOP expression and using [23, (Eq. 3.381.4)]
!
ð1 hE Þ2 m2E along with some algebraic manipulations, the asymptotic
1F0 LE mE ; ; (32) SOP can be derived in (37), at the top of this page. It is evi-
g2E h2E
! dent that the secrecy diversity gain is given by Gd ¼ 2LB mB ,
ð1 hB Þ2 which provides useful insights on the effect of the involved
1F0 LB mB ; ; ;
ð1 þ hB Þ2 key parameter on the overall system performance. To this
effect and recalling the Pochhammer symbol identities and
for the truncation error of the SOP over h-m fading channels. the properties of the hypergeometric functions as well as
Based on this, a similar closed-form upper bounds can be carrying out some algebraic manipulations, the following
deduced for the truncation error of the SOP under -m fad- compact closed-form asymptotic expression can be deduced
ing conditions. Hence, by performing the necessary change for the SOP over h-m fading channels at the high SNR
of variables and after some algebraic manipulations, the fol- regime:
lowing inequality is deduced
2L m þL m
1;h pGð2LB mB þ 2LE mE ÞhE B B E E
Gð2LB mB ÞGð2LE mE Þ Pout ðRS Þ ¼ L m
Tr;
Pout < F LB mB ; ; 2B G LE mE þ 12 ð1 þ hE Þ4LB mB þ2LE mE hBB B
1 1 0
G LB mB þ 2 G LE mE þ 2
1
2L m 2L m 1
! (33) gE B B mB B B 21þ2LB mB RS 2LE mE ð1 þ hB Þ2LB mB (38)
ð2E mE Þ2 2L m 2L m
1F0 LE mE ; ; 2 : LB g B B B mE B B GðLB mB ÞGðLE mE ÞG LB mB þ 12
g E ð1 2E Þ2
2F1 ðLB mB þ LE mE ; LB mB þ C1 ; C1 ; C2 Þ;
It is evident that the derived closed-form upper bounds in
where C1 and C2 are constant values given by
(32) and (33) are rather compact and tractable.
1
4.4 Simple Asymptotic SOP Representations C1 ¼ LE mE þ ; (39)
2
Capitalizing on the exact derived expressions for the SOP and
over h-m and -m fading channels, we can investigate their
ð1 hE Þ2
behavior at the high SNR regime in order to develop useful C2 ¼ : (40)
insights on the role of the involved parameters on the sys- ð1 þ hE Þ2
tem performance. To this end, assuming g B ! 1 and fixed
The above expression provides useful insights on the
g E ,4 our aim is to find an approximate expression of
1 G impact of the involved parameters on the overall system
Pout ðRS Þ in the form[29] Pout ðRS Þ Gc g B d , where Gc repre-
performance. Furthermore, it can be used to derive a similar
sents the secrecy array gain and Gd denotes the secrecy
expression for the SOP over -m fading channels. Thus, by
diversity gain. The secrecy diversity gain is defined as[30]
performing the necessary variable transformation and after
1 some algebraic manipulations, the following closed-form
: log Pout ðRS Þ
Gd ¼ lim ; (34) expression is deduced
g B !1 log ð
gBÞ
5. It has generally been shown that the lower bound of the SOP is in
close agreement with the exact SOP over the entire SNR regime (see
4. Intuitively, due to the high probability of successful eavesdrop- [11], [14] and the references therein). Since finding a solution for the
ping, the diversity gain associated with the case wherein gE ! 1, is lower bound of the SOP is easier than the exact SOP one, we use the for-
zero. mer to derive our asymptotic expressions of the SOP at high SNR.
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 113
Fig. 3. Exact average secrecy capacity versus gB for different number of
antennas and B ¼ 0:4, mB ¼ 0:5, E ¼ 0:1, mE ¼ 1:5 at g E ¼ 10 dB.
Fig. 2. Exact average secrecy capacity versus gB for different number of
antennas and hB ¼ 1:6, mB ¼ 2:5, hE ¼ 0:7, mE ¼ 0:5 at gE ¼ 10 dB.
In this section, representative numerical examples are plotted The ASC corresponding to hB ¼ 1:6; mB ¼ 3 and hE ¼ 0:7;
along with Monte Carlo simulations to assess the accuracy of mE ¼ 0:5 outperforms the other ASC performances corre-
the derived mathematical results. Without loss of generality, sponding to other fading parameters. This observation shows
it is assumed that RS ¼ 1 bit/s/Hz throughout the paper, that the fading parameters affect the overall security perfor-
unless otherwise stated. In this section, we present the ASC mance of the underlying system. In the same context, Fig. 3
and SOP performance metrics for various values of the fading illustrates the achieved ASC for different antennas values
parameters h, and m. It is worthwhile to recall that h repre- under -m fading conditions. It is shown that the number of
sents the scattered-wave power ratio between the in-phase antennas is critical also in this case as the achieved ASC dou-
and quadrature components of the wireless signal within any bles in the moderate and high SNR regimes when LB ¼ 2 is
one cluster and has real positive values, i.e., 0 < h < 1. On changed to LB ¼ 8 for fixed fading conditions and LE ¼ 2.
the other hand, 1 < < 1 represents the correlation coeffi- Fig. 4 illustrates the impact of h and m fading parameters
cient between the scattered-wave in-phase and quadrature on the achieved ASC. It is evident that the power ratio of
components of any multipath cluster, while m > 0 is the the in-phase and quadrature components of the wireless
number of multipath clusters in both fading models. signal in the two paths affects the overall system perfor-
Fig. 2 shows the average secrecy capacity for various num- mance even for a fixed number of used antennas. Particu-
ber of antennas at Bob and Eve, and different fading parame- larly in the case of considerably higher hE compared to hB , a
ters. It can be seen that the analytical results are in good non-negligible ASC degradation occurs across all SNR
agreement with the Monte-Carlo simulations across all the regimes. Likewise, this is also the trend regarding the effects
scenarios, which verifies the validity of the derived expres- of the correlation between the scattered-wave in-phase and
sions. Although the fading parameters for the main and quadrature components in Fig. 5. It is noticed that changes
eavesdropper links vary, we note that as the number of anten- of over 20 percent for fixed number of antennas. Therefore,
nas at Eve increases, the ASC decreases. This is expected since the variations of the number of antennas along with the
a high number of antennas at Eve yields some antenna diver- fading parameters h, and m affect the achieved ASC levels
sity and therefore a secure eavesdropping which can be detri- significantly. Moreover, a performance comparison between
mental to the security performance of the system. the underlying system over h-m, -m and the one over
In addition, different combinations of the fading parame- Rayleigh fading channels is provided. It can be seen that
ters for the main and eavesdropper channels are considered. there is a performance improvement or deterioration of the
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
114 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
Fig. 7. Exact average secrecy capacity versus gB for different gE values
Fig. 5. Exact average secrecy capacity versus gB for different fading and number of antennas for B ¼ 0:1, mB ¼ 2:5, E ¼ 0:1, mE ¼ 1:5.
conditions and LB ¼ 4, LE ¼ 4 at gE ¼ 10 dB.
Fig. 8. Exact secrecy outage probability versus g B for different gE values
Fig. 6. Exact average secrecy capacity versus gB for different g E values and hB ¼ 0:5, mB ¼ 0:5, hE ¼ 1:5, mE ¼ 2, and system parameters
and number of antennas for hB ¼ 2:5, mB ¼ 3:5, hE ¼ 0:85, mE ¼ 0:5. LB ¼ 2, and LE ¼ 4.
h-m or -m over the Rayleigh depending on the values the values can ensure acceptable SOP levels even at moderate
fading parameters of the generalized fading. For example, to high g E values. Conversely, low gB values lead to dra-
the ASC with hB ¼ 1, mB ¼ 1:5 and hE ¼ 3, mE ¼ 0:5 outper- matically poor performance even for low g E values since
forms the ASC corresponding to the Rayleigh from medium the achieved SOP is shown to lie around unity. Therefore,
to high-SNR regimes, while the ASC corresponding to besides ensuring that g B must be considerably greater than
hB ¼ 4, mB ¼ 0:5 and hE ¼ 2, mE ¼ 1:5 is outperformed by gE , it must be also ensured that the value of gB is not low
the one in the Rayleigh case. From these observations, one since the corresponding SOP will be rather poor regardless
can infer the scenarios under which the generalized h-m pro- of the value of g E and the number of antennas. These figures
vides a better security performance with respect to the well- also exhibit the accuracy of the derived closed-form asymp-
known fading models. totic expressions for the SOP, which exhibit a tight match
Figs. 6 and 7 demonstrate the effect of g E on the corre- with the respective exact results in both moderate and high
sponding ASC performance for fixed fading conditions and gB regimes.
two different antenna combinations. It is evident that the The behavior of the SOP over h-m and -m fading condi-
values of g E is, as expected, rather critical on the achieved tions is also illustrated in Figs. 10 and 11. It is evident that
ASC levels. For example, an over 20 percent ASC enhance- the derived secrecy diversity gain is confirmed, i.e.,
ment or degradation is observed when the value of g E Gd ¼ 2LB mB . Moreover, it is verified, as in the ASC case,
decreases or increases, respectively at both severe and light that for a fixed number of antennas at Bob, an increasing
fading conditions. Furthermore, it becomes evident that the number of antennas at Eve reduces considerably the overall
value of g E is comparably crucial to the number of selected security performance, whereas an increasing number of
antennas at Bob and Eve. antennas at Bob (for a fixed number of antennas at Eve),
The above behavior is also observed in Figs. 8 and 9, improves the overall system performance. Furthermore,
which demonstrate the SOP for the considered fading con- since it has been shown that the diversity gain is a function
ditions. It is noticed that changing the value of gE typically of mB , an increasing value of mB yields a performance
leads to SOP variations of about an order of magnitude improvement. Fig. 12 shows the SOP performance for vari-
across all g B values. In addition, it is shown that high gB ous target secrecy rates RS in h-m conditions. It can be noted
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 115
Fig. 10. Exact secrecy outage probability versus gB for different LE
values and hB ¼ 0:8, mB ¼ 1:5, hE ¼ 0:5, mE ¼ 2:5, and LB ¼ 1 at
gE ¼ 10 dB.
Fig. 12. Exact secrecy outage probability versus gB for various secrecy
that as RS increases, the SOP increases due to the fact that target rates RS . LE values and B ¼ 0:8, mB ¼ 1:5, E ¼ 0:5, mE ¼ 2:5,
and LB ¼ 1 at gE ¼ 10 dB.
the outage event is likely to occur. It is worth noting that the
aforementioned behavior can also be seen for different fad-
ing parameters and number of antennas.
Fig. 13 shows the SOP versus gB for different values
of the target secrecy rates RS for some well-known fading dis-
tributions, i.e., One-sided Gaussian, Rayleigh and Nakagami-
m. In this scenario, LB ¼ 4, LE ¼ 2, and g E ¼ 10 dB and it can
be seen that under the Rayleigh distribution (hi ! 1,
mi ¼ 0:5), the results coincide with the ones in [31, Fig. 3].
Moreover, this figure depicts the SOP performance under the
one-sided Gaussian and Nakagami-m with m ¼ 2. It is noted
that as RS decreases, the SOP performance improves in all sce-
narios, which corroborates the results in Fig. 12.
Finally, the overall impact of the involved different
parameters on the overall performance under the effects of
h-m and -m fading is demonstrated in Table 1, which
Fig. 13. Exact secrecy outage probability versus gB for various well-
depicts the corresponding ASC and the SOP for different known fading distributions. LB ¼ 4, LE ¼ 2 at gE ¼ 10 dB. Solid lines
values of hB , hE , mB , mE against LB , LE , gB and g E values. correspond to RS ¼ 0:1 and dashed lines correspond to RS ¼ 1.
The two considered cases are analyzed for the following real-
istic scenarios: (a) severe hB and mB with favorable hE and fading. In addition, we assume g E ¼ 0 dB for the case of ASC
mE ; (b) favorable hB and mB with severe hE and mE ; (c) severe and g E ¼ 15 dB for the case of SOP. In this context, it is first
B and mB with favorable E and mE ; and (d) favorable B verified that the ASC performance is affected by the different
and mB with severe E and mE . Without loss of generality, the values of the fading parameters hi , i and mi . For example, in
following parameters are used: gB ¼ f15 dB; 25 dBg and the h-m case, it can be noted that for various number of anten-
hi ¼ 4; mi ¼ 2 (severe), hi ¼ 1; mi ¼ 5 (light or favorable) for nas or main link SNR values, the ASC under severe hB ; mB
the case of h-m fading, where i ¼ fB; Eg, and i ¼ 0:9; mi ¼ 2 and light hE ; mE outperforms the one with favorable hB ; mB
(severe) and i ¼ 0:1; mi ¼ 5 (favorable) for the case of -m and severe hE ; mE . Similar observations are made for the case
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
116 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
TABLE 1
Effects of Fading Parameters on the SOP and ASC for Different System Parameters
15 3.81980 0.893285
LB ¼ 1, LE ¼ 1 25 7.08452 0.021604
severe hB -mB , light hE -mE 15 4.31824 0.950617
h m Fading Conditions
LB ¼ 2, LE ¼ 2 25 7.61596 0.001489
15 4.51655 0.976031
LB ¼ 3, LE ¼ 3 25 7.82320 0.000115
15 4.01385 0.815588
LB ¼ 1, LE ¼ 1 25 7.29112 0.003253
light hB -mB , severe hE -mE 15 4.43691 0.909554
LB ¼ 2, LE ¼ 2 25 7.73743 0.000073
15 4.60384 0.952317
LB ¼ 3, LE ¼ 3 25 7.91168 1.8091e-6
15 3.35750 0.864553
LB ¼ 1, LE ¼ 1 25 6.41581 0.047707
severe B -mB , light E -mE 15 1.16424 0.692372
m Fading Conditions
LB ¼ 2, LE ¼ 2 25 4.55584 0.006152
15 0.214783 0.371675
LB ¼ 3, LE ¼ 3 25 1.441140 0.000886
15 3.84456 0.705777
LB ¼ 1, LE ¼ 1 25 6.92600 0.001010
light B -mB , severe E -mE 15 3.24434 0.565260
LB ¼ 2, LE ¼ 2 25 5.53814 5.7843e-8
15 1.86507 0.295097
LB ¼ 3, LE ¼ 3 25 3.09455 1.919e-13
of -m fading. Furthermore, as the number of antennas LB SOP, which have been validated through comparisons with
and LE increases, the ASC improves regardless of the sce- respective results from Monte-Carlo simulations. Also, the
nario and the values of the main link SNR (e.g., 15 dB, 25 dB). asymptotic analysis for the SOP has shown to approximate
More importantly, it is seen that in both cases and for any well the exact SOP at the high SNR regime. The offered analyt-
number of antennas, the ASC in scenario (b) outperforms the ical results were then used in quantifying the effects of the dif-
one in scenario (a). It is worth mentioning that, the perfor- ferent parameters on the achieved PHY layer security. It was
mance difference between the two scenarios increases with shown that PHY depends significantly on these parameters,
the number of antennas between the corresponding values of which verifies the need for accurate characterization and
the main link SNR. In the same context, the effects of the fad- modeling of multipath fading conditions. Also, the offered
ing parameters on the SOP are discussed. In the h-m case, it results are expected to provide insights on the involved
can be seen that as the number of antennas increases, the SOP computational complexity and sustainability, since emerging
performance slightly deteriorates for gB ¼ 15 dB. Intuitively, communication systems are largely characterized by stringent
this can be explained as follows: the wiretap SNR is set to 15 quality of service and computational complexity require-
dB, i.e., the eavesdropper channel has the same quality as the ments. Possible future extensions to this work involve multi-
main link. However, as the main link SNR improves (i.e., ple-input multiple-output (MIMO) wiretap channels with
g B ¼ 25 dB), a substantial performance improvement is noted transmit antenna selection. MIMO systems have emerged as
as the number of antennas increases in both scenarios. good candidates for current and future communication sys-
As aforementioned, it is seen that hi , i and mi have affect tems and therefore, its consideration in the context of physical
the PHY security of the underlying system. It is noted that later security is of paramount importance. Another direction
such effects can be detrimental or can improve the system of research could investigate the impact of outdated channel
performance depending on the values of hi and mi for the information in transmit antenna selection.
main and wiretap channels. The offered conclusions on the
effects of the fading parameters on the system performance
ACKNOWLEDGMENTS
provide interesting insights that might be useful to future
designs since the issue of security is addressed in this work. The work of J. M. Moualeu was supported in part by the
NRF Funding Incentive for Rated Researchers. The work of
P. C. Sofotasios and S. Muhaidat was supported in part by
6 CONCLUSIONS Khalifa University under Grant No. KU/RC1-C2PS-T2/
This contribution addressed the secrecy performance of a 8474000137 and Grant No. KU/FSU-8474000122. This paper
wireless communication system over the generalized fading was presented in part at the IEEE International Conference
conditions characterized by the h-m and -m fading models. on Advanced Communication Technologies and Network-
Novel analytical expressions were derived for the ASC and ing (CommNet 2019).
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
MOUALEU ET AL.: PHYSICAL-LAYER SECURITY OF SIMO COMMUNICATION SYSTEMS OVER MULTIPATH FADING CONDITIONS 117
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.
118 IEEE TRANSACTIONS ON SUSTAINABLE COMPUTING, VOL. 6, NO. 1, JANUARY-MARCH 2021
Authorized licensed use limited to: ULAKBIM-UASL - Abant Izzet Baysal Univ Library. Downloaded on April 11,2021 at 14:52:32 UTC from IEEE Xplore. Restrictions apply.