i
Secure Vehicular Communications through
Reconfigurable Intelligent Surfaces
Yun Ai, Felipe A. P. de Figueiredo, Long Kong, Michael Cheffena,
Symeon Chatzinotas, Senior Member, IEEE, and Björn Ottersten, Fellow, IEEE
Abstract—Reconfigurable intelligent surfaces (RIS) is con-
arXiv:2011.14899v2 [eess.SP] 2 Dec 2020
sidered as a revolutionary technique to improve the wireless
system performance by reconfiguring the radio wave propagation
environment artificially. Motivated by the potential of RIS in
vehicular networks, we analyze the secrecy outage performance of
RIS-aided vehicular communications in this paper. More specifi-
cally, two vehicular communication scenarios are considered, i.e.,
a vehicular-to-vehicular (V2V) communication where the RIS
acts as a relay and a vehicular-to-infrastructure (V2I) scenario
where the RIS functions as the receiver. In both scenarios,
a passive eavesdropper is present attempting to retrieve the
transmitted information. Closed-form expressions for the secrecy
outage probability (SOP) are derived and verified. The results
demonstrate the potential of improving secrecy with the aid of
RIS under both V2V and V2I communications.
Index Terms—Physical layer security, reconfigurable intelligent
surfaces (RIS), vehicular communications, V2V, V2I.
I. I NTRODUCTION
ECONFIGURABLE intelligent surfaces (RIS) have re-
R cently appeared as a revolutionary technique to en-
hance network coverage and overcome the high attenuation
of millimeter wave (mmWave) and THz systems [1]. By
intelligently controlling a large number of low-cost passive
reflecting elements, the electromagnetic waves can be adapted
to the propagation environment. Thereby, the RIS functions
as a reconfigurable lens or reconfigurable mirror to beamform
the transmitted signals towards the desired user [2]. RIS are
also widely known as intelligent reconfigurable surfaces (IRS),
software-controllable surfaces, digitally controllable scatterers,
and large intelligent surfaces in literatures [1], [2].
Physical layer security (PLS) is widely considered as a
complement to conventional application layer encryption tech-
niques to enhance the communication secrecy in future (5G
and beyond) communication systems [3]. It has been demon-
strated both theoretically and experimentally that channel
fading, which is usually regarded as an adverse factor in
terms of reliability, can be utilized to enhance communication
security against eavesdropping [3]–[5]. Due to potential of RIS
and PLS technologies in future networks, investigation of the
combination of PLS and RIS-assisted systems has attracted
attention from the community recently [5]–[7]. The optimized
beamforming and phase shift design for secrecy rate of an
RIS-assisted mmWave system is investigated in [5]. In [6], the
secrecy outage probability (SOP) of an RIS-aided system is
Y. Ai and M. Cheffena are with the Norwegian University of Science and
Technology (NTNU), F. Figueiredo is with Instituto Nacional de Telecomuni-
cações (INATEL), Brazil, L. Kong, S. Chatzinotas, and B. Ottersten are with
the University of Luxembourg.
Fig. 1: Considered PLS scenarios for RIS-aided vehicular networks.
studied while the authors in [6] focus on the SOP of an RIS-
aided non-orthogonal multiple access (NOMA) system. The
expressions, in integral form, for the average secrecy capacity
(ASC) of an RIS-assisted vehicular network are derived in [7].
The realization of future autonomous vehicles requires
robust connections and high quality-of-service (QoS) be-
tween vehicles (i.e., vehicular-to-vehicular (V2V)) as well
as between vehicle and infrastructure (i.e., vehicular-to-
infrastructure (V2I)) [8]. The aforementioned advantages of
RIS technique make RIS-assisted vehicular communication an
appealing option to enhance vehicular network connectivity
[9]. Motivated by the latest advances in PLS analysis of
RIS-assisted systems as well as the potential of RIS-assisted
communication in vehicular networks, we study herein the se-
crecy performance of RIS-assisted vehicular communications
under passive eavesdropping. More specifically, we consider
the SOP performance under two communication scenarios of
vehicular communication, i.e., V2V and V2I scenarios. In the
V2V scenario, the RIS assists two vehicles that are blocked
by other objects to communicate with high QoS. In the V2I
scenario, the vehicle sends essential information to the RIS
that is close to the receiver to ensure robust transmission of
important messages to the intelligent transport infrastructure.
The main contributions of this paper are: (i) We analyze the
secrecy performance of RIS-assisted vehicular communication
under two realistic cases, where RIS are used as part of dual-
hop system and part of receiver, respectively; (ii) By avoiding
applying the central limit theorem (CLT) and instead adopting
a more versatile approach in the RIS analysis, the obtained
results are also valid when the number of RIS elements is
small; and (iii) We present some accurate or exact statistics
for RIS related signal-to-noise ratios (SNRs) (in Propositions
1, 2 and 3), which can be useful for RIS-related analysis.
Notations: [0, x]+ = max(x, 0), E[·] is the expectation
operator, Γ(·) and Γ(·, ·) are Gamma and incomplete Gamma
 ii

functions, respectively [10, Eq. 8.3], Kv (·) is the modified

1
Bessel function of second kind with order v [10, Eq. 8.407], Verification of Eq. (7)

Jv (·) is the Bessel function of first kind [10, Eq. 8.402], -1

0.2 10
Gm,n m,n
p,q (·) is the Meijer G-function [10, Eq. (9.3)], Hp,q (·) is Verification of Eq. (5)
m,n:s,t:i,j
the Fox H-function [11, Eq. 1.2], and Hp,q:u,v:e,f (·) is the
10-2
extended generalized bivariate Fox H-function [11, Eq. 2.56]. 0.15

10-3
II. C HANNEL AND S YSTEM M ODELS 0.1

In this paper, we consider the secrecy outage probability 10-4

0.05
of RIS-assisted V2V and V2I systems. The considered classic
Wyner’s wiretap model is illustrated in Fig. 1. 10-5
0
0 10 20 30 40 50 0 5 10 15 20
A. V2V Communications
Fig. 2: Verifications of statistics functions in Eqs. (5) and (7).
In the V2V case, a vehicle S communicates secret informa-
tion with another vehicle D that has blockage between them
with the aid of an RIS with N elements. The signals sent by S √ PN
where γ D = N dpP1s dp1 , an =αn βn , and A = γ D n=1 an .
is overheard by an eavesdropper E close to S. All vehicles are 0 SR RD
When N is large, the CLT can be applied and the random
assumed to be equipped with single antennas for simplicity.
variable (RV) A can be approximated by a Gaussian RV
The received signal at the receiver vehicle D via the RIS is
p and the RV γD can be considered to follow noncentral-χ2
yD = Ps · hTSR ωhRD · s + w0 , (1) distribution [2]. Nevertheless, the CLT approximation becomes
inaccurate while N is not large. In [13], an approximation of
where Ps is the transmit power of S, s is the transmit- γD is obtained by considering the RV an = αn βn as Gamma
ted signal with unit energy, w0 is the zero-mean additive distributed with αn and βn being independent and identically
white Gaussian noise (AWGN) with variance N0 , ω = distributed (i.i.d.) RVs. However, it is unrealistic to assume that
diag(̟1 (φ1 )ejφ1 , . . . , ̟N (φN )ejφN ) is the diagonal matrix αn and βn are identically distributed since they correspond to
consisting of the reflection coefficients produced by each two completely different propagation links. Here, we present
reflection element of the RIS. The vector hSR contains the in Proposition 1 the statistics of γD under the assumption of
channel gains from S to each element of RIS and the vector αn and βn being independent but not identically distributed
hRD includes the channel gains from each element of RIS to (i.n.i.d.) RVs by approximating the RV A as a Gamma RV.
D, which are expressed as [12] Proposition 1: When hSR,n ∼ CN (0, νSR ) and hRD,n ∼
p1
hSR
−
= αT Θ · dSR2 , (2a) CN (0, νRD ) are i.n.i.d. complex Gaussian RVs with |hSR,n | =
PN
2
T − 1
p αn and |hRD,n | = βn , then the RV γD = γ D · n=1 αn βn
hRD = β ϕ· dRD2 , (2b) under both small and large values of N can be accurately
where the column vectors α and β contain the amplitudes of described by the following probability density function (PDF)
the corresponding channel gains. Each element of α follows and cumulative distribution function (CDF):
independent Rayleigh distribution resulting from the scattering  √ 
1 x
 
kD −2

around the vehicle. Similarly, every element of β is also fγD (x) = kD

·x 2
· exp − , (5)
2Γ(kD )ηD ηD
independent Rayleigh distributed. Θ = [e−jθ1 , · · · , e−jθN ]  √ 
and ϕ = [e−jϕ1 , · · · , e−jϕN ] with θn and ϕn , n = 1, · · · , N , 1 x
FγD (x) =1 − · Γ kD , , (6)
being phase of the corresponding link; dxy is the distance Γ(kD ) ηD
between nodes x and y; and p1 is the path loss exponent for √ √
N π2 γ D (16−π 2 ) νSR νRD
the link from or to the RIS. where kD = 16−π 2 , and ηD = .
4π
From (1) and (2), the instantaneous SNR γD at D becomes Proof : Please refer to Appendix A and Fig. 2.
N 2 Due to the high mobility of vehicles S and D, perfect phase
αn βn ̟n (φn ) · ej(φn −θn −ϕn )
P
Ps · estimation required for ideal RIS might be challenging. Next,
n=1
γD = . (3) we consider the worst case of RIS phase shifting, where the
N0 dpSR
1
dpRD
1
phase errors σnD are uniformly distributed in complex plane to
We first consider perfect knowledge of the channel state evaluate the impact of imperfect RIS on secrecy performance.
D 2
information (CSI) at RIS as in [5]–[7], which enables ideal To obtain the statistics of RV γD = γ D ·
PN jσn
n=1 αn βn e ,
phase shifting (i.e., ̟n (φn ) = 1 and σnD = φn − (θn + ϕn )
we interpret the mathematical problem as an isotropic two-
= 0). Then, the maximum instantaneous SNR can be achieved
dimensional random walk, where the n-th step size is αn βn
at D and is expressed as
and n-th direction is σnD that is uniformly distributed.
!2 !2
XN X N Proposition 2: When hSR,n ∼ CN (0, νSR ) and hRD,n ∼
γD = γ D · αn βn = γD · an = A2 , (4) CN (0, νRD ) are i.n.i.d. complex Gaussian RVs with |hSR,n | =
n=1 n=1 αn and |hRD,n | = βn , and σnD is uniformly distributed
 iii

between 0 Pand 2π, the exact PDF and CDF of the RV III. S ECRECY P ERFORMANCE A NALYSIS
N D 2
jσn
γD = γ D · n=1 αn βn e are given by A. Secrecy Outage Probability (SOP)
N −1
2·x 2
 r 
x The secrecy rate indicates the maximum achievable rate
fγD (x) = · KN −1 2 , (7) the main channel can achieve in secrecy. The instantaneous
B γ D νSR νRD
N +1   secrecy rate Cs of the considered wiretap model is [15]
x 2 2,1 x 1−N
FγD (x) = · G1,3 2
N +1 , (8) +
B
N −1 N −1
γ D νSR νRD 2 ,− 2 ,− 2 Cs (γD , γE ) = [ln(1 + γD ) − ln(1 + γE ), 0] , (14)
N +1
where B = Γ(N ) · (γ D νSR νRD ) 2 . where γD and γE are the instantaneous SNRs of the main link
Proof : Please refer to Appendix B and Fig. 2. from S to D and wiretap channel from S to E, respectively.
For the link between vehicles S and E, the double-bounce Under passive eavesdropping, the legitimate transmitter S
scattering components caused by scatterers around both vehi- and receiver D have no channel state information (CSI) of the
cles’ local environments lead to a cascaded Rayleigh fading eavesdropper E. Then, the node S cannot adapt the coding
process [8]. Therefore, we use the double Rayleigh model to scheme to E’s channel state, but resorts to set the secrecy rate
characterize of the dynamic fading link between S and E. The to a constant target rate Rs . When the instantaneous secrecy
PDF and CDF of the instantaneous SNR γE are expressed as rate is larger than the target rate, i.e., Cs > Rs , perfect secrecy
2
 r 
x can be guaranteed. Otherwise, when the instantaneous secrecy
fγE (x) = · K0 2 , (9) rate is less than the target rate, i.e., Cs ≤ Rs , secrecy will
γE γE
  be compromised and secrecy outage occurs, the probability of
x x
r r
FγE (x) = 1 − 2 · K1 2 , (10) which is given by the secrecy performance metric SOP [3].
γE γE The SOP is mathematically expressed as [15]
where γ E = N Pdsp2 , and p2 is the path loss exponent for the Po =Pr [Cs (γD , γE ) ≤ Rs ] = Pr [γD ≤ ΘγE + Θ − 1]
0 SE
links between vehicles. Z ∞ Z (1+γE )Θ−1
= fγD ,γE (γD , γE ) dγD dγE , (15)
B. V2I Communications 0 0

Under the V2I scenario, a vehicle S sends essential informa- where Θ = exp(Rs ) ≥ 1, and fγD ,γE (·, ·) is the joint PDF of
tion to the intelligent transportation infrastructure D while an the RVs γD and γE .
eavesdropper E close to S attempts to eavesdrop the signals Next, we investigate the secrecy outage performance under
sent by S. The infrastructure D consists of an RIS and RF V2V and V2I scenarios, respectively.
receiver, where the RIS is deployed close to the RF receiver
such that the channel attenuation between them can be ignored B. V2V Communications
[2]. Therefore, the RIS and RF receiver together are considered
With the RVs γD and γE being independent, and utilizing
as a receiver from the perspective of analytical analysis.
the Parseval’s formula for Mellin’s transform, the SOP can be
Under the described V2I scenario with ideal phase shifting,
alternatively evaluated as [16]
the instantaneous SNR at D can be written as
!2 Z ∞ Z (1+γE )Θ−1 
N
X Po = fγD (γD ) dγD · fγE (γE )dγE
γD = γ D · αn = γD · C 2, (11)
Z0 0
n=1 M[fγE (x), s]
Ps PN = M[FγD (Θx+Θ−1), 1−s] · ds, (16)
where γ D = p1
N0 dSD
and C = n=1 αn . To solve the statistics L1 2πj
of C is equivalent to obtain the statistics of the received where L1 is the integration path from c − j∞ to c + j∞ with
signal for the single-input multiple-output (SIMO) system with c being some constant, and M[f (x), s] represents the Mellin
equal gain combining, where the exact closed-form solution transform of the function f (x) [10, Eq. 17.41].
is unavailable for N > 2. Next, we present an accurate Lemma 1: The SOP of the RIS-assisted vehicular commu-
approximation of γD in Proposition 3. nication under the V2V scenario as illustrated in Fig. 1 with
Proposition 3: When hSD,n ∼ CN (0, νSD ) and αn = perfect RIS phase shifting can be expressed as
PN 2
|hSD,n |, the PDF and CDF of the RV γD = γ D · n=1 αn
 
0,1:1,1:1,2 E1 E2 E3 ηD Θγ E
can be closely approximated by Po = D1 · H1,0:2,2:1,2 √ , , (17)
− F2 F3 Θ − 1 Θ−1
xN −1
 
x
fγD (x) = · exp − , (12) where D1 = γ (Θ−1) , E1 = (2; 1, 1); E2 = (1−kD , 1), (1, 1);
(γ D νSD )N ΩN Γ(N ) γ νSD ΩD E ΘΓ(kD )
D D 1
F2 = (0, 1), (1, 2 ); E3 = (1, 1), (1, 1); and F3 = (1, 1).

1 x
FγD (x) =1 − · Γ N, , (13) Proof : Please refer to Appendix C.
Γ(N ) γ D νSD ΩD
2 Lemma 2: The SOP of the RIS-assisted vehicular com-
where ΩD = 1 + Γ( 32 ) · (N − 1).

munication under the V2V scenario as shown in Fig. 1 with
Proof : The results follow by employing the result in [14, uniform distributed phase estimation error can be written as
Eq. (22)] for the special case of independent Rayleigh RVs.  
0,1:1,2:1,2 E4 E5 E6 γ D νSR Θγ E
Under the V2I scenario, the PDF and CDF of the eaves- Po = D2 ·H1,0:2,3:1,2 −1 Θ−1 , (18)
,
dropper’s SNR are given as in (9) and (10). − F5 F6 (Θ−1)νRD
 iv

100 100

10-1 10-2

10-4
10-2
SOP

SOP
10-6
10-3

10-8
10-4

10-10
-15 -10 -5 0 5 10 15 20 16 20 25 30 35 40 45 50 55 60 64

Ps
Fig. 3: SOP vs. N0
for varying number of RIS elements under V2V. Fig. 4: SOP vs. N for varying number of RIS elements under V2I.

N +3
where D2 = (BΘ)−1 (θ − 1) 2 , E4 = ( N 2+5 ; 1, 1); E5 = when only phase shifting is considered. The large differences
( 3−N 1+N 3+N
2 , 1), ( 2 , 1), ( 2 , 1); F5 = ( N2−1 , 1), ( 3+N
2 , 1);
between the lower and upper limits demonstrate the adverse
E6 = (1, 1), (1, 1); and F6 = (1, 1). effect of imperfect phase shifting for the RIS-aided system
Proof : The results follow by applying the same rationale as performance.
in Appendix C. Figure 4 shows the SOP performance improvement with
the increase of the number of RIS elements N under the V2I
C. V2I Communications scenario. As expected, when the number of RIS elements N
increases, the SOP performance improves even when the signal
Under the V2I scenario and with the independence between attenuation for the legitimate receiver is much larger than that
the RVs γD and γE , the SOP can be evaluated as [17] to the eavesdropper because of longer signal transmission. It
Z ∞
is also observed that the SOP in the logarithm scale exhibits a
Po = FγD (Θx + Θ − 1) · fγE (x) dx. (19)
0
linear relation with respect to the number N when N is large.
The slope of the linear relation is dependent on the distances
Lemma 3: The SOP of the RIS-aided vehicular network
between the communicating entities and is irrelevant to the
under the V2I scenario depicted in Fig. 1 can be evaluated by
transmission power.
N −1 − Θ−1 k   k−j
1 X e γ D ΩD X
k k 1
Po =1 − · · Θ · · 1− A PPENDIX A: P ROOF
OF P ROPOSITION 1
γE (γ D ΩD )k · k! j=0 j Θ
k=1 √
We assume that the RV A = γ D · N
P
 −(j+1)   n=1 αn βn in (4) can
Θ γ D ΩD −j be approximated by a Gamma RV Z with shape parameter
· · G2,1
1,2 . (20)
γ D ΩD Θγ E 0,0 kD and scale parameter ηD . It follows immediately that the
Proof : Please refer to Appendix D. first and second moments of the RV Z are E[Z] = kD ηD and
E[Z 2 ] = kD ηD2
, respectively. Since hSR,n ∼ CN (0, νSR ) and
IV. N UMERICAL R ESULTS AND D ISCUSSIONS βn ∼ CN (0, νRD ), we √ have that αn = |hSR,n | is a Rayleigh
πνSR
RV with E[αn ]√= 2 and E[α2n ] = νSR . Similarly, we
In this section, we numerically evaluate the secrecy outage πνRD
performance under the considered V2V and V2I scenarios. have E[βn ] = 2 and E[βn2 ] = νRD . Next, we find the
The path loss exponents are set as p1 = 2.1 and p2 = 2.3 for first and second moments of the RV A as follows:
N √
simulation purpose. p X π 2 νSR νRD
Figure 3 demonstrates the SOP in terms of the transmit SNR E[A] = γ D · E[αn ] E[βn ] = √ , (21)
n=1
4 · ( γ D · N )−1
Ps
γt = N under the V2V scenario. It can be seen that the SOP
0  N 2  N
performance improves significantly by even a relatively small 2
X X
E α2n E βn2
p    
E[A ] =E γD · αn βn = γD ·
increase in the number of RIS antenna elements. However,
n=1 n=1
when the SNR is large enough, the SOP performance stagnates N N
and further increasing the transmit SNR (namely the transmit
XX
+ γD · E[αn ] · E[βn ] E[αm ] E[βm ]
power) does not improve the SOP performance any longer. n=1 m6=n
Instead, increasing the number N can significantly improve
π 2 νSR νRD · N (N − 1)
 
this performance bound that can not be enhanced by increasing =γ D · N νSR νRD + . (22)
the transmission power. It is obvious that the curves for perfect 16
IRS shifting and random phase shifting represent the upper and Solving equalities E[A] = E[Z] and E[A2 ] = E[Z 2 ], we
lower limits of the secrecy outage performance, respectively, obtain the values of kD and ηD , which determines the statistics
 v

of Gamma RV Z. Finally, with change of RV γD = Z 2 , we Substituting (29) into (28), we obtain after some algebra
obtain the statistic function of RV γD as in Proposition 1.  s−1
(c) Γ(1−s) Θ 1
M[FγD (Θx+Θ−1), 1−s] = · ·
A PPENDIX B: P ROOF OF P ROPOSITION 2 Γ(kD ) Θ−1 2πj
√ −ξ
D
Γ(kD +ξ)Γ(−ξ)Γ(s+ ξ2 −1)
PN
Θ−1
Z
jσn
The RV r = n=1 αn βn e can be interpreted as the · · dξ, (30)
distance to origin after N ’random walks’ with the n-th step L2 Γ(1−ξ)Γ( 2ξ ) ηD
of length rn = αn βn in the direction of σnD . We first obtain
the statistics of the RV rn , which is the product of two i.n.i.d where (c) is obtained with the aid of [10, Eq. (3.194.3)] with
Rayleigh RVs. L2 being some contour.
2 2 Rewriting the Bessel function in (9) in Meijer G-function’s
∞ Z ∞ − x − r
1 4r·e νSR νRD x2 form, the Mellin transform M[fγE (x), s] can be solved as
Z r
frn (r) = fαn(x)fβn dx = dx
0 x x 0 xνSR νRD Z ∞ s−1
Γ(s) · Γ(s) γ sE
 
x x −
M[fγE (x), s] = · G2,0
 
4r 2r 0,2 0,0 dx = .
= · K0 . (23) 0 γE γE γE
νSR νRD νSR νRD (31)
Since the angle density functions of σnD are uniform, the
Next, substituting (30) and (31) into (16), the SOP can
conditional PDF of RV r is given with the Kluyver’s result in
be expressed, after some mathematical manipulations, as the
terms of integral over Bessel functions [18, p. 420]
Z ∞ following double contour integral:
rJ0(rx)
fr (r|r1 , · · · , rN ) = J0(r1 x) . . . J0(rN x)dx. (24) 2 Z Z  ξ
(Θ − 1)

0 x−1 1 ηD
Po = · · √
Then, the PDF of the RV r can be obtained directly from γ E ΘΓ(kD ) 2πj L1 L2 Θ−1
 s
NZ ∞ Z ∞Z ∞ Θ · γE Γ(1 − s) · Γ(s) · Γ(s)
· · · Γ(kD + ξ)
  
4 2r1
fr (r) = r xJ0 (rx) · · · r1 K0 Θ−1 Γ(1 − ξ) · Γ( 2ξ )
νSR νRD 0 νSR νRD
 0 0
· Γ(−ξ) · Γ(s + ξ − 1) dξds. (32)
2rN
·J0 (r1 x)dr1 . . . rN K0 · J0 (rN x)drN dx. (25)
νSR νRD Recalling the representation of bivariate Fox H-function in
Next, utilizing the following equalities [18, Chpt. 13] terms of double contour integral [11, Eq. 2.56] for (32), the
Z ∞ exact expression for the SOP can be expressed in terms of
(νSR νRD )2
 
2rn
rn K0 J0 (rn x)drn = , (26) bivariate Fox H-function as shown in Lemma 1.
0 νSR νRD (νSR νRD x)2 +4
41−N KN −1 νSR2rνRD
Z ∞

(νSR νRD )2N J0 (rx)x
dx = , (27) A PPENDIX D: P ROOF OF L EMMA 2
0 [(νSR νRD x)2 +4]N Γ(N )(νSR νRD r)1−N
in (25) leads to the PDF of the RV r. With change of RV Expressing the incomplete Gamma function in (13) in series
γD = γ D r2 , we obtain the exact PDF of the RV γD in (7). [10, Eq. (8.354)] and then substituting the resulting expression
The CDF of γD in (8) follows by using its relation with the of CDF and (9) into (19), the SOP can be rewritten as
PDF.
∞ N−1
−1 X[(1+x)Θ−1]k
 r  
2
Z
x − (1γ+x)Θ
A PPENDIX C: P ROOF OF L EMMA 1 Po = K0 2 · 1−e D ΩD
· dx
0 γE γE (γ D ΩD )k · k!
We first solve the function M[FγD (Θx + Θ − 1), 1 − s]. k=1
N −1
Based on the definition of Mellin transform, we have 2 − Θ−1 X 1
Z ∞ =1 − · e γ D ΩD ·
γ E (γ D ΩD )k · k!
M[FγD (Θx + Θ − 1), 1 − s] = x−s FγD (Θx + Θ − 1) dx Z ∞  r  k=1
√ 0 x − Θx k
(a)
Z ∞ −s
x Θx + Θ − 1 1

(b) (2πj)
−1 · K0 2 · e γ D ΩD · [(1+x)Θ − 1] dx. (33)
= 1,1
· G1,2 0 γ
kD ,0 dx =
E
0 Γ(kD ) ηD Γ(kD )
Z Z ∞ √ −ξ k
Rewriting [(1 + x)Θ − 1] in series and relevant terms in
−s Γ(kD +ξ)Γ(−ξ) Θx+Θ−1
· x dxdξ, (28) Meijer G-functions, the definite integral in (33) becomes
L2 0 Γ(1−ξ) ηD
k
Θk · kj Z ∞ j 1,0 Θx −

   
where (a) is obtained by converting the incomplete Gamma 2,0 x −
X
I= · x ·G0,1 ·G0,2 dx
function in (6) into Meijer G-function; and (b) is derived by 1 j−k 0 γ D ΩD 0 γ E 0,0


j=0 1−Θ
representing the Meijer G-function in terms of contour integral
1 k−j
k
Θk · kj · 1 − Θ


 
and changing the integration order. The inner integral in (28) 2,1 γ D ΩD −j
X
= · G1,2 . (34)
can be solved with the help of [10, Eq. (3.194.3)] as
j=0 2 · γ ΘΩD

(j+1) Θγ E 0,0
D
Γ(1 − s)Γ(s + 2ξ − 1)
Z ∞
x−s
ξ dx = ξ
1−s . (29) Finally, substituting (34) into (33) leads to the closed-form
0 (Θx + Θ − 1) 2 Γ( ξ2 )(Θ − 1) 2 Θ−1
Θ
expression for the SOP under the V2I scenario in Lemma 2.

R EFERENCES
[1] Q. Wu and R. Zhang, “Towards smart and reconfigurable environment:
Intelligent reflecting surface aided wireless network,” IEEE Communi-
cations Magazine, vol. 58, no. 1, pp. 106–112, January 2019.
[2] E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M.-S. Alouini, and
R. Zhang, “Wireless communications through reconfigurable intelligent
surfaces,” IEEE Access, vol. 7, pp. 116 753–116 773, September 2019.
[3] J. Barros and M. R. Rodrigues, “Secrecy capacity of wireless channels,”
in Proceedings of IEEE International Symposium on Information Theory
(ISIT). Seattle, Washington, USA: IEEE, July 2006, pp. 356–360.
[4] Y. Ai, M. Cheffena, T. Ohtsuki, and H. Zhuang, “Secrecy performance
analysis of wireless sensor networks,” IEEE Sensors Letters, vol. 3,
no. 5, pp. 1–4, May 2019.
[5] J. Qiao and M.-S. Alouini, “Secure transmission for intelligent reflect-
ing surface-assisted mmWave and Terahertz systems,” IEEE Wireless
Communications Letters, vol. 9, no. 10, pp. 1743–1747, June 2020.
[6] L. Yang, Y. Jinxia, W. Xie, M. Hasna, T. Tsiftsis, and M. Di Renzo,
“Secrecy performance analysis of RIS-aided wireless communication
systems,” IEEE Transactions on Vehicular Technology, vol. 69, no. 10,
pp. 12 296–12 300, July 2020.
[7] A. U. Makarfi, K. M. Rabie, O. Kaiwartya, X. Li, and R. Kharel, “Phys-
ical layer security in vehicular networks with reconfigurable intelligent
surfaces,” in Proceedings of IEEE Vehicular Technology Conference
(VTC-Spring). Antwerp, Belgium: IEEE, May 2020, pp. 1–6.
[8] Y. Ai, M. Cheffena, A. Mathur, and H. Lei, “On physical layer security
of double Rayleigh fading channels for vehicular communications,”
IEEE Wireless Communications Letters, vol. 7, no. 6, pp. 1038–1041,
December 2018.
vi
[9] Y. Liu, X. Liu, X. Mu, T. Hou, J. Xu, Z. Qin et al., “Reconfig-
urable intelligent surfaces: Principles and opportunities,” arXiv preprint
arXiv:2007.03435, 2020.
[10] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
Products, 7th ed. Burlington, MA, USA: Academic Press, 2007.
[11] A. M. Mathai, R. K. Saxena, and H. J. Haubold, The H-Function: Theory
and Applications. New York, NY, USA: Springer, 2009.
[12] L. Yang, F. Meng, J. Zhang, M. O. Hasna, and M. Di Renzo, “On the
performance of RIS-assisted dual-hop UAV communication systems,”
IEEE Transactions on Vehicular Technology, vol. 69, no. 9, pp. 10 385–
10 390, June 2020.
[13] S. Atapattu, R. Fan, P. Dharmawansa, G. Wang, J. Evans, and T. A.
Tsiftsis, “Reconfigurable intelligent surface assisted two-way com-
munications: Performance analysis and optimization,” arXiv preprint
arXiv:2001.07907, 2020.
[14] Z. Hadzi-Velkov, N. Zlatanov, and G. K. Karagiannidis, “An accurate
approximation to the distribution of the sum of equally correlated
Nakagami-m envelopes and its application in equal gain diversity
receivers,” in Proceedings of IEEE International Conference on Com-
munications (ICC). Dresden, Germany: IEEE, June 2009, pp. 1–5.
[15] Y. Ai, A. Mathur, H. Lei, M. Cheffena, and I. S. Ansari, “Secrecy
enhancement of RF backhaul system with parallel FSO communication
link,” Optics Communications, vol. 475, November 2020.
[16] Y. Ai, L. Kong, and M. Cheffena, “Secrecy outage analysis of double
shadowed Rician channels,” Electronics Letters, vol. 55, no. 13, pp.
765–767, June 2019.
[17] Y. Ai, A. Mathur, M. Cheffena, M. R. Bhatnagar, and H. Lei, “Phys-
ical layer security of hybrid satellite-FSO cooperative systems,” IEEE
Photonics Journal, vol. 11, no. 1, pp. 1–14, February 2019.
[18] G. N. Watson, A Treatise on the Theory of Bessel Functions. Cambridge,
UK: Cambridge University Press, 1922.