Mobility Improves NOMA Physical Layer Security

Full text

Mobility Improves NOMA Physical Layer Security
Jie Tang∗†, Long Jiao†, Ning Wang†,PuWang
†, Kai Zeng †, Hong Wen∗
∗University of Electronic Science and Technology of China, ChengDu, China
†George Mason University, Fairfax, Virginia 22030, U.S.A
Email: {jtang20; ljiao; nwang5; pwang20; kzeng2}@gmu.edu; sunlike@uestc.edu.cn
Abstract—Physical layer security of non-orthogonal multiple
access (NOMA) systems has attracted great attentions. However,
the impact of mobility on physical layer security of NOMA
systems has not been well studied. In this paper, to fill this gap,
we investigate the impact of random mobility on physical layer
security of NOMA systems. Considering scenarios where a base
station (BS) or access point (AP) communicates to two random
mobile users with a passive eavesdropper in two concentric
circles, we study the secrecy performance with combinations of
two typical random mobility models: random waypoint (RWP)
and random direction (RD). A general analytical framework to
numerically calculate the average secrecy rates of NOMA mobile
users under steady state is provided. By comparing secrecy
performance of mobile users with static users, we find that RWP
mobile users can achieve higher average secrecy rates than the
users with other mobility combinations. Meanwhile, two types
of secrecy fairness for mobile users are fully considered and we
propose a novel sum average secrecy rate maximization problem,
subject to average power limits and users’ QoS (quality of
service) requirements. Considering eavesdropper’s channel state
information (CSI) is unknown to BS, we propose a threshold
power allocation strategy to improve the sum average secrecy
rate of NOMA mobile users. Extensive numerical simulations
are conducted to validate our model and theoretical analysis.
Index Terms—5G, physical layer security, non-orthogonal
multiple access (NOMA), secrecy rate, mobility
I. INTRODUCTION
Physical layer security [1, 2] has been recognized as a
promising technique to directly leverage the unique properties
of wireless medium or radio to enhance the security for 5G
wireless networks. Work [1] investigated physical layer secu-
rity under wireless fading channel and proposed the average
secrecy rate (secrecy capacity) and secrecy outage probability
(SOP) to measure the secrecy performance. Recently, this
concept has been applied to non-orthogonal multiple access
(NOMA) systems [3], in order to realize the secure transmis-
sion [4–7], which has attracted great attentions.
NOMA [3] can exploit the power domain for multiple
access, which has been recognized as a promising technique
in 5G networks. Work [4] investigated NOMA physical layer
security in large-scale networks with stochastic geometry. By
assuming that the legitimate users are uniformly distributed
in the circular region and eavesdroppers follow Poisson point
process (PPP) in an infinite two dimension plane, the authors
derives the exact expressions of the SOP under quasi-static
Rayleigh fading. In [4], a protected zone around base station
(BS) is employed, in which no eavesdroppers are allowed to
locate inside. For multiple NOMA users, work [5] investigated
the problems of minimizing the transmit power subject to SOP
and QoS constraints. It has found that the SOP constraint
does not change the optimal NOMA decoding order of each
user. Work [7] studied the sum secrecy rate maximization of
NOMA systems subject to all users’ QoS requirements. Work
[6] investigated the physical layer security for cooperative
NOMA systems, where both amplify-and-forward (AF) and
decode-and-forward (DF) protocols are considered.
However, most existing works only study NOMA physical
layer security under static scenarios. In many scenarios of
wireless networks, the users can be mobile, e.g., cellular users
walking on the street or riding on a bus, or students walking
on a campus while connecting with campus WiFi networks.
For the networks with random mobile users, work [8] derives
the probability density of received power in mobile receivers
under random waypoint (RWP) [9] and random direction (RD)
[10] mobility models. However, it only studies receive signal
quality in random mobile networks [8], the impact of mobility
on NOMA physical layer security is not well understood.
To fill this gap, in this paper, we investigate the impact
of random mobile users on physical layer security of NOMA
systems. Considering scenarios where a BS or AP communi-
cates to two random mobile users with a passive eavesdropper
in two concentric circles, we study the secrecy performance
with combinations of two typical random mobility models:
RWP and RD. A general analytical framework for numeri-
cally calculating the average secrecy rates of NOMA mobile
users under steady state is provided. By comparing secrecy
performances of mobile users with static users, we find that
RWP mobile users can achieve highier average secrecy rates
than users with other mobility combinations. Meanwhile, for
mobile users, two types of secrecy fairness are considered and
different from [7], we propose a sum average secrecy rates
maximization problem which are subject to average power
limits and users’ QoS requirements. Considering eavesdrop-
per’s CSI is unknown to BS, we propose a threshold power
allocation strategy to improve the sum average secrecy rate
of NOMA mobile users. Extensive numerical simulations are
conducted to validate our model and theoretical analysis.
II. SYSTEM MODEL
A. NOMA physical layer security model with mobile users
The proposed model can be shown in Fig. 1, considering
four single antenna nodes are located in two concentric
circular regions (cell 1 and cell 2) with radius R1and R2
(R1≤R2), respectively. Alice is the BS or AP located in
978-1-5386-4727-1/18/$31.00 ©2018 IEEE

Figure 1: NOMA security model with random mobile users
the center of two cells, communicating with legitimate mobile
users, Bob and Charlie. A passive eavesdropper Eve is located
somewhere in the region of cell 2. The legitimate users (Alice,
Bob and, Charlie) do not know the exact location of Eve.
During the communication period, Bob and Charlie move
randomly and independently in circular cell 1 and cell 2,
respectively. We study the NOMA physical layer security of
mobile Bob and Charlie under combinations of two typical
random mobility models: RWP [9] and RD [10] which are
described below.
RWP Bob and Charlie: For RWP Bob, at first, he
randomly chooses a point D0in cell 1. Then, he randomly
chooses a coordinate D1(D1is uniformly distributed in
cell 1) as his next destination point and moves to it with
a constant speed. After Bob arrives at the destination point
D1, he chooses a new destination D2and moves to it with a
constant speed. Bob continues this process until the end of the
mobility. The RWP Charlie conducts the same mobile pattern
as Bob in cell 2.
RD Bob and Charlie: For RD Bob, he randomly starts at
point D0of cell 1. Then, he randomly chooses a direction
θ1(0≤θ1≤2π) and moves with a constant speed. After
a random period of time, he chooses a new direction θ2and
moves to it with a constant speed for a new random travel
time, and continues this process. The RD Bob may dash on
the border of the region. Here we assume Bob rebounds with
a new direction θi=θi+π/2mod2πat the border [10].
The RD Charlie conducts the same mobile pattern as Bob in
cell 2.
RWP Bob and RD Charlie: Bob moves according to RWP
model in cell 1 and Charlie moves according to RD model in
cell 2.
The RWP and RD models can interpret the basic properties
of random mobile users and many real-world mobility patterns
can be imitated by RWP and RD, such as vehicles and
unmanned aerial vehicles. Though the real humans mobility
behavior sometimes can be more complicated than RWP and
RD, the tractable spatial distributions of RWP and RD can
provide us more insights about NOMA secrecy under mobility.
Eavesdropping model: Assuming Alice employs a secrecy
protect zone [4] around him with radius RG
Eto guarantee Eve’s
distance dE≥RG
Eto her. Considering the typical downlink
connectivity where Alice transmits secrecy massages to Bob
and Charlie. For the worst case consideration for secrecy,
assuming Eve knows Alice’s location and RG
E, then she can
always keep eavesdropping at the boundary of the guard zone
with a distance dE=RG
E(Eve can be static or move around
the guard zone). This model is illustrated in Fig. 1.
B. Mathematical problem formulation
For users moving in a continuous time period T,we
discretize Tto time slots ti∈(t1,t
2, ..., tN),i =1,2, ..., N .
When tiis very small, this discretization can fully approximate
the continuous case. At each time ti, Alice employs NOMA
to communicate with Bob and Charlie simultaneously with
power PT,i. Assume Eve always keeps eavesdropping with a
minimum distance of dE,i =RG
Eto Alice. At time slot ti, the
instant channel gain (ICG) of Bob, Charlie, and Eve can be
denoted by as:
hB,i =|˜
hB,i|2
da
B,i
,h
C,i =|˜
hC,i|2
da
C,i
,h
E,i =|˜
hE,i|2
(RG
E)a,(1)
where ˜
hB,i,˜
hC,i and ˜
hE,i are time-varying complex channel
fading coefficients of Alice to Bob, Alice to Charlie, and
Alice to Eve channels at ti, respectively. dB,i,dC,i are
the instantaneous distances from Alice to Bob, and Alice
to Charlie at ti, respectively. ais the path loss coefficient
(a≥2) depends on the propagation environments [11]. For
each ti, based on NOMA protocol, Alice will allocate ratio
ρ(0 <ρ<1) of total power PT,i for a user with higher ICG
(good user) and 1−ρfor the other user with lower ICG (poor
user). Assume Bob and Charlie return ICG to Alice at each
time slot ti, thus there are two situations below.
Situation 1: Bob has higher ICG than Charlie at ti
where hB,i ≥hC,i. Thus Bob conducts SIC (Successive
Interference Cancellation) [3] to decode his message. Let
rB,i,rC,i denote the instantaneous received SNRs at Bob and
Charlie, respectively, then
rB,i =ρPT,ihB,i
σ2
B
,r
C,i =(1 −ρ)PT,ihC,i
ρPT,ihC,i +σ2
B
.(2)
Without loss of generality, let σ2
B=σ2
C=σ2
E=σ2, then
rB,i =ηρhB,i,r
C,i =(1 −ρ)hC,i
ρhC,i +η,(3)
where η=PT,i/σ2denotes the SNR. At the same time slot,
the upper bound of Eve’s instantaneous received SNRs for
overhead massages to Bob and Charlie can be denoted as:
wB,i ≤ηρhE,i,w
C,i ≤η(1 −ρ)hE,i (4)
Situation 2: Bob has worse ICG than Charlie where hB,i ≤
hC,i. Then Charlie conducts SIC to decode his message.

Similarly, the instantaneous received SNRs rB,i and rC,i for
Bob and Charlie can be written as
rB,i =(1 −ρ)hB,i
ρhB,i +η,r
C,i =ηρhC,i,(5)
The upper bound of Eve’s instantaneous received SNRs for
massages overhead to Bob and Charlie can be denoted as:
wB,i ≤η(1 −ρ)hE,i,w
C,i ≤ηρhE,i (6)
The secrecy rate [1] is defined as the maximum achievable
transmit bit rate with perfect secrecy per channel use. Thus
for ti, the lower bound of instantaneous secrecy rate of Bob
and Charlie can be represented by
Cj,i
S= [log(1 + rj,i (dj,i)) −log(1 + wj,i ((rE))]+(7)
where j=(B,C)are subscript abbreviations of Bob and
Charlie, [x]+=max(0,x).
Average secrecy rate: For classic physical layer secrecy
with static nodes, the distance dj,i keeps constant. However,
in our work, Bob and Charlie keep moving randomly thus
dB,i,dC,i are time-varying, so as to rB,i,rC,i,wB,i and wC,i.
Thus during T, the lower bound of average secrecy rate ¯
Cj
S
for Bob and Charlie can be written as
¯
Cj
S=1
N
N

i=1
Cj,i
S(8)
The average secrecy rate in (8) measures the ergodic secrecy
rate for the different random mobile users.
Secrecy fairness: Considering the secrecy performance gap
of Bob and Charlie, we define ΔCFto measure the absolute
value of average secrecy rate gap between mobile users Bob
and Charlie as:
ΔCF=1
N
N

i=1 CB,i
S−CC,i
S(9)
From (9), a lower bound of secrecy fairness can be written as
ΔCF≥1
NN

i=1 CB,i
S−
N

i=1 CC,i
S=¯
CB
S−¯
CC
S(10)
A smaller ΔCFmeans Bob and Charlie can achieve similar
secrecy rate. More detailed analysis about (8) will be provided
in the next section.
III. SECRECY ANALYSIS FOR NOMA MOBILE USERS
A. Spatial distance PDF of mobile users
After moving for an initial time, the mobile user will run
into steady state [9]. In steady state, the spatial node distance
probability density distribution function (SPDF) f(d)[9]
illustrates the probability density of a dynamic user appearing
in a spatial position with distance dfrom the central point. For
RWP mobility users in a circle with radius R, a well known
SPDF is given by C.Bettstetter [9] (CB) as:
fCB(d)= 4d
R2−4d3
R4,(0 <d≤R)(11)
The SPDF in (11) has been verified to be very close to the real
RWP mobility process [9]. For RD users, work [10] has proved
that the SPDF is approaching to the uniform distribution in
the circle region. Thus, for steady state, the RD user has
the same SPDF with mean static (MS) [2] users (the static
user appear in the circle region uniformly). By using the
polar coordinate integral, the SPDF for both RD and MS
Bob in the circular area with unit radius can be derived
as fRD(d)=fMS (d)=2π
0
d
πR2dθ|R=1 =2d. For the
simplicity of representation, we normalize distance variables
by making the following substitutions as (0 <m
j,i,λ ≤1):
mj,i =da
j,i
Ra
v
,λ=(RG
E)a
Ra
2
(12)
where v=1,2denotes the cell 1 and 2, respectively. There-
fore, when j=B, v =1, and when j=C, v =2. Then,
the SPDF with a path loss order a(i.e., PDF of mj,i =da
j,i)
can be written as follows:
fm(m)=fd(¯
da)=fd(m1
a)
∂(m1
a)
∂m 
(13)
From (9) and (12), we can get SPDF of CB and RD (MS)
with variable m(0 <m≤1) as:
fCB(m)= 4
am2
a−1−m4
a−1,f
RD(m)= 2
am2
a−1(14)
B. Average secrecy rate of mobile users
Let hB,h
C,h
Erepresent random variables corresponding
to hB,i,h
C,i,h
E,i, based on the complete probability formula,
the average secrecy rate in (8) can be derived in (15) at the bot-
tom of this page. In (15), the parameter j(k)=B(C)or C(B)
represents combination of Bob and Charlie. Cj
S(mj,m
k|hj≥
hk)denotes the condition secrecy rate when hj≥hk. The
probability Pr(hj≥hk|mj,m
k)denotes the conditional
probability under given distances mj,m
k.f(mj),f(mk)are
SPDF of Bob and Charlie with variable m, which can refer to
(14). Based on (15), we can numerically calculate the average
secrecy rates of Bob and Charlie. For Rayleigh fading channel,
based on work [1], Pr(hj≥hk|mj,m
k)can be written as
Pr(hj≥hk|mj,m
k)= 1
1+mj/mk
(16)
¯
Cj(k)
S=1
01
0
[Cj
S(mj,m
k|hj≥hk)Pr(hj≥hk|mj,m
k)+Cj
S(mj,m
k|hj<h
k)Pr(hj≤hk|mj,m
k)]f(mj)f(mk)dmjdmk
(15)

Then Cj
S(mj,m
k|hj≥hk)can be written as [1]
Cj
S(mj,m
k|hj≥hk)=F(ηρ
mjRa
v
)−F(ηρ
mjRa
v+λRa
2
)
(17)
Under Rayleigh fading channel, F(x)can be written
as F(x)=∞
0
1
xlog2(1 + u)e(−u
x)du [1]. Similarly,
Cj
S(mj,m
k|hj<h
k)in (15) can be written as:
Cj
S(mj,m
k|hj<h
k)=∞
0∞
0
[log(1 + (1 −ρ)rj
ρrj+η)−
log(1 + η(1 −ρ)w)]+f(rj)f(w)drdw
(18)
where f(rj),j =(B,C)and f(w)are PDF of rj=|hj|2and
w=|hE|2under given mj,m
k, respectively. For Rayleigh
fading channel, f(rj)and f(w)under given mj,m
kfollow
the exponential distribution [11], which can be written as
f(rj)=Ra
vmjeRa
vmjrj,f(w)=Ra
2λeRa
2λw (19)
Taking (16)-(19) to (15), we can get the average secrecy rates
of Bob and Charlie in (8), respectively.
IV. POWER ALLOCATION UNDER AVERAGE POWER
LIMITS AND QOSCONSTRAINTS
In this section,we propose a novel power allocation strategy
for NOMA mobile users. Different from [7], we maximize the
average sum secrecy rates of mobile users which are subjected
to average power limits and QoS requirements.
A. Average secrecy rate maximization problem definition
For each time slot ti, considering Alice transmits with
power PT,i(1 ≤i≤N)corresponding to ti. At each time
slots, Bob and Charlie return hB,i and hC,i to Alice and Alice
treated the one with higher ICG as good user, and the other
one as the poor user:
hg,i =max{hB,i,h
C,i},h
p,i =min{hB,i,h
C,i}(20)
Alice allocates power (P1,i,P
2,i)for the good and poor user at
ti, respectively. Then the power allocation subjected to average
power limits and QoS constraints can be characterized as:
max
(P1,i,P2,i )
1
N
N

i=1
Cg
S,i +Cp
S,i (21)
s.t. 1
N
N

i=1
PT,i ≤Pmax (22)
PT,i =P1,i +P2,i ≤Ppeak,i (23)
RB
g,i ≥RB
1,i;(24)
RB
p,i +βi≥RB
2,i (25)
where PT,i is the total transmit power at time ti, which is
limited by the instantaneous power limit Ppeak,i (Ppeak,i ≥
Pmax). Pmax is the average power limit during time T.
RB
1,i and RB
2,i are instantaneous Qos required for good and
Algorithm 1 The threshold power allocation strategy
Input: Pmax,PS,P
peak,i,R
B
1,i,R
B
2,i
Output: P1,i,P
2,i
For ti,(i≤N, i ++)
If PS≤Ppeak,i
Calculating (P∗
1,i,P∗
2,i)and ΔPi=PS−P∗
1,i −P∗
2,i
Else
Calculating (P∗
1,i,P∗
2,i)and ΔPi=Ppeak,i −P∗
1,i −P∗
2,i
End If
If ΔPi≥0
P1,i =P∗
1,i +ΔPi,P
2,i =P∗
2,i
Else Declare outage
End If
End For
poor users at ti.βiis the rate loss on poor user which is
related to the power strategy of Alice. Cg
S,i +Cp
S,i denotes the
instantaneous sum secrecy rate of good and poor user at ti.
RB
g,i and RB
p,i are the channel capacity of good user and poor
user at time ti. Let σ2
B=σ2
C=σ2
E=σ2=1, then
RB
g,i(P1,i)=log(1+P1,ihg,i)(26)
RB
p,i(P2,i )=log(1+ P2,ihp,i
P1,ihp,i +1)(27)
However, because Alice is unknown hE,i, and even unknown
the statistical information of hE,i. Thus the average secrecy
rate maximization problems in (21) cannot be easily solved.
However, we can propose a threshold power allocation strategy
to improve the sum secrecy rate of NOMA mobile users below.
B. The threshold power allocation strategy
Considering for mobile users, the power allocation strategy
should be realized with low complexity, because the channel
is time varying and the secrecy performance will be very
sensitive to the processing delay. Here we propose a threshold
strategy for Alice when she is completely unknown Eve’s
ICG. In each time slot ti, Alice transmits signal if hg,i ≥τ.
Else if hg,i <τ, Alice suspends the communication (Alice
can reserve transmit power of this time slot and increase the
transmit power in other time slots) and inform Bob and Charlie
a suspension. Considering the average power limit in (22),
Alice can calculate statistic information Pr(hg≥τ)and let
PS=Pmax/P r (hg≥τ). According to [7], allocating power
to the user with good instantaneous channel conditions can
improve the sum secrecy rate most. Therefore, at each time
slot, Alice can improve the good user’s secrecy rate as much
as possible. The proposed strategy is shown in Algorithm 1.
Firstly, let βi=0and assuming P∗
1,i,P∗
2,i are the minimum
power which satisfies (23), (24) and (25). From (26), (27), we
can get
P∗
1,i ≥2RB
1,i −1
hg,i
,P∗
2,i ≥(2RB
2,i −1)(P∗
1,i +h−1
p,i )(28)

Figure 2: The sum average secrecy rates of NOMA mobile users,
comparing to static users.
At each ti, Alice employs all remain power ΔPi=PS−
P∗
1,i −P∗
2,i for good user, then the instantaneous secrecy rate
increases at good user which can be written as
ΔCS
g,i = log( 1+(P∗
1,i +ΔPi)hg,i
1+(P∗
1,i +ΔPi)hE,i
)−log( 1+P∗
1,ihg,i
1+P∗
1,ihE,i
)
(29)
However, Alice adding power at good user can decrease the
channel rate at poor user as:
βi= log(1 + P∗
2,ihp,i
P∗
1,ihp,i +1)−log(1 + P∗
2,ihp,i
(P∗
1,i +ΔPi)hp,i +1)
(30)
Thus the increase of instantaneous sum secrecy rates can
be denoted by ΔCS
i=ΔCS
g,i(ΔPi)−ΔCS
p,i(ΔPi), where
ΔCS
p,i(ΔPi)denotes the secrecy rate loss at poor user as
ΔCS
p,i(ΔPi)=βi+ log(1 + ΔPihE,i
1+P∗
2,ihE,i
)(31)
From (10), we can also define another kind of secrecy fairness
which can measure the secrecy rates gap for good and poor
users as
ΔCF,2=1
N
N

i=1 Cg,i
S−Cp,i
S≥1
NN

i=1 Cg,i
S−
N

i=1 Cp,i
S
(32)
Discuss: If Alice employs all the remain power to poor user,
the good user will not be interfered. However, without known
hE,i,ΔCS
imay be negative. It can be further illustrated
in Fig. 5 and Fig. 6. Thus for maximum sum secrecy rate
consideration, Alice should employ remain power for good
user when she is completely unknown Eve’s ICG.
V. N UMERICAL RESULTS AND DISCUSSION
This section will illustrate secrecy performances of NOMA
mobile users by numerical simulations. Figs. 2-4 show the
numerical results of the average secrecy rates and fairness of
Figure 3: The average secure rates of Bob and Charlie.
Figure 4: The average secrecy rates and fairness of good and poor
users.
Bob and Charlie. We simulated 4 types of mobility combina-
tions for comparing. The abbreviation 2RWP denotes Bob and
Charlie are both RWP mobile users, RWPRD denotes Bob is
RWP and Charlie is RD user, 2RD denotes Bob and Charlie
are RD users, and 2MS means Bob and Charlie are static and
appear in the cell 1 and 2 uniformly. The parameters in Figs. 2-
4 are set as R1=R2=1,a=4,ρ=0.5with SNR=1 and 10.
From Fig. 2, it can be seen that all of the sum average secrecy
rates increase with protect zone radius increased. However,
the sum average secrecy rates of 2RWP users are the highest
among all mobility combinations (higher than others about 10-
20% for most protect zone radius). Besides, the performance
of RWPRD users are always higher than 2RD users, 2RD users
have the same performance with 2MS users. This phenomenon
can be fully illustrated by Figs. 3 and 4 below.
Fig. 3 shows the average secrecy rates between Bob and
Charlie with secrecy fairness performances of (10) under 4
types of mobility combinations. From Fig. 3, we can see that
Bob and Charlie can achieve similar secrecy rates if they have
the same mobility pattern, which results in a good secrecy
fairness. By the contrary, RWPRD users have much bigger

Figure 5: The sum average secrecy rates under average power limits
and Qos constraints.
Figure 6: The secrecy rates increase for good and poor user.
secrecy rate gap than other combinations, which results in a
poor fairness performance. Fig. 4 shows average secrecy rates
of good and poor users with secrecy fairness performances
in (32) (denoted by term ’good-poor’). We can see that even
the poor user of all mobility combinations can achieve much
lower secrecy rates, 2RWP good user can achieve much higher
secrecy rates than RWPRD and 2RD users, which leads the
highest sum secrecy rates of 2RWP users in Fig. 2.
Fig. 5 shows the sum average secrecy rates under average
power limits and Qos constraints. The parameters are set as
RB
1,i =0.5,R
B
2,i =0.1,R1=R2=10,a =2,τ =1/64.
The average power limit Pmax varies from 0 to 35 dB. The
abbreviation ONF denotes the proposed threshold strategy in
Algorithm 1. DP good and DP poor denote that Alice directly
employs ΔPifor good and poor users at each ti. Fig. 5 shows
that all the performances improve with Pmax increased. DP
good scenario can get much better performances than DP poor
scenario. The proposed ONF strategy can improve secrecy
rates effectively under all types of mobility combinations. Fig.
6 shows the sum secrecy rates increase when Alice employs
ΔPifor the good and poor users under average power limits
and QoS constraints. It can be seen that for all mobility
combinations, Alice adding ΔPifor poor user can hardly
contribute to sum secrecy rates (close to 0). The main reason
is that the poor user’s ICG may be lower than Eve with high
probability. If Alice uses more power for poor user, the sum
secrecy rate will decrease inversely. Therefore, Alice should
put ΔPifor good user at each time slot when she unknown
about Eve.
VI. CONCLUSION
This work investigated the impact of random mobility
on physical layer security of two NOMA access users. We
studied the secrecy performance of users with combinations
of RWP and RD mobility models and found that RWP users
can achieve the highest sum average secrecy rate among
the static and other mobile users. Then we proposed a
sum average secrecy rate maximization problem, subjected
to average power limits and users’ QoS requirements. In the
future, we would like to conduct more real-world experiments
to demonstrate the practicality of the proposed models and
analysis. Meanwhile, the scenarios where the NOMA systems
has more than two mobile users will be fully investigated.
REFERENCES
[1] M. Bloch, J. Barros, M. R. D. Rodrigues, and S. W. McLaugh-
lin, “Wireless information-theoretic security,” IEEE Transac-
tions on Information Theory, vol. 54, no. 6, pp. 2515–2534,
Jun. 2008.
[2] X. X. L. Tao, Z. Yan and Z. Shidong, “Mean physical-layer
secrecy capacity in mobile communication systems,” Journal
of Tsinghua University, vol. 55, no. 11, pp. 1241–1245, 2015.
[3] L. Dai, B. Wang, Y. Yuan, S. Han, C. l. I, and Z. Wang, “Non-
orthogonal multiple access for 5g: solutions, challenges, op-
portunities, and future research trends,” IEEE Communications
Magazine, vol. 53, no. 9, pp. 74–81, September 2015.
[4] Y. Liu, Z. Qin, M. Elkashlan, Y. Gao, and L. Hanzo, “En-
hancing the physical layer security of non-orthogonal multiple
access in large-scale networks,” IEEE Transactions on Wireless
Communications, vol. 16, no. 3, pp. 1656–1672, March 2017.
[5] B. He, A. Liu, N. Yang, and V. K. N. Lau, “On the design of
secure non-orthogonal multiple access systems,” IEEE Journal
on Selected Areas in Communications, vol. 35, no. 10, pp.
2196–2206, Oct 2017.
[6] J. Chen, L. Yang, and M. S. Alouini, “Physical layer security
for cooperative noma systems,” IEEE Transactions on Vehicular
Technology, vol. PP, no. 99, pp. 1–1, 2018.
[7] Y. Zhang, H. M. Wang, Q. Yang, and Z. Ding, “Secrecy sum
rate maximization in non-orthogonal multiple access,” IEEE
Communications Letters, vol. 20, no. 5, pp. 930–933, May
2016.
[8] K. Govindan, K. Zeng, and P. Mohapatra, “Probability density
of the received power in mobile networks,” IEEE Transactions
on Wireless Communications, vol. 10, no. 11, pp. 3613–3619,
Nov. 2011.
[9] C. Bettstetter and C. Wagner, “The spatial node distribution
of the random waypoint mobility model.” WMAN, vol. 11, pp.
41–58, 2002.
[10] P. Nain, D. Towsley, B. Liu, and Z. Liu, “Properties of random
direction models,” in IEEE INFOCOM, vol. 3, Mar. 2005, pp.
1897–1907 vol. 3.
[11] A. Goldsmith, Wireless communications. Cambridge university
press, 2005.