Secured Scheme for RF Energy Harvesting Mobile Edge Computing Networks based on NOMA and Access Point Selection

Full text

Secured Scheme for RF Energy Harvesting Mobile
Edge Computing Networks based on NOMA and
Access Point Selection
Van-Truong Truong
Faculty of Electrical-Electronic Engineering
Duy Tan University
Da Nang, 550000, Vietnam
truongvantruong@dtu.edu.vn
Dac-Binh Ha
Faculty of Electrical-Electronic Engineering
Duy Tan University
Da Nang, 550000, Vietnam
hadacbinh@duytan.edu.vn
Abstract—In this paper, we study an RF energy harvesting
mobile edge computing network where an energy-constrained
user harvests the RF energy from the power station and of-
floading its tasks to two access points based on non-orthogonal
multiple access (NOMA). In particular, the user can offload
its confidential tasks to the trusted access point, and non-
confidential tasks can be offload to both trusted and untrusted
access points by using the harvested energy. We propose two
protocols based on NOMA, non-access point selection (NAPS),
and access point selection (APS) schemes, namely NOMA-NAPS
and NOMA-APS. The closed-form exact expressions of successful
computation probability (SCP) are derived to evaluate these
proposed protocols’ system performance. Numerical results are
shown that the NOMA-APS protocol outperforms the NOMA-
NAPS one. The Monte-Carlo simulation results also verify the
correctness of our analysis.
Index Terms—mobile edge computing, non-orthogonal multiple
access, access point selection, successful computation probability.
I. INTRODUCTION
With the rapid development of the Internet of Things (IoT),
many new applications and services with big data have been
inspired in practice. These new services and applications
contain intensive computation, which introduces a challenge to
system design, specifically in resource-constrained networks,
e.g., wireless sensor networks. In this kind of networks, the
computation ability of nodes is limited. To solve this challenge,
mobile edge computing (MEC) technique has been proposed,
in which the function of servers or access points moves
towards the network edge to support the intensive computation
needs of nodes [1]–[6].
Meantime, due to the energy-constrained battery and diver-
sity functions of wireless devices, the energy problem is also
another challenge on the device design. Several new energy
harvesting techniques have been proposed and deployed to
power the users, e.g., radio frequency energy, solar energy,
wind energy, thermal energy, magnetic energy, and so on. The
radio frequency energy harvesting (RF EH) is an emerging
technique that can prolong mobile devices’ lifetime and main-
tain the coverage of wireless networks [7]. The prior research
results have shown that the user computation performance can
be improved by integrating the RF EH technique into MEC
networks [3], [8]. Besides RF EH, the NOMA technique has
been recognized as a promising solution for future wireless
networks due to its ability to serve multiuser by using the
same time and frequency resources [9]–[11].
The combination of RF EH and NOMA techniques in MEC
networks is proposed in some prior works to improve the sys-
tem performance [8], [12]–[14]. In [12], three different modes
of the offloading scheme, namely the partial computation
offloading, the complete local computation, and the complete
offloading, were proposed for a NOMA MEC network. The
optimal solutions for an optimization problem to maximize
the successful computation probability were proposed by
jointly optimizing this considered scheme’s parameters. The
authors in [13] proposed a computation efficiency maximiza-
tion framework for wireless-powered MEC networks based on
uplink NOMA with partial and binary computation offloading
modes. The iterative and alternative optimization algorithms
were proposed to solve the computation efficiency non-convex
problem. In the work of [8], the efficient algorithms were
proposed to solve the weighted sum-energy minimization
problems under both cases with partial and binary offloading
for multi-antenna NOMA multiuser MEC system. The NOMA
MEC networks for both uplink and downlink transmissions
were studied in [14]. It is shown that the deployment of
NOMA can efficiently reduce the latency and energy con-
sumption of MEC offloading compared to their conventional
orthogonal multiple access approaches.
In this work, we consider the scenario that a user harvests
the RF energy from the power station and partially offloads
its tasks to two MEC access points, in which the confidential
task must be offloaded to the trusted access point by applying
downlink NOMA scheme. The main contributions of our paper
are as follows.
•We propose two quadra-phase protocols for the con-
sidered mobile edge computing system based on the
downlink NOMA scheme, namely NOMA-NAPS and
NOMA-APS.
•Accordingly, we derive the closed-form exact expressions

of successful computation probability for these protocols
to evaluate the system performance.
•Numerical results are provided to investigate the impacts
of key parameters, i.e., transmit power, time switching ra-
tio, and power allocation ratio on the system performance
to verify these proposed protocols’ effectiveness.
The remainder of this paper is organized as follows. The
system model is described in Section II. Section III presents
the detailed derivations of the closed-form exact expression
of the successful computation probability used to analyze the
considered system performance. Section IV provides numeri-
cal results and discussions. Finally, Section V concludes the
paper.
II. SY ST EM MO DE L DESCRIPTION
We define the notations adopted in the next part of this work
in Table I.
TABLE I
NOTATIO NS
Notation Meaning
g0, g1, g2Channel power gains of the links P-U,U-AP1,U-
AP2
λiFading parameter of the links P-U,U-AP1,U-AP2
LLength of the task
L1Length of the confidential task
L2Length of the non-confidential task
αTime switching ratio
TTransmission block time
ηEnergy conversion efficiency
P0Transmit power of the power station
γ0Average transmit SNR
PTTransmit power of the user
ciNumber of required CPU cycles for each bit of APi
fiCPU-cycle frequency at the APi
bPower allocation ratio
BBandwidth
C1Channel capacity of link U-AP1
C2Channel capacity of link U-AP2
P rsSuccessful computation probability
Fig. 1 describes an RF energy harvesting mobile edge
computing network where an energy-constrained user (U)
harvests RF energy from a power station (P) and offloading
its tasks to two access points (APs). Due to the constraint of
latency requirement and its limited computation ability, Umay
not execute its tasks locally. Hence, it offloads its tasks to MEC
servers located at APs, which have a more strong computation
ability. In particular, Umust offload its confidential tasks to
the trusted AP, denote as AP1, and the non-confidential
tasks can be offloaded to both trusted and untrusted access
points, denote as AP2. Assuming that Uhas a number of
tasks with the same length of Lbits to be executed, each
task consists of L1-bit confidential subtask and L2-bit non-
confidential subtask (L1+L2=L) [15]–[17]. Thus, the L1-
bit confidential subtasks are designed to offloaded to AP1
to ensure the security requirement of this considered system.
Moreover, due to the limited battery problem, Uoffloads its
tasks by using all energy harvested from Pand by applying the
downlink NOMA scheme. Finally, APs return the computing
Fig. 1. System model for the RF EH NOMA MEC network
Fig. 2. Time flowchart of the considered RF EH NOMA MEC system
results to Uby uplink NOMA scheme. The entire protocol
can divide into four phases as the time flowchart depicted in
Fig. 2.
Assuming that all the channels have block Rayleigh fading,
i.e., the channel power gain is constant over each block
but vary independently between different blocks and follows
Rayleigh distribution. We also assume that P,U, and APs
are single antenna devices and operate in half-duplex mode.
Let gi,i∈ {0,1,2}, denote the channel power gains of the
wireless links P-U,U-AP1and U-AP2, respectively
with gi∼ CN(0, σ2). In this work, we propose two different
protocols for this considered system as follows.
A. NOMA-NAPS protocol
In this subsection, we describe the NOMA-NAPS protocol
as follows.
•In the first phase (energy harvesting phase), Uharvests
energy from Pduring the duration of τ0=αT , where α
denotes the time switching ratio, i.e., 0< α < 1, and T
stands for transmission block time.
•In the second phase (offloading phase), Uoffloads L1-bit
confidential subtasks to AP1and L2-bit non-confidential
subtasks to AP2by applying downlink NOMA scheme
during duration τ1.
•In the third phase (computing phase), all offloaded tasks
are computed at the corresponding MEC APs during
duration τ2.
•In the last phase (result returning phase): After successful
computation, the APs feedback the computed results to
Uwithin τ3. Here, the computed results are small, and it
means that τ3is very small compared to τ0,τ1, as well
as τ2and thus, is neglected [12], [13].

In a summary manner, the NOMA-NAPS protocol is proposed
for Uas Algorithm 1.
Algorithm 1 NOMA-NAPS Algorithm for U
1: Uharvests energy in duration of τ0=αT
2: dividing L-bit task into L1-bit subtask (Task 1) and L2-bit
subtask (Task 2)
3: applying NOMA, Uoffloads Task 1 to AP1and Task 2
to AP2
4: waiting for the computed results
5: download results from corresponding APs
B. NOMA-APS protocol
In this subsection, we present the NOMA-APS protocol as
follows.
•The first phase (energy harvesting phase) is the same as
NOMA-NAPS protocol.
•In the second phase (offloading phase), if channel of link
U−AP1is better than of link U−AP2then Uoffloads
whole task to AP1, else Uoffloads L1-bit confidential
subtasks to AP1and L2-bit non-confidential subtasks to
AP2by applying downlink NOMA scheme.
•The third phase (computing phase) is the same as the
NOMA-NAPS protocol.
•In the last phase (result returning phase) is the same as
the NOMA-NAPS protocol.
In a summary manner, the NOMA-APS protocol is proposed
for Uas Algorithm 2.
Algorithm 2 NOMA-APS Algorithm for U
1: Uharvests energy in duration of τ0=αT
2: procedure SELE CT IO N(AP1,AP2)
3: if g1> g2then AP1selected
4: Uoffloads L-bit task to AP1
5: goto 11
6: else divide L-bit task into L1-bit subtask (Task 1) and
L2-bit subtask (Task 2)
7: applying NOMA, Uoffloads Task 1 to AP1and Task
2 to AP2
8: goto 11
9: end if
10: waiting for the computed results
11: download results from corresponding APs
During the energy harvesting phase, the harvested energy
of Uis obtained by
Eh=ηP0g0αT , (1)
where 0< η ≤1stands for the energy conversion efficiency
of the energy receiver [18], P0denotes the transmit power of
P. The transmit power PTcalculated as follows
PT=Eh
(1 −α)T−τ=aP0g0,(2)
where a∆
=ηαT
(1−α)T−τ,τ= max{τ21, τ22 }in which τ2iis the
computing time for Li-bit at APi,i∈ {1,2}.τ2iis defined as
follows:
τ2i=ciLi
fi
,(3)
where cistands for the number of required CPU cycles for
each bit of APi, and fidenotes the CPU-cycle frequency at
the APi,i∈ {1,2}.
During the offloading phase, when applying NOMA, U
transmits superimposed signal
x=√bs1+p(1 −b)s2,(4)
to APs, where s1and s2are the data messages corresponding
to L1-bit and L2-bit subtasks, respectively; bstands for the
power allocation coefficient. Due to the priority for the trusted
AP1, we assume that the transmit power is allocated to AP1
more than AP2, thus bis selected to satisfy the condition 0.5<
b < 1to apply NOMA scheme.
The i.i.d. quasi-static Rayleigh channel power gains gi,
i∈ {0,1,2}, follows exponential distributions with parameters
λi. Therefore, the cumulative density function (CDF) and
probability density function (PDF) of gi(i= 0,1,2) are
respectively given by
Fgi(x)=1−e−x
λi,(5)
fgi(x) = 1
λi
e−x
λi.(6)
III. PERFORMANCE ANALYSIS
In this section, we present the performance analysis of this
proposed MEC system in terms of successful computation
probability. The successful computation probability, namely
P rs, is defined as the probability that all tasks are successfully
executed within a given time T > 0. For this proposed system,
P rsis expressed as
P rs= Pr [τ1+τ≤(1 −α)T].(7)
In order to evaluate the performance of this considered MEC
system, we obtain the following theorems.
Theorem 1: Under quasi-static Rayleigh fading, the closed-
form exact expression of the successful computation proba-
bility P rN AP S
sfor this considered MEC system based on the
proposed NOMA-NAPS scheme is given by
P rN AP S
s=(0,if b < ρ
uK1(u),if b > ρ (8)
where u= 2r1
λ0β1
λ1+β2
λ2,β1=
2
L1
(1−α)BΩ1−1
aγ0"b−(1−b) 2
L1
(1−α)BΩ1−1!#,β2=2
L2
(1−α)BΩ2−1
a(1−b)γ0,
Ω1= (1 −α)T−c1L1
f1,Ω2= (1 −α)T−c2L2
f2,
ρ= 1 −1
2
L1
(1−α)Ω1B
, and Kν(.)is the modified Bessel
function of the second kind and νth order.
Proof. See in Appendix A.

Theorem 2: Under quasi-static Rayleigh fading, the closed-
form exact expression of the successful computation proba-
bility P rAP S
sfor this considered MEC system based on the
proposed NOMA-APS scheme is given by
P rAP S
s=




v0Kν(v0)−λ2
(λ1+λ2)v1Kν(v1),if b < ρ
v0Kν(v0)−λ2
(λ1+λ2)v1Kν(v1)+
v2Kν(v2)−λ1
(λ1+λ2)v3Kν(v3),if b > ρ
(9)
where v0= 2qβ
λ0λ1,v1= 2qβ(λ1+λ2)
λ0λ1λ2,v2=
2r1
λ0β1
λ1+β2
λ2,v3= 2qβ2(λ1+λ2)
λ0λ1λ2,β=2
L
(1−α)BΩ−1
aγ0,
Ω = (1 −α)T−c1L
f1.
Proof. See in Appendix B.
IV. NUMERICAL RESULTS AND DISCUSSION
In this section, we provide numerical results in terms
of the successful computation probability. The Monte-Carlo
simulations are also used to verify the analytical results. The
parameters to consider are introduced in Table II.
TABLE II
SIMULATION PARAMETERS
Parameters Notation Typical Values
Environment Rayleigh
Fading parameter λ0,λ1,λ21
Number of antennas of APs 1
Average transmit SNR γ00-30 dB
Energy conversion efficiency η0.75
CPU-cycle frequency of APsf1, f25 GHz
The number of CPU cycles for each bit c1,c210
Channel bandwidth B100 MHz
The threshold of latency T0.5s
The length of confidential task L132 Mbits
The length of non-confidential task L248 Mbits
In Figs. 3 and 4, we examine the effect of average transmit
SNR and time switching ratio on system performance in terms
of P rs. From these figures, we can see that when the transmit
power increases, P rsincreases. It means that the increase of
transmit power can improve system performance. Meanwhile,
the variation of P rsis not the same, specifically, when α
increases from 0 to 1, P rsincreases to the highest value and
then decreases to 0. It means that there exists an optimal value
of α, called α∗, making P rsto achieve the optimal value.
Similarly, in Figs. 5 and 6, we examine the effect of
average transmit SNR and power allocation ratio on system
performance in terms of P rs. From these figures, we also
see that the increase in transmit power can improve system
performance. Meanwhile, when the power allocation ratio b
increases from 0.5 to 1, P rsincreases to the highest value,
and then degrades. It means that there exists an optimal value
of b, called b∗, making P rsto achieve the optimal value.
From the above figures 3-6, we can conclude that the
NOMA-APS protocol outperforms the NOMA-NAPS one in
terms of successful computation probability. Finally, we also
Fig. 3. P rsvs. the average transmit SNR with different values of α
Fig. 4. P rsvs. the time switching ratio with different values of γ0
observe from these figures that the analysis and simulation
results are very good matching. This has verified the accuracy
of our analysis.
V. CONCLUSION
In this paper, we proposed two quadra-phase protocols,
namely NOMA-NAPS and NOMA-APS, for the RF energy
harvesting mobile edge computing system based on downlink
NOMA scheme. Accordingly, we have derived the closed-form
exact expressions of successful computation probability for
these protocols to evaluate the system performance. Finally,
the numerical results are provided to investigate the impacts
of key parameters, i.e., transmit power, time switching ratio,
and power allocation ratio on the system performance to verify
the effectiveness of these proposed protocols. The numerical
results have shown that the NOMA-APS protocol outperforms

Fig. 5. P rsvs. the average transmit SNR with different values of b
Fig. 6. P rsvs. the power allocation ratio with different values of γ0
the NOMA-NAPS one in terms of successful computation
probability.
APPENDIX A
PROO F OF TH EO RE M 1
This appendix provides the detailed proof for Theorem 1.
From equation (7), we derive the closed-form expression of
P rN AP S
sas (A-1) at the top of the next page. This ends our
proof of Theorem 1.
APPENDIX B
PROO F OF TH EO RE M 2
This appendix provides the detailed proof for Theorem 2.
P rAP S
sis calculated as follows.
P rAP S
s= Pr g1> g2, C > L
Ω
| {z }
P1
+ Pr g1< g2, C1>L1
Ω1
, C2>L2
Ω2
| {z }
P2
,
(10)
where Cdenotes the channel capacities of link U-AP1.P1
and P2are calculated as (B-1) at the top of the next page.
This ends our proof of Theorem 2.
REFERENCES
[1] Y. Mao, C. You, J. Zhang, K. Huang, and K. B. Letaief, “A survey
on mobile edge computing: The communication perspective,” IEEE
Communications Survey Tutorials, vol. 19, no. 4, pp. 2322–2358, 2017.
[2] S. Chen, H. Wen, J. Wu, A. Xu, Y. Jiang, H. Song, and Y. Chen,
“Radio frequency fingerprint-based intelligent mobile edge computing
for internet of things authentication,” Sensors, vol. 19, no. 16, p. 3610,
2019.
[3] F. Zhou, Y. Wu, R. Q. Hu, and Y. Qian, “Computation rate maximization
in UAV-enabled wireless powered mobile-edge computing systems,”
IEEE Journal on Selected Areas in Communications, vol. 36, no. 9,
pp. 1927–1941, 2018.
[4] H. Sun, F. Zhou, and R. Q. Hu, “Joint offloading and computation
energy efficiency maximization in a mobile edge computing system,”
IEEE Transactions on Vehicular Technology, vol. 68, no. 3, pp. 3052–
3056, 2019.
[5] Y. Zhang, X. Lan, Y. Li, L. Cai, and J. Pan, “Efficient computation
resource management in mobile edge-cloud computing,” IEEE Internet
of Things Journal, vol. 6, no. 2, pp. 3455–3466, 2019.
[6] Q. Li, J. Zhao, and Y. Gong, “Computation offloading and resource
allocation for mobile edge computing with multiple access points,” IET
Communications, vol. 13, no. 17, pp. 2668–2677, 2019.
[7] X. Lu, P. Wang, D. Niyato, D. I. Kim, and Z. Han, “Wireless networks
with RF energy harvesting: A contemporary survey,” IEEE Communi-
cations Surveys and Tutorials, 17(2), pp. 757–789, 2014.
[8] F. Wang and X. Zhang, “Dynamic computation offloading and resource
allocation over mobile edge computing networks with energy harvest-
ing capability,” in IEEE International Conference on Communications
(ICC), Kansas City, MO, USA, May 2018.
[9] D.-B. Ha and Q. S. Nguyen, “Outage performance of energy harvesting
DF relaying NOMA networks,” Mobile Networks and Applications,
2017.
[10] D.-D. Tran, H.-V. Tran, D.-B. Ha, and G. Kaddoum, “Cooperation in
NOMA networks under limited user-to-user communications: Solution
and analysis,” in IEEE Wireless Communications and Networking Con-
ference (WCNC), 15-18/4/2018, Barcelona, Spain.
[11] D.-D. Tran and D.-B. Ha, “Secrecy performance analysis of QoS-based
non-orthogonal multiple access networks over nakagami-m fading,” in
The International Conference on Recent Advances in Signal Processing,
Telecommunications and Computing (SigTelCom), HCMC, Vietnam,
2018.
[12] Y. Ye, G. Lu, R. Q. Hu, and L. Shi, “On the performance and optimiza-
tion for MEC networks using uplink NOMA,” in IEEE International
Conference on Communications Workshops (ICC Workshops), Shanghai,
China, 2019.
[13] F. Zhou, Y. Wu, R. Q. Hu, and Y. Qian, “Computation efficiency in
a wireless-powered mobile edge computing network with NOMA,” in
IEEE International Conference on Communications (ICC), Shanghai,
China, 20-24 May 2019, 2019.

P rN AP S
s= Pr L1
C1
+c1L1
f1≤(1 −α)T, L2
C2
+c2L2
f2≤(1 −α)T
= Pr (1 −α)Blog21 + abγ0g0g1
a(1 −b)γ0g0g1+ 1>L1
Ω1
,(1 −α)Blog2[1 + a(1 −b)γ0g0g2]>L2
Ω2
=(0,if b<ρ
R∞
0h1−Fg1β1
xih1−Fg2β2
xifg0(x)dy, if b>ρ
=


0,if b < ρ
2r1
λ0β1
λ1+β2
λ2K12r1
λ0β1
λ1+β2
λ2,if b > ρ
(A-1)
where C1and C2denote the channel capacities of link U-AP1and U-AP2, respectively, Bis the channel bandwidth,
γ0
∆
=P0
σ2is the average transmit signal-to-noise ratio (SNR).
P1= Pr g1> g2, g0g1>2L
(1−α)BΩ−1
aγ0!= Pr g1> g2, g1>β
g0=Z∞
0Z∞
β
y
Fg2(x)fg1(x)fg0(y)dxdy
(a)
= 2rβ
λ0λ1K1 2rβ
λ0λ1!−2λ2
(λ1+λ2)sβ(λ1+λ2)
λ0λ1λ2K1
2sβ(λ1+λ2)
λ0λ1λ2
.
P2=(0,if b<ρ
Pr g1< g2, g1>β1
g0, g2>β2
g0,if b>ρ
=(0,if b<ρ
R∞
0R∞
β2
yhFg1(x)−Fg1β1
yifg2(x)fg0(y)dxdy, if b>ρ
(b)
=


0,if b < ρ
2r1
λ0β1
λ1+β2
λ2K12r1
λ0β1
λ1+β2
λ2−2λ1
(λ1+λ2)qβ2(λ1+λ2)
λ0λ1λ2K12qβ2(λ1+λ2)
λ0λ1λ2,if b > ρ
(B-1)
where step (a) and (b) are obtained by substituting (5) and (6) into the integrals.
[14] Z. Ding, P. Fan, and H. V. Poor, “Impact of non-orthogonal multiple
access on the offloading of mobile edge computing,” IEEE Trans.
Commun., vol. 67, no. 1, pp. 375–390, 2019.
[15] Y. Ye, R. Q. Hu, G. Lu, and L. Shi, “Enhance latency-constrained
computation in mec networks using uplink noma,” IEEE Transactions
on Communications, vol. 68, no. 4, pp. 2409–2425, 2020.
[16] L. P. Qian, B. Shi, Y. Wu, B. Sun, and D. H. Tsang, “Noma-enabled
mobile edge computing for internet of things via joint communication
and computation resource allocations,” IEEE Internet of Things Journal,
vol. 7, no. 1, pp. 718–733, 2019.
[17] F. Fang, K. Wang, and Z. Ding, “Optimal task assignment and power
allocation for downlink noma mec networks,” in 2019 IEEE Globecom
Workshops (GC Wkshps). IEEE, 2019, pp. 1–6.
[18] D.-B. Ha, D.-D. Tran, V. Tran-Ha, and E.-K. Hong, “Performance
of amplify-and-forward relaying with wireless power transfer over
dissimilar channels,” Elektronika ir Elektrotechnika Journal, vol. 21, no.
5, pp. 90–95, 2015.