Enhancing Physical Layer Security for Cooperative Non-Orthogonal Multiple Access Networks with Artificial Noise

This paper does the study on the performance of the physical layer secrecy of nonorthogonal multiple access (NOMA) in downlink cooperative. The given system includes one source, multiple legitimate user pairs in the form of an eavesdropper. By applying the decode-and-forward (DF) scheme, a good user will take the information from the source to send it to the bad user in every pair, we assume that the eavesdropper will spend e ﬀ ort to decode the message from the bad user. To enhance the secrecy performance of given system, the artiﬁcial noise cooperative transmission scheme named ANCOTRAS is suggested. To assess the performance of the suggested scheme, we obtained the lower bound and exact closed-form expressions of secrecy outage probability by implementing statistical characteristics of signal-to-noise ratio (SNR) and signalto-interference-plus-noise ratio (SINR). Furthermore, the secrecy performance of given system is studied basing on key parameters (including the power allocation ratio), average transmit power and amount of user pair for verifying the suggested scheme. At the end, the accuracy of ﬁnal analytical outcome is reassured by using the Monte-Carlo simulation results.


Introduction
In the recent years, the portable devices with wireless features and integrated smart operation system such as smartphones, smart watches or smart home devices became an integral part of human society, all thanks to their flexibility and multi-function.It also leads us to the compromising booming of fifth-generation network (5G) as a replacement for 4G.The reason for this changes are because the growing number of wireless devices also go hand in hand with the need for low cost and low latency networks, which the 4G and earlier generation networks, using tradition orthogonal multiple access (OMA) such as FDMA, TDMA, and CDMA, cannot meet because of the limited channel resource and the spectral efficiency loss.To adapt to the requirements of 5G network, the non-orthogonal multiple access (NOMA) technique is the most promising one requirements [1][2][3][4][5].With the ability in serving multiple users in the same period of time, frequency or code, NOMA can make sure that the spectral efficiency can be maintained.Besides, the equality of user can be increased in NOMA in comparison with OMA.Furthermore, the cooperative relaying technique which can be applied to NOMA networks can enhance the effectiveness and widen the covered area of the wireless networks [6,7] and it also can attract the attention of other researchers on this topic [8][9][10][11][12].In specific, a cooperative NOMA transmission scheme is suggested in the work [8] to V.L. Nguyen et al. explore the previous information in NOMA system, in which the better channel condition helps the users decode the message from others.For user having poor connection, the given information takes the form of the relays to increase the reception reliability.The NOMA cooperative relaying system is studied in the work [9] which also suggested the most suitable relay selection (BRS) scheme.In the outcome, it could be seen that in the case of increasing number of relays, the NOMA-based BRS acquired more rate gain than the traditional BRS.The [10,11] study is NOMA cooperative relaying system with energy harvesting and [12] provided another communication protocol for cooperative NOMA system.In them, the writer give the conclusion that in the weak transmission conditions their protocols can deal with the lack of direct link between the paired users.
Moving to the eavesdropping issue, while the information security is becoming more and more vital in the information age because the value of our confidential information in using for illegal activities, our information become more vulnerable than ever before because of the broadcast nature of wireless communications.The activities of eavesdropping by wiretappers are hard to be discovered and convicted while information transmission between transceivers can be eavesdropped with not many efforts.Although there are some provided solutions such as RSA and DES but most of them are used at the higher layers such as application or network layers.In addition to it, they are also limited in error-free condition of the link between the transmitter and receiver and the ability of eavesdroppers are assumed in the low computing power level and lacking of efficient algorithms [13].In the given situation, the physical layer secrecy (PLS) is suggested to accomplish the secure transmission to improve the security level of wireless networks by using the dynamic characteristics of the wireless channels [14][15][16][17][18].The secrecy ability of NOMA network can be increased by applying this approach.But the involvement of successive interference cancellation (SIC) in NOMA technique can differentiate it from conventional OMA technique.In the recent time, the physical layer secrecy (PLS) of NOMA relaying networks [19][20][21][22][23] attracted a lot of works.For example, the [19] investigated the PLS of a downlink NOMA system with the assuming that the required users quality of service (QoS) to deploy NOMA and all channels undergo Nakagami-m fading.From the outcome, it's can be seen that, the secrecy performance of the overall communication process can be improved when the difference in the level of priority between legitimate users is low.The [20] analyzed the secrecy performance of a two-user downlink NOMA system.The researcher included two single-input single-output and multiple-input single-output systems which is different in transmit antenna selection (TAS).The secrecy outage probability (SOP) for different TAS schemes is also expressed in precise and estimated closed form.The outcome is that the SOP for the far user with fixed power allocation scheme drop when the transmit power increase, as the transmit power beyond the threshold, and, after that, it touch the floor with the rising of the interference from the near user.In [21], the authors studied the PLS of large-scale NOMA networks through the SOP.In this article, the locations of NOMA users and eavesdroppers were mapped by using stochastic geometry approaches.And the authors in [22] produced the artificial noise (AN) at the base station to increase the security of a beamformingaided multiple-antenna system.They suggested The secrecy beamforming scheme for multiple-input singleoutput non-orthogonal multiple access (MISO-NOMA) system, in which, confidential information of two NOMA assisted legal users was protected by using AN, so the system secrecy performance is enhanced.In [23], the PLS for cooperative NOMA systems including amplify-and-forward (AF) and decode-and-forward (DF) protocols is considered.The conclusion showed that AF and DF schemes reach approximately similar secrecy performance and this secrecy performance is separated from the channel conditions between the relay and the poor user.
Different from the works above, our work focusses on the PLS of downlink NOMA cooperative relay communication system in combination with AN to increase the PLS performance.To be specific, our major subject is the PLS performance of the NOMA system, in that system, the base station (BS) simultaneously delivers information to user pairs based on powerdomain, then, the information will be forwarded from the better user (user with better channel condition) to the worse user (users with poor channel conditions), assuming that only the worse users are eavesdropped.AN is applied in the BS to blur the eavesdropper to improve the performance of PLS.Our article contributes the ideas below: 1. Suggesting AN with cooperative transmission scheme (ANCOTRAS).
2. Deriving the lower bound and exact closed-form expressions of secrecy outage probability for each user and overall system.
3. Evaluating the PLS performance using tools of secrecy outage probability expressions.
4. Studying the above system behavior in distinct key parameters, including power allocation ratio, average transmit power and the number of user pairs.
Enhancing Physical Layer Security for Cooperative Non-Orthogonal Multiple Access Networks with Artificial Noise 5. Comparing between the PLS performance of ANCOTRAS scheme and non-AN scheme to identify the advantages of integrated AN NOMA cooperative system.
The rest of this paper is organized as follows.The system model is presented in Section II.Secrecy performance of the considered system is analyzed in Section III.The numerical results are shown in Section IV.Finally, Section V draws the conclusion of our paper.

Network and Channel Models
The Fig. 1 illustrated a A cooperative communication system for downlink NOMA.In which, the information is intended to be transmitted to M mobile user's represented as D i (i = 1 ≤ m < n ≤ M), by the source S i.e., base station, with the under the presence of an eavesdropper E. In this system, it is ok to split the M users to multiple pairs, such as {D m , D n } , m < n, to perform NOMA [8], and in the information signal will be exchanged (by forwarding information) between two paired users.This means m th user and the n th user are paired to deploy cooperative NOMA.Using the applying successive interference cancellation (SIC) to cancel the interference and detect the D m 's signal the better user, then, forward the information of the worse user D n .Without loss of generality, assuming that all the channel gains between S and users follow the order of

Phase 1
In this phase, the source S broadcasts information to the M users.The received signals at D m and at D n are respectively, where n SD m and n SD n are the AWGN with zero mean and variance The instantaneous SINR at the n th user to detect s n transmitted from S can be given by where γ = P S σ 2 is denoted as the average transmit SNR of Similarly, the instantaneous SINR at the m th user to detect s n transmitted from S can be written as where b 3 = At the same time, the received signal at E is given as follows where n SE is the AWGN with zero mean and variance σ 2 E .Due to assuming that the eavesdropper only tries to detect s n , therefore the instantaneous SINR at E is given by where b 5 =

Phase 2
In phase 2, D m uses the power P Dm to forward s n to D n and S and, at the same time, uses the power P S to broadcast an AN to users and eavesdropper.The instantaneous SINR at D n in the second phase is as follows: where Equally, the instantaneous SINR at E in this phase is given by where To identify the advantages of ANCOTRAS protocol, a study on the case of non-AN also is done.In this, the instantaneous SNR at D n in the phase 2 is as follows: Similarly, the instantaneous SNR at E in the last phase is given by Considering i.i.d.Rayleigh channels, the channel gains |h SDm | 2 , |h SDn | 2 , |h SE | 2 and |h mn | 2 follow exponential distributions with parameters λ SDm , λ SDn , λ SE and λ mn , respectively.In order statistics, the probability density function (PDF) and the cumulative distribution function (CDF) of U , where U ∈ {X 1 , Y 1 }, are respectively given by [24] where i ∈ {m, n} The PDF and CDF of V , where V ∈ (X 2 , Z 1 , Z 2 ) are respectively expressed as where λ ∈ {λ mn , λ SE , λ DmE }.
Adding more calculation, we derive CDFs and PDFs of γ s n SD m , γ SD n , γ SE , γ mn , γ D m E .Drawn from above outcomes, The CDF of γ SDn are calculated as below: Given that step , respectively.Equally, the CDF and PDF of γ SE are derived respectively as belows: The CDF of the γ mn is calculated as follows Similarly the CDF of the γ D m E is given by The PDF of the γ D m E is expreesed as follows:

Analyzing Secrecy Performance
This part analyzed secrecy performance is analyzed in term of secrecy outage probability (SOP).SOP is an crucial performance metric which is applied to describe the secrecy performance of a wireless communication system.In this paper, the secrecy performance is investigated in terms of SOP at S and at D m with the assumption that E wants to take out the message of D n .

Preliminaries
The instantaneous capacities of the legitimate channel and eavesdropper channel can be respectively defined by where The instantaneous secrecy capacity for S − D m , S − D n and D m D n are given by respectively.Here, for simplicity we assume B = 1Hz.SOP is defined as the probability that the instantaneous secrecy capacity falls below a predetermined secrecy rate threshold R S > 0, given by SOP = P r(C S < R S ).Notice that, in this considered system we only consider the case of the eavesdropper tries to hear the message of D n at S and at D m .In the next subsection, we present the calculation of the SOPs at S and at D m .

Secrecy outgage probability at S
The SOP at S of S − D n link (SOP 1 ) can be calculated as follows: Because the equation ( 28) is intractable to obtain the closed-form expression, only the lower bound of SOP 1 is obtained here.From the equation (3), γ SD n < a n a m can be seen, so the lower bound of SOP 1 should be as follows: (30)

Secrecy outgage probability at D m
In this situation, the secrecy outage event happens when s n canâĂŹt be noticed by D m or D m can detect s n but the secrecy capacity is under the secrecy threshold.As a result, the SOP at the D m can be calculated as below: -SOP at D m with AN.
Proposition 1.Under Rayleigh fading, the SOP of the link D m − D n with AN is given by where -SOP at D m without AN.
Proposition 2. Under Rayleigh fading, the SOP of the link D m − D n without AN is given by where Proof.See Appendix B.

Secrecy outgage probability of overall system
As a result of D n applying selection combining scheme, the instantaneous secrecy can be calculated as belows: The SOP of overall system is given by 36) Proposition 3.Under Rayleigh fading, the SOP of overall system with AN is given by Similarly, SOP of the overall system without AN is given by

Nummerical results and disscussion
In this section, the PLS performance of ANCOTRAS protocol for this considered cooperative NOMA system is analyzed by numerical results.And our analytical results is also verified using Monte-Carlo simulation results.The outcomes of SOP of the system at S is illustrated in Figs. 2 and Figs. 3. figures displayed that the average transmits SNR from S to user D n ( γ) then the SOP at the S reduces when the distance from S to the eavesdropper E(d SE ) is increased.That means there are increase in the PLS capability of this model.Besides, in these two results, when we increase the average transmit SNR from S to the E ( γE ), the SOP at S of the system also increases.

Secrecy outage probability at D m
Figs. 4 and Figs. 5 depict the results of the SOP of the system at D m .In these two results, the secrecy performance is analyzed in consistent with the changes of the parameters belows: γ, γDm and γE .In these figures, We can see that the SOP at D m reduces when γDm (the average transmit SNR from D m to D n ) increase.From Figs. 4, we can see that the secrecy performance can be improved if we use AN from S to D n ( γ at second phase).Similarily, in Figs. 5 it is clear that the average transmit SNR of AN is increased from S to E ( γE ), the secrecy performance is improved.However, introducing AN when the average transmit SNR is low ( γE = 10 dB), the security performance of the system will not be improved compared to the case of not using AN.To be more specific, in this case, we should use γE The survey results in Figs. 6 showed that the SOP at D m of the system reduce along with the increase in γDm and reduce in γDmE .This result settles once again that using AN the will increases the PLS performance of the system.Seeing the changes in the figures of users (M), which using AN, in figs.7, it is showed that the SOP at the D m of system decreases when M increases.Using he formula of SOP 3 , the conclusion is that, C S 3 increases when M increase so SOP 3 decreases.Particularly, the simulation results in figs.7 showed that with 3 sequential M values: 4,6 and 8, SOP 4 do not experience major change.So it can be seen that when the system according to the changes of M and γDm is

Secrecy outage probability overall system
Fig. 8 illustrates the SOP of the overall system.The result displays that the higher the average transmit SNR from D m to D n ( γDm ) is the lower SOP of the overall system is observed.It means that secrecy performance will rise.On another side, it can be seen that when the average transmit SNR from S to D n ( γ) rise, then SOP of the overall system with reduce, but, if γ is big enough (particularly when γ ≥ 20 (dB) in this figure) then SOP of the overall system with AN will rise.In phase 2, γ is the average transmit SNR of the AN, so if γ is greater than a appropriate value (particularly when γ ≥ 10 (dB) in this figure) the security performance of the system with AN will reduce.We can see the SOP of the overall system with AN reduced when increasing the average transmit SNR from S to E ( γE ) in Fig. 9.In this figure, the increase of γE go hand in hand with the decrease of SOP.However, γE have to be at suitable (particularly when γE ≥ 10 (dB) in this figure), the security performance of the system with AN better than that the security performance of the system without AN.

Conclusion
In this work, we inspected the secrecy performance of a downlink cooperative NOMA network with artificial noise.Particularly, in this paper, the method of broadcasting artificial noise from the base station is used to interfere the eavesdroppers and we also improved the information security.Besides, new analytical expressions were carried out in terms of the secrecy outage probability to regulate the system secrecy performance.In the meantime,we used the numerical results to confirm the examines.Basing on the analytical and simulation results, it can be settled that the security level of the network model which is suggested in this paper relies on the distance: from 1) base station to the better user; worse users; eavesdropping device and from 2) better users to worse users, eavesdropping devices.Using artificial noise can improve the security performance of the system.As long as appropriate values are used, the securecy performance of the system with AN is better than that without AN.APPENDIX A: Here, we derive the expression of SOP at D m in the case of with AN.
where γ t is the threshold to detect s n and Φ 1 , Φ 2 , Φ 3 are calculated as follows

Figure 1 .
Figure 1.System model for secure cooperative NOMA denoted as the ordered channel gains of the m th user and the n th user.Denote that |h mn | 2 is the channel gains of the links between the m th user and the n th user; |h SE | 2 and |h DmE | 2 are channel gains of the links of S − E and D m − E, respectively.We assume that all the nodes are single-antenna devices and operate in a halfduplex mode.All wireless links are assumed to undergo independent frequency non-selective Rayleigh block fading and additive white Gaussian noise (AWGN) with zero mean and the same variance σ 2 .We denote d SD m , d SD n , d mn , d SE , d DmE as the Euclidean distances of S − D m , S − D n , D m − D n , S − E, D m − E, respectively and θ denote the path-loss exponent.

1 , x < b 4 b 3
(a) and (b) are gained by assuming the following condition holds x < b 2 b