Reliable and Secure Transmission in Multiple Antennas Hybrid Satellite-Terrestrial Cognitive Networks Relying on NOMA

We study a hybrid satellite-terrestrial cognitive network (HSTCN) relying on non-orthogonal multiple access (NOMA) interconnecting a satellite and multiple terrestrial nodes. In this scenario, the long distance communication is achieved by the satellite equipped multiple antennas to send information to a multi-antenna destinations through the base station acting as relay. The secure performance is necessary to study by exploiting the appearance of an eavesdropper attempting to intercept the transmissions from relay to destinations. We explore situation of hardware imperfections in secondary network and deign of multiple antennas need be investigated in term of the physical-layer security by adopting the decode-and-forward (DF) relay strategy. Specifically, we guarantee coverage area by enabling relaying scheme and keep outage probability (OP) performance satisfying required data rates. Moreover, suppose that only the main channels’ state information is known while the wiretap channels’ state information is unavailable due to the passive eavesdropper, we analyze the secrecy performance in term of intercept probability (IP) of the HSTCN by driving the closed-form expressions of such performance metric. Finally, the presented simulation results show that: 1) The outage behaviors of NOMA-based HSTCN network does not depend on transmit signal to noise ratio (SNR) at source at high SNR; 2) Numerical results show that the such system using higher number of transceiver antennas generally outperform the system with less antennas in terms of OP and IP and reasonable selection of parameters is necessary to remain the secrecy performance of such systems; and 3) By allocating different power levels to tow users, the second user has better secure behavior compared with the first user regardless of other set of satellite links or the number of antennas, which means that the superiority of the second user compared with user the first user in terms of OP and IP are same.


I. INTRODUCTION
To provide advances such as navigation assistance, ubiquitous large coverage and fast services in disaster areas, land mobile satellite (LMS) communication systems benefit to deployment of future wireless networks [1]. Besides these advantages, LMS systems meet the major challenges such as the masking effect due to nonexistence of a direct The associate editor coordinating the review of this manuscript and approving it for publication was Mohammad Tariqul Islam . line-of-sight (LoS) link between the satellite and terrestrial user equipments (UEs) related to heavy rain, fog attenuation and/or poor angles [2]. In particular, the low-power and low-cost terrestrial UEs located in a tunnel or a building meet difficulties to send information to the satellite due to limited transmit power. A hybrid terrestrial-satellite relaying communications have has been recommended in [3]- [5] to overcome the effect of masking. Since then, such system has attracted a significant amount of works. For instance, in [3], the authors studied the impact of co-channel interference (CCI) from the terrestrial network for the context of amplify-and-forward (AF) protocol deployed at the terrestrial relay. More importantly, they derived the approximate statistical distributions signal-to-interference-plus noise ratio (SINR) and hence exhibited some asymptotical computations of bit-error rate (BER) as main evaluation for the system performance. To obtain more insights into the system performance. Later, a hybrid satellite-terrestrial relay system employing multiple amplify-and-forward relays was explored by further exploiting a multi-antenna satellite to communicate with multiple users [4]. A max-max user-relay selection approach is provided to minimize the outage probability of the considered system [4]. The authors in [5] presented the average symbol error rate (SER) and obtained expression of the diversity order in the considered network in which technique for channel estimation and detection of transmitted signal are adopted related transmission between the terrestrial UE and the satellites. Moreover, the authors in [6] examined performance of AF hybrid terrestrial-satellite relaying networks with opportunistic Scheduling by deriving expressions in term of the ergodic capacity.
Actually, raised issues need be studied due to the challenges of low spectrum resource utilization and high cost in implementation of satellite mobile users [7]. As a model of cognitive radio (CR), spectrum sharing among satellite and terrestrial networks is known as promising candidate to design hybrid satellite-terrestrial cognitive networks (HSTCNs) [8]- [10]. By looking at the requirements of both spectrum efficiency and reliability, HSTCN can provide comprehensive wireless coverage as well as improve spectrum resource usage. The authors in [10] provided the closed-form formula of the outage probability (OP) for the considered system in the presence of interference power constraints imposed by multiple adjacent terrestrial primary users (PUs). The authors in [11] studied the interweave architecture for HSTCN to allow the satellite share idle spectrum with terrestrial networks. They indicated that cognitive satellite systems can not only improve the spectrum utilization but also benefit their operational revenues. However, finding out the realtime idle spectrum is difficult task for the terrestrial networks. To this end, the underlay paradigm recommended in HSTCNs as [12], [13], which, however, would cause unavoidable CCI among secondary users (SUs) and primary users (PUs). Therefore, as further difficult task due to the unknown of channel state information (CSI) exchange, a key issue regarding reliable coordination between satellite and terrestrial networks has become necessary open problems. Reference [2] addressed its achievable rate maximization. They first converted the CCI threshold into transmit power constraints, and then formulate the maximization of the achievable rate as an optimization problem. The authors in [14] explored the security problem for energy harvestingassisted cognitive satellite-terrestrial network containing a base station, some mobile users (MUs) and energy receivers (ERs), in which a multi-beam satellite sub-network shares the portion of millimeter wave bands with multiple cellular networks. The work in [15] considered system model to allow the secondary network selection to achieve optimal outage probability of the primary satellite system and then provide spectrum sharing opportunities.
As further paradigm providing spectrum efficiency for forthcoming networks, non-orthogonal multiple access (NOMA) [16]- [20] was investigated as promising wireless access technique. In principle, at the transmitter NOMA employs non-orthogonal transmission and users' superposed information achieved in power domain for higher spectrum efficiency. NOMA can serve multiple users over the same resource block, which is different from traditional orthogonal multiple access (OMA) [21], [22] and hence it can effectively improve sum rate. At the receiver, Successive Interference Cancellation (SIC) is implemented to decode the users' information. Specifically, by considering other users as interference the user with the best channel condition can be decoded firstly In [21], the authors studied the relaying scheme employed in the secondary network of the proposed CR-NOMA and the relay can be harvest energy from the secondary transmitter to serve signal forwarding to distant secondary users. They considered the complex model of wireless powered CR-NOMA in terms of outage behavior and throughput performance as awareness on imperfect SIC at the receiver. The authors [22] examined the secondary users (SUs) in the CR-NOMA network opportunistically access the licensed spectrum resources to foster the number of accessible SUs sharing the limited and dynamic spectrum resources. Moreover, partial relay selection architecture is exploited at full-duplex (FD) and half duplex (HD) relays to enhance the performance of far users. Reference [23]- [26] introduced lots of scenarios of CR-NOMA systems along with the system performance analysis. For example, [24] designed a CR-NOMA model and an spectrum efficiency optimization problem which was solved by optimizing the sensing sub-slot. Since CR-NOMA combines the advantages of the energy harvesting and device-to-device transmission mode, it is considered as a promising approach for the coming 5G communication system [25], [26].
The authors in [8], [27], [28] investigated benefits of NOMA in HSTCN. For example, [27] explored the performance of an underlay cognitive hybrid satellite-terrestrial network comprising a primary satellite source with its terrestrial receiver and the secondary transmitter (ST) with its pre-paired users on the ground. Particularly, the ST uses a cooperative NOMA scheme to permit a nearby NOMA user operating as a relay and detects and forwards signals to the distant NOMA user during the cooperation phase. Reference [8] investigated the outage probability (OP) performance of amplify-and-forward (AF) HSTCN with the NOMA scheme. In such, half-duplexing terrestrial secondary networks cooperate with a primary satellite network to provide dynamic spectrum access. By adopting pertinent heterogeneous fading models the authors in [28] obtained the closed-form expressions for the OP of primary satellite network and secondary terrestrial network. Further, we derived the asymptotic outage behavior at high signal-to-noise-ratio (SNR) for the primary and secondary networks, and thereby calculated the achievable diversity orders.

A. RELATED WORK AND MOTIVATION
The authors in [29] proposed beamforming (BF) schemes to utilize the interference from the terrestrial network as a green source to improve the physical-layer security for the HSTRN network, provided that the two networks share the portion of millimeter-wave frequencies. In [30], HSTRN system is studied in the situation that the satellite sends information to a destination through multiple relays at the existence of an eavesdropper attempting to intercept the transmissions from both the satellite and relays. They explored the context of the single-relay selection and multi-relay selection as well as round-robin scheduling schemes to address the physical-layer security of this considered HSTRN by adopting the decodeand-forward (DF) relay scheme. Reference [31] investigated the secure transmission for HSTRN where the terrestrial base station serving as a green interference resource to enhance the security of the satellite link. The authors adopted a stochastic model for the channel state information uncertainty and proposed a secure and robust beamforming framework to minimize the transmit power, while satisfying a range of outage (probabilistic) constraints at the satellite user and the terrestrial user.
However, they only considered the secure transmission for HSTRN. To the best of our knowledge, there is no prior work consider secure performance of NOMA-HSTCNs in the contexts of multiple-antenna and cooperative spectrum sharing. Therefore, we explore outage performance and secure performance metric for such systems.

B. OUR CONTRIBUTION AND ORGANIZATION
The detailed contributions of this paper are outlined as follows: • This framework is different from the system models of [29]- [31]. In particular, we experience the practical situation that the eavesdropper is able to intercept from only the terrestrial network. More specifically, we propose a practical framework of physical layer security in a NOMA-HSTCN containing nodes with hardware impairments. In such, we design one satellite to communicate to two terrestrial destination through terrestrial relay at the existence of one terrestrial eavesdropper.
• We present analytical expressions to indicate two important performance metrics such as the outage probability (OP) and intercept probability of the such system, which is applicable to evaluate reliable and secure transmission of such NOMA-HSTCN which benefits from design of multiple-antenna.
• To provide further insight, we show trends of for OP and IP at high SNR region. This is interesting guidelines to configure such system in practice. Then, we indicate which parameters are used to improve performance of specific user to satisfy specific demands in the context of NOMA.
• To provide details of our contribution, we summarize main advances in Table 1 which indicates some achieved improvements compared with other recent studies. The remainder of the paper is organized as follows. Section II provides details of the system model. In Section III, closed-form expressions of OP and IP are provided. In Section IV presents numerical results and discussion. Finally, Section V provides some concluding remarks along with future research directions.

II. SYSTEM MODEL
As depicted in Fig. 1, we consider the secondary network of the NOMA-HSTCN which consists a base station (S) facilitated M antenna, a relay (R) adopted half-duplex along with decode-and-forward (DF) protocol. It is noted that relay equipped single antenna due to low-cost and limited power, two destinations D i with N i antennas, a eavesdropper (E) with N E antennas. In primary network, we assumed the primary destination (PD) equipped single antenna has significant impacts from secondary network. Furthermore, we assume no direct link between S and D i [32] due to obstructions. Then, the transmit signal of S is given as In the first phase, S transmits the signal x S to relay R. Thus, the received signal at R is given as According to the principle of maximum ratio transmission (MRT) [4], . The signal to interference plus 215046 VOLUME 8, 2020  noise ratio (SINR) to detect signal x 2 at R is given by where Further, the SINR to detect x 1 at R is given by Due interference constraint, it is necessary to limit interference of the PD being beyond an acceptable level P P , the transmit power at S and R are given as respectively [33] and In the second phase, the relay employs DF protocol and transmits the information to D i and E. Then, the received signal at D i and E are given respectively by and Based on the maximal-ratio combining (MRC) scheme [35], we have . Then, the SINR to detect x 2 at D 2 is given as where H i = h RD i 2 F . Moreover, the SINR to detect the signal x 2 at D 1 is given as Applying SIC, the SINR at D 1 to detected the own signal x 1 is given as According to the NOMA principle, E successfully eliminates the signal x 2 and treats the signal x 1 as noise. Then, we can obtain as where Moreover, E successfully eliminates the signal x 1 of D 1 with SIC. Thus, we can obtain as VOLUME 8, 2020

III. PERFORMANCE ANALYSIS
As most of research on physical layer security, a couple of common criterion to characterize the secrecy performance are secure outage probability (SOP) and intercept probability (IP). 1 Therefore, we measure the OP and IP performance, which provide important guidelines to deploy such NOMA-HSTCN systems.

A. CHANNEL MODEL
First, we assume the channel vector h j independent and identically distributed (i.i.d.) Shadowed-Rician fading. Moreover, the probability density function (PDF) of h where m j are the fading severity parameter, j and 2b j are the average power of LOS and multi-path components,  (14) is rewritten as is the Beta function. Regarding channels in ground, the PDF of |h RP | 2 is given as where = λ m , m and are present the fading severity factor and mean, respectively.
Moreover, the PDF of H v in which v ∈ {1, 2, E} are given as respectively [36] The outage probability is evaluated as ability to a user detect its signal at related nodes. In this manner, signal x 1 of user of D 1 is processed at relay and destination D 1 . Such outage probability at D 1 is expressed as where γ th 1 = 2 2R 1 − 1 and R i is the target rates for user D i . Proposition 1: The component of overall OP, A 1 can be obtained as (20), as shown at the bottom of the page.
Proof: See Appendix A. Proposition 2: The second term of (19) an be calculated as Proof: See Appendix B. Finally, the closed-form expression for the OP of user D 1 is given by 2) OUTAGE PROBABILITY OF D 2 It is worth noting that signal x 2 is detected at user D 1 , and hence OP of user D 2 can be characterized by evaluating related SINRs compared with the thresholds. Therefore, the OP of user D 2 can be formulated by 215048 VOLUME 8, 2020 Proposition 3: The closed-form expression of B 1 is given as (24), as shown at the bottom of the page.
Proof: With the help of (3), B 1 is written as where . It then can be rewritten by Similar in appendix A, (24) is obtained. The proof is completed.
Then, putting (9) into (23) we can write B 2 as By performing similar computation, B 2 , B 3 can be expressed respectively by Then, we can compute OP for user D 2 as below

C. DIVERSITY ORDER ANALYSIS
In this paper, we consider asymptotic at high SNR. It is worth noting that the diversity order for D 1 , D 2 are defined by Whenρ → ∞, we apply a series representation of the incomplete Gamma function [45,Eq. 8.354.1] we have Then, we can write P out,∞ D 1 as (33), as shown at the bottom of the next page.
Next, P out,∞ D 2 is written as where B ∞ 1 is express as (35), as shown at the bottom of the next page. Hence, B ∞ 2 and B ∞ 3 are given in the bottom of the next page, respectively. Therefore, the diversity orders of D 1 and D 2 are equal zero.

D. INTERCEPT PROBABILITY ANALYSIS 1) INTERCEPT PROBABILITY OF D 1
The intercept probability of D 1 is expressed as [37] IP out (38) where γ E i is the secrecy SNR threshold of D i . Proposition 4: The closed-form expression for IP of D 1 is given as (39), as shown at the bottom of the next page.
The proof is completed.

E. INTERCEPT PROBABILITY OF D 2
The intercept probability of D 2 is given as [37] IP out According to the above explanations, with the closed-form expression of user D 2 , i.e. IP out D 2 is obtained as (46), as shown at the bottom of the page.

F. THROUGHPUT PERFORMANCE
It can be further evaluated other metric, i.e. the overall throughput can be achieved based on obtained outage probabilities. In delay-limited mode, at fixed target rates R 1 , R 2 the throughput can be achieved. As a result, the overall throughput can be formulated as [38] T total = 1 − P out

IV. NUMERICAL RESULTS
In this section, to verify mathematical analysis, it is necessary to simulate and illustrate for proposed assisted CR-NOMA scheme. Here, the shadowing scenarios of the satellite links H j , including the heavy shadowing (HS) with b j , m j , j = (1, 0.063, 0.0007) and average shadowing (AS) with b j , m j , j = (5, 0.251, 0.279) as [39]. Next, we set power allocation factors a 1 = b 1 = 0.2, a 2 = b 2 = 0.8,ρ =ρ S =ρ R , k = k RD i = k SR = k E , N = N 1 = N 2 = N E , the target rates R 1 = R 2 = 0.5 bit per channel use (BPCU) except for specific cases, the channel gains λ 1 = 1, λ 2 = 2, λ RP and m = m 1 = m 2 = m RP = m E . Moreover, case of m = 1 is equivalent with the Rayleigh fading channel model. In these following figures, Monte-Carlo simulations are performed to validate the analytical results. Fig. 2 shows the comparison of OP performance with HS and AS schemes for two users. By increasing average SNR at the source, OP decreases significantly, especially in high region of SNR, i.e.ρ. As can be observed, analytical results agree well with the Monte Carlo simulations, which  confirm the corrections of derived expressions. Moreover, an increasing ofρ obviously improves the SINR, then the corresponding OP performance of two users can be improved. However, such OP depends on target rates, then these curves become saturated. It is worth noting that when adjusting the power allocation coefficient reasonably, we can reduce OP performance gap among two users. The reason is that the OP depends on SINR, while SINR contain power allocation factors. This phenomenon implies that a reliable transmission can be achieved by introducing the higher SNR at source and reasonable selection of power allocation factors. Fig. 3 demonstrates the OP performance of the considered NOMA-HSTCN for different antenna configurations, where the satellite-relay link undergoes HS case. As expected, the OP improves with an increase in the number of antennas, demonstrating the advantages of employing multiple antennas and beamforming in such systems. For example, the antenna configuration with M = N = 2 can achieve a significant enhancement compared with systems with case of M = N = 1. Similarly, the power constraint with primary network contributes to vary OP of two users as shown in Fig. 4. In this observation,ρ = 30 is reported as better OP  for two users under evaluation trends of OP by adjusting ρ D from −10 to 30 dB.
Under impact of hardware imperfection, OP changes once we vary value of k, shown in Fig. 5. It can be concluded that user D 2 has little impact by hardware imperfection. In contrast, there are gaps among three cases of k = 0, 0.2, 0.4 as considering OP performance of user D 1 . The reason is that SINRs depend on k.
The impacts of power allocation factors a 1 , b 1 on OP performance can be seen clearly as Fig. 6. As seen from (9), (10), and (11), OP can be decided by varying SINRs while such SINRs depend on power allocation factors. The opposite trends of OP for two users once we increase a 1 , b 1 from 0 to 0.5. Of course, systems relying on OMA do not depend on such power factors, it is indicated as straight lines regardless of varying of a 1 , b 1 .
In Fig. 7, we can see trends of the IP for two users under various configurations of satellite link modes (AS and HS) as well as antennas. It can be confirmed that user D 2 more secure compared with D 1 regardless of other set of satellite links or the number of antennas, which means that the superiority of user D 2 compared with user D 1 in term of OP and IP are same. In addition, in Fig. 8 there are similar   performance gap regarding IP of two users when we vary value of k which is reported impact of hardware noise to IP performance. Of course, ideal hardware exhibits the best IP performance. Therefore, it can help when designing NOMA-HSTCN communication systems to combat undesired effects related hardware.
In Fig. 9, we can see how transmit SNR at source contributes to change of the throughput for two users under various configurations of satellite links as well as the number of antennas. It can be reported from (20) and (24) to confirm that the trend of throughput for two users can be predicted, i.e. it is very high at high SNR region.

V. CONCLUSION
In this paper, we have investigated the effects of hardware impairments on the reliability and security for NOMA-HSTCN systems by enabling multiple antennas design. The practical factor of hardware impairments are studied insight fully. Specifically, we derive the closed-form analytical expressions of the OP and IP to highlight as important performance metrics, and analyze the limitation of OP in the high SNR region. The results obtained have been verified by Monte Carlo simulation. These results show that as the level of hardware impairment increases, the user's OP reduce significantly, and the IP decreases as well. The different trends of curves in term of OP is caused by the power allocation parameters. It can be concluded that the severity of the performance degradation depends on several factors, including the power allocation ratio, the transmit SNR at the source, and the quality of channels. We can extend to generic framework to analyse performance of larger number of NOMA users in future work.

APPENDIX A
With the help (4) and (5), the term A 1 can be written as VOLUME 8, 2020 where σ 2 = σ 2 R = σ 2 i = σ 2 E ,ρ S =P S σ 2 and ρ D =P P σ 2 . First, we denote the first term of (48) is A 1_1 and it can be computed by where . Moreover, substituting (16) into (49) we can be calculate A 1_1 as Second, the second term of (48) is A 1_2 and it can be expressed as Furthermore, it can be calculated as After some algebraic manipulations, A 1_2 can be obtain as Finally, by substituting (52) and (55) into (48) we obtain (20) and the proof is completed.

APPENDIX B
By substituting (5) and (11) into (19), A 2 is rewritten as Similarly, we denote the first and second term of (56) are A 2_1 and A 2_2 respectively. Then, A 2_1 is express as . With the help of (17), (18) and after some variable substitutions and manipulations, we can obtain A 2_1 as Furthermore, the term A 2_2 is expressed as Similar in (55), A 2_2 is given by Now, using (58) and (60), the expected result can be attained.
This completes the proof.  He got Creative Young Medal in 2015. He published one book and five book chapters. He has authored or coauthored over 85 technical articles published in peer-reviewed international journals (SCIE) and over 60 conference papers. He led as a lead guest editor in several special issues in peer-reviewed journals. He serves as Associate Editor in seven ISI/Scopus journals. His research interests include signal processing in wireless communications networks, MIMO, NOMA, UAV networks, satellite systems, physical layer security, device-to-device transmission, and energy harvesting.
MIROSLAV VOZNAK (Senior Member, IEEE) received the Ph.D. degree in telecommunications from the Faculty of Electrical Engineering and Computer Science, VSB-Technical University of Ostrava, in 2002, and the Habilitation degree, in 2009. He was appointed as a Full Professor in electronics and communications technologies, in 2017. His research interests generally focus on information and communication technologies, especially on quality of service and experience, network security, wireless networks, and big data analytics. He is the author or coauthor of more than 100 articles in SCI/SCIE journals. He has served as a member of editorial boards for several journals, including Sensors, Journal of Communications, Elektronika Ir Elektrotechnika, and Advances in Electrical and Electronic Engineering.