Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Power Allocation Strategies for Secure Spatial Modulation

Power Allocation Strategies for Secure Spatial Modulation Power Allocation Strategies for Secure Spatial Modulation Guiyang Xia, Linqiong Jia, Yuwen Qian, Feng Shu, Zhihong Zhuang, Jiangzhou Wang, Fellow, IEEE Abstract—In secure spatial modulation (SM) networks, power As an efficient way to improve the security of SM networks, allocation (PA) strategies are investigated in this paper under the power allocation (PA) has an important impact on SR perfor- total power constraint. Considering that there is no closed-form mance. However, there are only little research of making an expression for secrecy rate (SR), an approximate closed-form investigation on PA strategies for SM. In [11], the optimal expression of SR is presented, which is used as an efficient metric PA factor was given for precoding SM by maximizing the SR to optimize PA factor and can greatly reduce the computation complexity. Based on this expression, a convex optimization (CO) performance with exhaustive search (ES). Thus, no closed- method of maximizing SR (Max-SR) is proposed accordingly. form SR expression can be developed for PA, which will Furthermore, a method of maximizing the product of signal-to- result in a high computational complexity to complete ES. leakage and noise ratio (SLNR) and artificial noise-to-leakage- This motivates us to find some closed-form solutions or low- and noise ratio (ANLNR) (Max-P-SAN) is proposed to provide an complexity iterative methods for different PA strategies. In analytic solution to PA with extremely low-complexity. Simulation results demonstrate that the SR performance of the proposed CO this paper, we will focus on the investigation of PA strategies method is close to that of the optimal PA strategy of Max-SR in secure SM (SSM) networks. The main contributions are with exhaustive search and better than that of Max-P-SAN in summarized as follows: the high signal-to-noise ratio (SNR) region. However, in the low 1) Due to the fact that SR lacks a closed-form expression and medium SNR regions, the SR performance of the proposed Max-P-SAN slightly exceeds that of the proposed CO. in SSM systems, its effective approximate simple ex- pression is defined as a metric, which can dramatically Index Terms—Spatial modulation, power allocation, secure reduce the evaluation complexity of SR values to op- transmission, finite-alphabet inputs. timize PA factor. Following this definition, a convex optimization (CO) method is proposed to address the I. INTRODUCTION optimization problem of maximizing SR (Max-SR). The simulation results show that the SR difference between S a promising and green technology in multiple-input- the proposed CO and the Max-SR with ES can be multiple-out (MIMO) systems, spatial modulation (SM) negligible for almost all SNR regions. [1] exploits both the index of activated antenna and amplitude 2) To reduce the computational complexity of the above phase modulation (APM) symbol to transmit messages. Due CO method and at the same time provide a closed-form to the broadcasting characteristic of wireless transmission [2]– PA strategy, a PA strategy of maximizing the product [4] , physical layer security becomes an urgent and important of signal-to-leakage plus noise ratio (SLNR) and AN- problem in wireless communication [5]–[7]. to-leakage plus noise ratio (ANLNR) (Max-P-SAN) is How to make SM have a capability to achieve a secure proposed, which can strike a good balance between max- transmission become an important issue for SM networks. In imizing SLNR and maximizing ANLNR. Its analytic [8], without the knowledge of Eve’s location, the confidential expression of PA factor is also given. Simulation results messages are securely transmitted from the SM transmitter to show that the SR performance of Max-P-SAN method, the desired receiver by projecting artificial noise (AN) onto the with extremely low complexity, tends to that of Max-SR null-space of the desired channel. In [9], the authors proposed with ES method and is slightly better than that of CO a full-duplex desired receiver, where the confidential messages in the low and medium SNR regions. is received and meanwhile AN is emitted to corrupt the illegal The reminder is organized as follows. In Section II, a SSM receiver (Eve). This scheme can provide a high capability to system with the aid of AN is described. In Section III, first, the combat eavesdropping. The authors in [10] proposed two new approximate simple formula of average SR is given, and two schemes of transmit antenna selection for secure SM networks: PA strategies, CO and Max-P-SAN, are proposed to maximize maximizing secrecy rate (SR) and leakage, where the proposed approximate SR and the product of SLNR and ANLNR, leakage-based antenna selection scheme achieve an excellent respectively. Subsequently, numerical simulations and analysis SR performance with a very low-complexity. are presented in Section IV. Finally, we make our conclusions in Section V. Guiyang Xia, Linqiong Jia, Yuwen Qian, Feng Shu and Zhihong Zhuang are with the School of Electronic and Optical Engineering, Nanjing University of Science and Technology, 210094, CHINA. Feng Shu is also with the School II. SYSTEM MODEL of Computer and information at Fujian Agriculture and Forestry University, Fuzhou, 350002, China. Fig. 1 sketches a SSM system with N transmit antennas Jiangzhou Wang is with the School of Engineering and Digital Arts, University of Kent, Canterbury CT2 7NT, U.K. (TAs) at Alice, N receive antennas (RAs) at Bob, and N r e arXiv:1808.00646v1 [eess.SP] 2 Aug 2018 2 symbol vectors. W is the covariance matrix of the last two terms of y in (2), i.e. AN plus noise, and W = 2 H H (1 − β)P C + σ I , where C = H TT H . Similarly, B N B B r B 2 H H W = (1 − β)P C + σ I where C = H TT H . E E N E E e E From [8], it is known that pre-multiplying y in (2) by −1/2 W from left is to whiten the AN plus noise into a white Gaussian noise. The linear transformation does not change the mutual information, thus I(x; y ) = I(x; y ), where B B −1/2 −1/2 ′ ′ y = W y , and n = W ( (1 − β)P H Tn+n ). B B B B B B B Similarly, I(x; y ) = I(x; y ). The average SR is defined as E E R = E [I(x; y ) − I(x; y ), 0] . (6) Fig. 1. A secure SM system model. s H ,HE B B E where [a] =max(a,0) and R (β) = I(x; y )− I(x; y ) is the B E instantaneous SR for a specific channel realization. Here, we RAs at Eve, respectively. Here, Eve intends to intercept the assume that the ideal channel knowledge of H and H per confidential messages. Additionally, we denote the size of B E channel use are available at transmitter [7]. In accordance with signal constellation M by M. As a result, log N M bits per the above equations, the optimization problem of maximizing channel use can be transmitted, where log N bits are used to SR over PA factor can be casted as select one active antenna, and the remaining log M bits are used to form a constellation symbol. max R (β) s.t. 0 ≤ β ≤ 1. (7) Referring to the SSM model in [10], the transmit signal with the aid of AN is represented by p p III. POWER ALLOCATION STRATEGY FOR SECRECY RATE s = βP e b + (1 − β)P Tn (1) i j MAXIMIZATION where β ∈ [0, 1] is the PA factor, P denotes the total transmit In this section, two new PA methods, called CO and Max- power. e is the ith column of identity matrix I , which i N P-SAN, are proposed. The former forms an iterative solution, means the ith antenna is chosen for transmitting symbol b , and the latter produces a closed-form PA expression. which is the input symbol equiprobably drawn from a M- ary constellation. T is the projection matrix of the AN vector N ×1 A. Proposed CO method n ∈ C with tr(TT ) = 1, where tr(·) denotes the matrix trace. The receive vector of symbols at the desired and Due to the absence of closed-form expression of SR, it eavesdropping receivers are is hard for us to design an efficient method to optimize PA p p factor directly. Although the ES method in [11] is employed y = βP H e b + (1 − β)P H Tn + n (2) B i j B B to find out the optimal PA factor for a given SNR, but its high p p y = βP H e b + (1 − β)P H Tn + n (3) complexity limits its applications to practical SM systems. E i j E E In view of this, the cut-off rate [4] with closed-form for N ×N N ×N r t e t where H ∈ C and H ∈ C are the complex B E traditional MIMO systems can be easily extended to the secure channel gain matrices from Alice to Bob and from Alice to SM systems, and may be adopted as an efficient metric to Eve, with each elements of H and H obeying the Gaussian B E optimize the PA factor as follows distributions with zero mean and unit variance, i.e., CN (0, 1). N ×1 N ×1 r e a B E Additionally, n ∈ C , and n ∈ C are complex B E R (β) = I − I , (8) s 0 0 Gaussian noise at desired and eavesdropping receivers with 2 2 where I is the cut-off rate for the desired receiver given by n ∼ CN (0, σ I ) and n ∼ CN (0, σ I ), respectively. B N E N B r E e Given a specific channel realization, the mutual information N M N M t t X X −βP B H H between Alice and Bob, and between Alice and Eve are I = ζ − log exp d H ω H d (9) B B ij 0 2 ij B i=1 j=1 I (s; y |H ) = log N M− B B t B 2   −1 where ζ = 2log N M, ω (β) = W . Similarly, the cut-off t B N M N M t  t  2 B X X 1 E ′ 2 rate I for the eavesdropper is E ′ log exp −f + kn k (4) 0 n b,i,j 2 B N M   i=1 j=1 N M N M t t X X −βP E H H I = ζ − log exp d H ω H d I (s; y |H ) = log N M− E E mk E E t 0 2 mk E E 2 ( ) m=1 k=1 NtM NtM X X ′ 2 (10) E log exp −f + kn k (5) e,m,k 2 E N M −1 m=1 k=1 where ω(β) = W . The detailed process of (8) refers to −1/2 ′ 2 Appendix A in [4]. Replacing the objective function in (7) by where f = k βP W H d + n k , f = b,i,j B ij e,m,k B B −1/2 (8) yields ′ 2 k βP W H d + n k , d = x − x , and d = E mk ij i j mk E E x − x . Here, x , x , x , or x is one possible transmit m k i j m k max R (β) s.t. 0 ≤ β ≤ 1. (11) vectors in the set of combining antenna and all possible 3 P H P H However, the objective function of problem (11) is non- where κ = tr(H H ), κ = tr(H H ), ω = B B E E B B E N N t t −1 concave. Note that ω (β) ≈ C in the high SNR P tr(C ), and ω = P tr(C ). Therefore, the promising E B E E (1−β)P E optimal values of β in (18) should satisfy the following region (i.e, σ → 0) when C is nonsingular, thus we have equation κ ˜ (β) ≈ log κ E E ϕ ϕ β − 2ϕ β + ϕ N M N M t t a o d d H H −1 ′ X X −βd H C H d F (β) = = 0 (19) E mk mk E E 2 2 = log exp . (12) (−ϕ β + ϕ β + ϕ ) b c d 4(1 − β) m=1 k=1 where ϕ = κ ω , ϕ = κ ω , ϕ = κ ω + σ κ N , a B E b E B c E B E e It can be seen that κ ˜ (β) is convex with respect to β and 2 2 2 ϕ = σ ω N + σ σ N N , and ϕ = ϕ − ϕ . Based d B r r e o b c B B E then the objective function becomes a difference between two on (19), it is seen that the denominator of the derivative and convex functions. To convert this difference to a concave ϕ in (19) are both greater than 0, we only need to solve the function, we have the linear under-estimator of κ ˜ (β) at the roots of equation ϕ β − 2ϕ β + ϕ = 0. Due to ϕ < 0 and o d d o feasible point β as follows k−1 2 △ = ϕ − ϕ ϕ > 0, this equation has two real-valued roots. o d In summary, the set of feasible solutions to (18) is κ ˜ ≥ κ ˜ (β ) + κ ˜ (β )(β − β ) = g (β ) (13) E E k−1 k−1 k k−1 E k √ √ ϕ + △ ϕ − △ where κ ˜ (β ) is the first derivative value of function κ ˜ at d d k−1 E S = β = , β = , β = 0, β = 1 , 1 2 3 4 β , and ϕ ϕ k−1 o o N M N M t t X X where β , and β are two solutions to the quadratic equation 1 2 1 −PQ −βPQ mk mk κ = exp (14) E in (19) while β = 0, and β = 1 are two end-points of 2 3 4 ln2 · κ 4(1 − β) 4(1 − β) m=1 k=1 the feasible search interval [0, 1]. Obviously, β = 0 means H H −1 that there is no confidential messages to be sent. In other where Q = d H C H d . For a given feasible mk E mk mk E words, SR=0. Thus, this point can be directly removed from solution β , the problem (11) can be solved by the following the solution set. β < 0 falls outside the feasible set [0, 1], approximate iterative sequence of convex problems 1 and can be deleted directly. Considering the function F (β) max G(β ) = g (β ) − κ ˜ (β ) s.t. 0 < β < 1. (15) k E k B k k is a continuous and differentiable function over the interval [0, 1], its first derivative is negative as β goes to one from It is clear that the objective function in (15) is concave. Start- the left. Thus, F (1) is the local minimum point, which rules ing with an feasible point β , the optimization problem (15) is out it from the the feasible solution set referring to the set of iteratively solved with different β , where {β } is the gener- k k maximizing F (β). Finally, we have the unique solution ated sequence of solutions corresponding to the kth iteration. This iterative process terminates until | G(β )−G(β ) |≤ ε, k k−1 β = (ϕ − △)/ϕ (20) 2 d o where ε is a prechosen threshold. due to the fact that F (β ) = 0 and β ∈ [0, 1]. 2 2 B. Proposed Max-P-SAN method Utilizing the leakage idea [12], [13], the SLNR, mainly C. Complexity Analysis and Comparison denoting the desired signal leakage to the eavesdropping Below, we present a complexity comparison among the direction, is given by three methods: CO, Max-P-SAN, and ES. Firstly, the complex- −1 2 2 H H 2 ity of the ES method in [11] is about C = 2N M lN ES samp SLNR = tr(H H ) tr(H H ) + σ N N /βP (16) t B B E t r B E B [2(N +N )N +N +N ] floating-point operations (FLOPs), r e r e In the same manner, the AN is viewed as the useful signal of where l denotes the number of searches depending on the the eavesdropper, the ANLNR from the wiretap channel to the required accuracy, and N (≥ 500) is the number of realiza- samp desired channels is as follows tions of noise sample points for accurately estimating expec- −1 tation operators. For the proposed CO, its computational com- ANLNR = tr(C ) tr(C ) + σ N /(1 − β)P (17) E E B e 2 2 2 plexity is approximated as C = 3N M D (2N + 2N ) CO ite t t t It is hard to jointly optimize the two objective functions FLOPs, where D is the number of iterations. Finally, it is ite SLNR and ANLNR . To simplify the joint optimization B E obvious that the proposed Max-P-SAN scheme has the lowest problem, we multiply the two functions to form a new product complexity among the three methods, and its complexity is of SLNR and ANLNR , which is used as a single objective 2 2 2 B E C = 2N (2N + 3N ) + 2N N + 2N N + N + Max-P-SAN r e t t t t r e function. This will significantly simplify our optimization N + N FLOPs. From the three complexity expressions, the r e manipulation. Maximizing their product means maximizing at dominant term in C is only quadratic. In general, Max-P-SAN least one of them, or both them. From simulations, we find that N ≫ N > N (N ), their complexities have an increasing samp t r e the proposed product method performs very well, and make a order as follows: Max-P-SAN, CO, and ES. good balance between performance and complexity. Then, the associated optimization problem can be written as IV. SIMULATION RESULTS κ · β (1 − β) · ω B E max F (β) = · In what follows, numerical simulations are presented to 2 2 κ · β + σ N N (1 − β) · ω + σ N E t r B e B E evaluate the SR performance for two proposed PA strategies, s.t. 0 ≤ β ≤ 1. (18) with ES method as a performance benchmark. Specially, the 4 β for three given SNRs: 0dB, 5dB, and 20dB, with ES as a performance benchmark. From this figure, it is obvious that ES Method Proposed Max-P-SAN all optimal values of β reduce as SNR increases from 0dB to Proposed CO Method 3.5 20dB. This can be readily explained as follows: a high SNR Typical PA factor: β=0.1 Typical PA factor: β=0.25 means a good channel quality. This implies that less power is Typical PA factor: β=0.5 required to transmit confidential messages, and more power is utilized to emit AN to corrupt eavesdroppers. Additionally, as 2.5 SNR increases, the optimal values of β corresponding to the two proposed PA are closer to that of β for ES. 1.5 V. CONCLUSION In this paper, we have made an investigation of PA strategies for SSM system. An efficient approximated expression of SR was given to simplify the computational complexity for 0.5 0 5 10 15 20 25 30 optimizing PA factor. Then, two PA strategies were proposed SNR (dB) to implement PA between confidential messages and AN. The first one is CO and the second one is Max-P-SAN. The former Fig. 2. Comparison of average SR for various PA methods with N = 4, N = 2 and N = 2. is iterative while the latter is closed-form. In accordance r e with simulations, we find: the proposed CO provides a SR performance being close to the ES method for almost all 2 2 noise variances are assumed to be identical, i.e., σ = σ . SNR regions, and the proposed Max-P-SAN can achieve the B E Modulation type is quadrature phase shift keying (QPSK). optimal SR performance in the low and medium regions with Fig. 2 plots the curves of SR versus SNR with N = 4, N = t r an extremely low-complexity. These strategies can be applied 2, and N = 2. Here, three typical PA strategies, β = 0.1, 0.25, e to the future SSM-based networks such as unmanned aerial and 0.5, are used as performance references. From Fig. 2, it is vehicle, future mobile networks, and intelligent transportation. seen that the proposed CO method can achieve the optimal SR performance being close to that of the ES method for almost all REFERENCES SNR regions. The proposed Max-P-SAN method approaches the ES performance in the low and medium SNR regions, but [1] R. Y. Mesleh, H. Haas, S. Sinanovic, C. W. Ahn, and S. Yun, “Spatial modulation,” IEEE Trans. Veh. Technol., vol. 57, no. 4, pp. 2228–2241, slightly worse than the ES in the high SNR region in terms Jul. 2008. of SR. Because Max-P-SAN has a closed-form expression, it [2] H. M. Wang, Q. Yin, and X. G. Xia, “Distributed beamforming for strikes a good balance between performance and complexity. physical-layer security of two-way relay networks,” IEEE Trans. Signal Process., vol. 60, no. 7, pp. 3532–3545, Jul. 2012. Also, the two proposed methods perform much better than [3] Q. Wu, G. Y. Li, W. Chen, D. W. K. Ng, and R. Schober, “An overview three typical fixed PA schemes. This means that they can of sustainable green 5G networks,” IEEE Wireless Commun., vol. 24, harvest appreciable SR performance gains. no. 4, pp. 72–80, Aug. 2017. [4] S. R. Aghdam and T. M. Duman, “Joint precoder and artificial noise design for MIMO wiretap channels with finite-alphabet inputs based on the cut-off rate,” IEEE Trans. Wireless Commun., vol. 16, no. 6, pp. 3913–3923, Jun. 2017. SNR= 20dB [5] X. Chen, X. Chen, and T. Liu, “A unified performance optimization Optimal β (ES) 3.5 for secrecy wireless information and power transfer over interference Proposed Max-P-SAN channels,” IEEE Access, vol. 5, pp. 12 726–12 736, 2017. Proposed CO Method [6] S. Yan, X. Zhou, N. Yang, B. He, and T. D. Abhayapala, “Artificial- noise-aided secure transmission in wiretap channels with transmitter- side correlation,” IEEE Trans. on Wireless Commun., vol. 15, no. 12, 2.5 pp. 8286–8297, Dec. 2016. SNR= 5dB [7] Y. Wu, J. B. Wang, J. Wang, R. Schober, and C. Xiao, “Secure transmission with large numbers of antennas and finite alphabet inputs,” IEEE Trans. Commun., vol. 65, no. 8, pp. 3614–3628, Aug. 2017. 1.5 SNR= 0dB [8] L. Wang, S. Bashar, Y. Wei, and R. Li, “Secrecy enhancement analysis against unknown eavesdropping in spatial modulation,” IEEE Commun. Lett., vol. 19, no. 8, pp. 1351–1354, Aug. 2015. [9] C. Liu, L. L. Yang, and W. Wang, “Secure spatial modulation with a full-duplex receiver,” IEEE Wireless Commun. Lett., vol. 6, no. 6, pp. 0.5 838–841, Dec. 2017. [10] F. Shu, Z. Wang, R. Chen, Y. Wu, and J. Wang, “Two high-performance 0 0.2 0.4 0.6 0.8 1 schemes of transmit antenna selection for secure spatial modulation,” IEEE Trans. Veh. Technol., pp. 1–1, 2018. [11] F. Wu, L. L. Yang, W. Wang, and Z. Kong, “Secret precoding-aided spatial modulation,” IEEE Commun. Lett., vol. 19, no. 9, pp. 1544–1547, Fig. 3. Comparison of the achievable SR for SNR=0, 5, 20dB with same Sep. 2015. configuration as Fig. 2. [12] M. Sadek, A. Tarighat, and A. H. Sayed, “A leakage-based precoding scheme for downlink multi-user mimo channels,” IEEE Transactions on Fig. 3 plots the curves of maximum achievable SR versus Wireless Communications, vol. 6, no. 5, pp. 1711–1721, May 2007. Average Secrecy Rate (bits/s/Hz) Average Secrecy Rate (bits/s/Hz) 5 [13] S. Feng, M. M. Wang, W. Yaxi, F. Haiqiang, and L. Jinhui, “An efficient power allocation scheme for leakage-based precoding in multi- cell multiuser MIMO downlink,” IEEE Commun. Lett., vol. 15, no. 10, pp. 1053–1055, Oct. 2011. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Electrical Engineering and Systems Science arXiv (Cornell University)

Power Allocation Strategies for Secure Spatial Modulation

Loading next page...
 
/lp/arxiv-cornell-university/power-allocation-strategies-for-secure-spatial-modulation-ql9bsim00e
ISSN
1932-8184
eISSN
ARCH-3348
DOI
10.1109/JSYST.2019.2918168
Publisher site
See Article on Publisher Site

Abstract

Power Allocation Strategies for Secure Spatial Modulation Guiyang Xia, Linqiong Jia, Yuwen Qian, Feng Shu, Zhihong Zhuang, Jiangzhou Wang, Fellow, IEEE Abstract—In secure spatial modulation (SM) networks, power As an efficient way to improve the security of SM networks, allocation (PA) strategies are investigated in this paper under the power allocation (PA) has an important impact on SR perfor- total power constraint. Considering that there is no closed-form mance. However, there are only little research of making an expression for secrecy rate (SR), an approximate closed-form investigation on PA strategies for SM. In [11], the optimal expression of SR is presented, which is used as an efficient metric PA factor was given for precoding SM by maximizing the SR to optimize PA factor and can greatly reduce the computation complexity. Based on this expression, a convex optimization (CO) performance with exhaustive search (ES). Thus, no closed- method of maximizing SR (Max-SR) is proposed accordingly. form SR expression can be developed for PA, which will Furthermore, a method of maximizing the product of signal-to- result in a high computational complexity to complete ES. leakage and noise ratio (SLNR) and artificial noise-to-leakage- This motivates us to find some closed-form solutions or low- and noise ratio (ANLNR) (Max-P-SAN) is proposed to provide an complexity iterative methods for different PA strategies. In analytic solution to PA with extremely low-complexity. Simulation results demonstrate that the SR performance of the proposed CO this paper, we will focus on the investigation of PA strategies method is close to that of the optimal PA strategy of Max-SR in secure SM (SSM) networks. The main contributions are with exhaustive search and better than that of Max-P-SAN in summarized as follows: the high signal-to-noise ratio (SNR) region. However, in the low 1) Due to the fact that SR lacks a closed-form expression and medium SNR regions, the SR performance of the proposed Max-P-SAN slightly exceeds that of the proposed CO. in SSM systems, its effective approximate simple ex- pression is defined as a metric, which can dramatically Index Terms—Spatial modulation, power allocation, secure reduce the evaluation complexity of SR values to op- transmission, finite-alphabet inputs. timize PA factor. Following this definition, a convex optimization (CO) method is proposed to address the I. INTRODUCTION optimization problem of maximizing SR (Max-SR). The simulation results show that the SR difference between S a promising and green technology in multiple-input- the proposed CO and the Max-SR with ES can be multiple-out (MIMO) systems, spatial modulation (SM) negligible for almost all SNR regions. [1] exploits both the index of activated antenna and amplitude 2) To reduce the computational complexity of the above phase modulation (APM) symbol to transmit messages. Due CO method and at the same time provide a closed-form to the broadcasting characteristic of wireless transmission [2]– PA strategy, a PA strategy of maximizing the product [4] , physical layer security becomes an urgent and important of signal-to-leakage plus noise ratio (SLNR) and AN- problem in wireless communication [5]–[7]. to-leakage plus noise ratio (ANLNR) (Max-P-SAN) is How to make SM have a capability to achieve a secure proposed, which can strike a good balance between max- transmission become an important issue for SM networks. In imizing SLNR and maximizing ANLNR. Its analytic [8], without the knowledge of Eve’s location, the confidential expression of PA factor is also given. Simulation results messages are securely transmitted from the SM transmitter to show that the SR performance of Max-P-SAN method, the desired receiver by projecting artificial noise (AN) onto the with extremely low complexity, tends to that of Max-SR null-space of the desired channel. In [9], the authors proposed with ES method and is slightly better than that of CO a full-duplex desired receiver, where the confidential messages in the low and medium SNR regions. is received and meanwhile AN is emitted to corrupt the illegal The reminder is organized as follows. In Section II, a SSM receiver (Eve). This scheme can provide a high capability to system with the aid of AN is described. In Section III, first, the combat eavesdropping. The authors in [10] proposed two new approximate simple formula of average SR is given, and two schemes of transmit antenna selection for secure SM networks: PA strategies, CO and Max-P-SAN, are proposed to maximize maximizing secrecy rate (SR) and leakage, where the proposed approximate SR and the product of SLNR and ANLNR, leakage-based antenna selection scheme achieve an excellent respectively. Subsequently, numerical simulations and analysis SR performance with a very low-complexity. are presented in Section IV. Finally, we make our conclusions in Section V. Guiyang Xia, Linqiong Jia, Yuwen Qian, Feng Shu and Zhihong Zhuang are with the School of Electronic and Optical Engineering, Nanjing University of Science and Technology, 210094, CHINA. Feng Shu is also with the School II. SYSTEM MODEL of Computer and information at Fujian Agriculture and Forestry University, Fuzhou, 350002, China. Fig. 1 sketches a SSM system with N transmit antennas Jiangzhou Wang is with the School of Engineering and Digital Arts, University of Kent, Canterbury CT2 7NT, U.K. (TAs) at Alice, N receive antennas (RAs) at Bob, and N r e arXiv:1808.00646v1 [eess.SP] 2 Aug 2018 2 symbol vectors. W is the covariance matrix of the last two terms of y in (2), i.e. AN plus noise, and W = 2 H H (1 − β)P C + σ I , where C = H TT H . Similarly, B N B B r B 2 H H W = (1 − β)P C + σ I where C = H TT H . E E N E E e E From [8], it is known that pre-multiplying y in (2) by −1/2 W from left is to whiten the AN plus noise into a white Gaussian noise. The linear transformation does not change the mutual information, thus I(x; y ) = I(x; y ), where B B −1/2 −1/2 ′ ′ y = W y , and n = W ( (1 − β)P H Tn+n ). B B B B B B B Similarly, I(x; y ) = I(x; y ). The average SR is defined as E E R = E [I(x; y ) − I(x; y ), 0] . (6) Fig. 1. A secure SM system model. s H ,HE B B E where [a] =max(a,0) and R (β) = I(x; y )− I(x; y ) is the B E instantaneous SR for a specific channel realization. Here, we RAs at Eve, respectively. Here, Eve intends to intercept the assume that the ideal channel knowledge of H and H per confidential messages. Additionally, we denote the size of B E channel use are available at transmitter [7]. In accordance with signal constellation M by M. As a result, log N M bits per the above equations, the optimization problem of maximizing channel use can be transmitted, where log N bits are used to SR over PA factor can be casted as select one active antenna, and the remaining log M bits are used to form a constellation symbol. max R (β) s.t. 0 ≤ β ≤ 1. (7) Referring to the SSM model in [10], the transmit signal with the aid of AN is represented by p p III. POWER ALLOCATION STRATEGY FOR SECRECY RATE s = βP e b + (1 − β)P Tn (1) i j MAXIMIZATION where β ∈ [0, 1] is the PA factor, P denotes the total transmit In this section, two new PA methods, called CO and Max- power. e is the ith column of identity matrix I , which i N P-SAN, are proposed. The former forms an iterative solution, means the ith antenna is chosen for transmitting symbol b , and the latter produces a closed-form PA expression. which is the input symbol equiprobably drawn from a M- ary constellation. T is the projection matrix of the AN vector N ×1 A. Proposed CO method n ∈ C with tr(TT ) = 1, where tr(·) denotes the matrix trace. The receive vector of symbols at the desired and Due to the absence of closed-form expression of SR, it eavesdropping receivers are is hard for us to design an efficient method to optimize PA p p factor directly. Although the ES method in [11] is employed y = βP H e b + (1 − β)P H Tn + n (2) B i j B B to find out the optimal PA factor for a given SNR, but its high p p y = βP H e b + (1 − β)P H Tn + n (3) complexity limits its applications to practical SM systems. E i j E E In view of this, the cut-off rate [4] with closed-form for N ×N N ×N r t e t where H ∈ C and H ∈ C are the complex B E traditional MIMO systems can be easily extended to the secure channel gain matrices from Alice to Bob and from Alice to SM systems, and may be adopted as an efficient metric to Eve, with each elements of H and H obeying the Gaussian B E optimize the PA factor as follows distributions with zero mean and unit variance, i.e., CN (0, 1). N ×1 N ×1 r e a B E Additionally, n ∈ C , and n ∈ C are complex B E R (β) = I − I , (8) s 0 0 Gaussian noise at desired and eavesdropping receivers with 2 2 where I is the cut-off rate for the desired receiver given by n ∼ CN (0, σ I ) and n ∼ CN (0, σ I ), respectively. B N E N B r E e Given a specific channel realization, the mutual information N M N M t t X X −βP B H H between Alice and Bob, and between Alice and Eve are I = ζ − log exp d H ω H d (9) B B ij 0 2 ij B i=1 j=1 I (s; y |H ) = log N M− B B t B 2   −1 where ζ = 2log N M, ω (β) = W . Similarly, the cut-off t B N M N M t  t  2 B X X 1 E ′ 2 rate I for the eavesdropper is E ′ log exp −f + kn k (4) 0 n b,i,j 2 B N M   i=1 j=1 N M N M t t X X −βP E H H I = ζ − log exp d H ω H d I (s; y |H ) = log N M− E E mk E E t 0 2 mk E E 2 ( ) m=1 k=1 NtM NtM X X ′ 2 (10) E log exp −f + kn k (5) e,m,k 2 E N M −1 m=1 k=1 where ω(β) = W . The detailed process of (8) refers to −1/2 ′ 2 Appendix A in [4]. Replacing the objective function in (7) by where f = k βP W H d + n k , f = b,i,j B ij e,m,k B B −1/2 (8) yields ′ 2 k βP W H d + n k , d = x − x , and d = E mk ij i j mk E E x − x . Here, x , x , x , or x is one possible transmit m k i j m k max R (β) s.t. 0 ≤ β ≤ 1. (11) vectors in the set of combining antenna and all possible 3 P H P H However, the objective function of problem (11) is non- where κ = tr(H H ), κ = tr(H H ), ω = B B E E B B E N N t t −1 concave. Note that ω (β) ≈ C in the high SNR P tr(C ), and ω = P tr(C ). Therefore, the promising E B E E (1−β)P E optimal values of β in (18) should satisfy the following region (i.e, σ → 0) when C is nonsingular, thus we have equation κ ˜ (β) ≈ log κ E E ϕ ϕ β − 2ϕ β + ϕ N M N M t t a o d d H H −1 ′ X X −βd H C H d F (β) = = 0 (19) E mk mk E E 2 2 = log exp . (12) (−ϕ β + ϕ β + ϕ ) b c d 4(1 − β) m=1 k=1 where ϕ = κ ω , ϕ = κ ω , ϕ = κ ω + σ κ N , a B E b E B c E B E e It can be seen that κ ˜ (β) is convex with respect to β and 2 2 2 ϕ = σ ω N + σ σ N N , and ϕ = ϕ − ϕ . Based d B r r e o b c B B E then the objective function becomes a difference between two on (19), it is seen that the denominator of the derivative and convex functions. To convert this difference to a concave ϕ in (19) are both greater than 0, we only need to solve the function, we have the linear under-estimator of κ ˜ (β) at the roots of equation ϕ β − 2ϕ β + ϕ = 0. Due to ϕ < 0 and o d d o feasible point β as follows k−1 2 △ = ϕ − ϕ ϕ > 0, this equation has two real-valued roots. o d In summary, the set of feasible solutions to (18) is κ ˜ ≥ κ ˜ (β ) + κ ˜ (β )(β − β ) = g (β ) (13) E E k−1 k−1 k k−1 E k √ √ ϕ + △ ϕ − △ where κ ˜ (β ) is the first derivative value of function κ ˜ at d d k−1 E S = β = , β = , β = 0, β = 1 , 1 2 3 4 β , and ϕ ϕ k−1 o o N M N M t t X X where β , and β are two solutions to the quadratic equation 1 2 1 −PQ −βPQ mk mk κ = exp (14) E in (19) while β = 0, and β = 1 are two end-points of 2 3 4 ln2 · κ 4(1 − β) 4(1 − β) m=1 k=1 the feasible search interval [0, 1]. Obviously, β = 0 means H H −1 that there is no confidential messages to be sent. In other where Q = d H C H d . For a given feasible mk E mk mk E words, SR=0. Thus, this point can be directly removed from solution β , the problem (11) can be solved by the following the solution set. β < 0 falls outside the feasible set [0, 1], approximate iterative sequence of convex problems 1 and can be deleted directly. Considering the function F (β) max G(β ) = g (β ) − κ ˜ (β ) s.t. 0 < β < 1. (15) k E k B k k is a continuous and differentiable function over the interval [0, 1], its first derivative is negative as β goes to one from It is clear that the objective function in (15) is concave. Start- the left. Thus, F (1) is the local minimum point, which rules ing with an feasible point β , the optimization problem (15) is out it from the the feasible solution set referring to the set of iteratively solved with different β , where {β } is the gener- k k maximizing F (β). Finally, we have the unique solution ated sequence of solutions corresponding to the kth iteration. This iterative process terminates until | G(β )−G(β ) |≤ ε, k k−1 β = (ϕ − △)/ϕ (20) 2 d o where ε is a prechosen threshold. due to the fact that F (β ) = 0 and β ∈ [0, 1]. 2 2 B. Proposed Max-P-SAN method Utilizing the leakage idea [12], [13], the SLNR, mainly C. Complexity Analysis and Comparison denoting the desired signal leakage to the eavesdropping Below, we present a complexity comparison among the direction, is given by three methods: CO, Max-P-SAN, and ES. Firstly, the complex- −1 2 2 H H 2 ity of the ES method in [11] is about C = 2N M lN ES samp SLNR = tr(H H ) tr(H H ) + σ N N /βP (16) t B B E t r B E B [2(N +N )N +N +N ] floating-point operations (FLOPs), r e r e In the same manner, the AN is viewed as the useful signal of where l denotes the number of searches depending on the the eavesdropper, the ANLNR from the wiretap channel to the required accuracy, and N (≥ 500) is the number of realiza- samp desired channels is as follows tions of noise sample points for accurately estimating expec- −1 tation operators. For the proposed CO, its computational com- ANLNR = tr(C ) tr(C ) + σ N /(1 − β)P (17) E E B e 2 2 2 plexity is approximated as C = 3N M D (2N + 2N ) CO ite t t t It is hard to jointly optimize the two objective functions FLOPs, where D is the number of iterations. Finally, it is ite SLNR and ANLNR . To simplify the joint optimization B E obvious that the proposed Max-P-SAN scheme has the lowest problem, we multiply the two functions to form a new product complexity among the three methods, and its complexity is of SLNR and ANLNR , which is used as a single objective 2 2 2 B E C = 2N (2N + 3N ) + 2N N + 2N N + N + Max-P-SAN r e t t t t r e function. This will significantly simplify our optimization N + N FLOPs. From the three complexity expressions, the r e manipulation. Maximizing their product means maximizing at dominant term in C is only quadratic. In general, Max-P-SAN least one of them, or both them. From simulations, we find that N ≫ N > N (N ), their complexities have an increasing samp t r e the proposed product method performs very well, and make a order as follows: Max-P-SAN, CO, and ES. good balance between performance and complexity. Then, the associated optimization problem can be written as IV. SIMULATION RESULTS κ · β (1 − β) · ω B E max F (β) = · In what follows, numerical simulations are presented to 2 2 κ · β + σ N N (1 − β) · ω + σ N E t r B e B E evaluate the SR performance for two proposed PA strategies, s.t. 0 ≤ β ≤ 1. (18) with ES method as a performance benchmark. Specially, the 4 β for three given SNRs: 0dB, 5dB, and 20dB, with ES as a performance benchmark. From this figure, it is obvious that ES Method Proposed Max-P-SAN all optimal values of β reduce as SNR increases from 0dB to Proposed CO Method 3.5 20dB. This can be readily explained as follows: a high SNR Typical PA factor: β=0.1 Typical PA factor: β=0.25 means a good channel quality. This implies that less power is Typical PA factor: β=0.5 required to transmit confidential messages, and more power is utilized to emit AN to corrupt eavesdroppers. Additionally, as 2.5 SNR increases, the optimal values of β corresponding to the two proposed PA are closer to that of β for ES. 1.5 V. CONCLUSION In this paper, we have made an investigation of PA strategies for SSM system. An efficient approximated expression of SR was given to simplify the computational complexity for 0.5 0 5 10 15 20 25 30 optimizing PA factor. Then, two PA strategies were proposed SNR (dB) to implement PA between confidential messages and AN. The first one is CO and the second one is Max-P-SAN. The former Fig. 2. Comparison of average SR for various PA methods with N = 4, N = 2 and N = 2. is iterative while the latter is closed-form. In accordance r e with simulations, we find: the proposed CO provides a SR performance being close to the ES method for almost all 2 2 noise variances are assumed to be identical, i.e., σ = σ . SNR regions, and the proposed Max-P-SAN can achieve the B E Modulation type is quadrature phase shift keying (QPSK). optimal SR performance in the low and medium regions with Fig. 2 plots the curves of SR versus SNR with N = 4, N = t r an extremely low-complexity. These strategies can be applied 2, and N = 2. Here, three typical PA strategies, β = 0.1, 0.25, e to the future SSM-based networks such as unmanned aerial and 0.5, are used as performance references. From Fig. 2, it is vehicle, future mobile networks, and intelligent transportation. seen that the proposed CO method can achieve the optimal SR performance being close to that of the ES method for almost all REFERENCES SNR regions. The proposed Max-P-SAN method approaches the ES performance in the low and medium SNR regions, but [1] R. Y. Mesleh, H. Haas, S. Sinanovic, C. W. Ahn, and S. Yun, “Spatial modulation,” IEEE Trans. Veh. Technol., vol. 57, no. 4, pp. 2228–2241, slightly worse than the ES in the high SNR region in terms Jul. 2008. of SR. Because Max-P-SAN has a closed-form expression, it [2] H. M. Wang, Q. Yin, and X. G. Xia, “Distributed beamforming for strikes a good balance between performance and complexity. physical-layer security of two-way relay networks,” IEEE Trans. Signal Process., vol. 60, no. 7, pp. 3532–3545, Jul. 2012. Also, the two proposed methods perform much better than [3] Q. Wu, G. Y. Li, W. Chen, D. W. K. Ng, and R. Schober, “An overview three typical fixed PA schemes. This means that they can of sustainable green 5G networks,” IEEE Wireless Commun., vol. 24, harvest appreciable SR performance gains. no. 4, pp. 72–80, Aug. 2017. [4] S. R. Aghdam and T. M. Duman, “Joint precoder and artificial noise design for MIMO wiretap channels with finite-alphabet inputs based on the cut-off rate,” IEEE Trans. Wireless Commun., vol. 16, no. 6, pp. 3913–3923, Jun. 2017. SNR= 20dB [5] X. Chen, X. Chen, and T. Liu, “A unified performance optimization Optimal β (ES) 3.5 for secrecy wireless information and power transfer over interference Proposed Max-P-SAN channels,” IEEE Access, vol. 5, pp. 12 726–12 736, 2017. Proposed CO Method [6] S. Yan, X. Zhou, N. Yang, B. He, and T. D. Abhayapala, “Artificial- noise-aided secure transmission in wiretap channels with transmitter- side correlation,” IEEE Trans. on Wireless Commun., vol. 15, no. 12, 2.5 pp. 8286–8297, Dec. 2016. SNR= 5dB [7] Y. Wu, J. B. Wang, J. Wang, R. Schober, and C. Xiao, “Secure transmission with large numbers of antennas and finite alphabet inputs,” IEEE Trans. Commun., vol. 65, no. 8, pp. 3614–3628, Aug. 2017. 1.5 SNR= 0dB [8] L. Wang, S. Bashar, Y. Wei, and R. Li, “Secrecy enhancement analysis against unknown eavesdropping in spatial modulation,” IEEE Commun. Lett., vol. 19, no. 8, pp. 1351–1354, Aug. 2015. [9] C. Liu, L. L. Yang, and W. Wang, “Secure spatial modulation with a full-duplex receiver,” IEEE Wireless Commun. Lett., vol. 6, no. 6, pp. 0.5 838–841, Dec. 2017. [10] F. Shu, Z. Wang, R. Chen, Y. Wu, and J. Wang, “Two high-performance 0 0.2 0.4 0.6 0.8 1 schemes of transmit antenna selection for secure spatial modulation,” IEEE Trans. Veh. Technol., pp. 1–1, 2018. [11] F. Wu, L. L. Yang, W. Wang, and Z. Kong, “Secret precoding-aided spatial modulation,” IEEE Commun. Lett., vol. 19, no. 9, pp. 1544–1547, Fig. 3. Comparison of the achievable SR for SNR=0, 5, 20dB with same Sep. 2015. configuration as Fig. 2. [12] M. Sadek, A. Tarighat, and A. H. Sayed, “A leakage-based precoding scheme for downlink multi-user mimo channels,” IEEE Transactions on Fig. 3 plots the curves of maximum achievable SR versus Wireless Communications, vol. 6, no. 5, pp. 1711–1721, May 2007. Average Secrecy Rate (bits/s/Hz) Average Secrecy Rate (bits/s/Hz) 5 [13] S. Feng, M. M. Wang, W. Yaxi, F. Haiqiang, and L. Jinhui, “An efficient power allocation scheme for leakage-based precoding in multi- cell multiuser MIMO downlink,” IEEE Commun. Lett., vol. 15, no. 10, pp. 1053–1055, Oct. 2011.

Journal

Electrical Engineering and Systems SciencearXiv (Cornell University)

Published: Aug 2, 2018

There are no references for this article.