Fast secrecy rate optimization for MIMO wiretap channels in the presence of a multiple-antenna eavesdropper
Introduction
Wireless communication technologies have progressed rapidly due to the open and shared nature. However, the broadcast characteristics of wireless communications make it difficult to prevent transmitted signals from unintended recipients on the communication link. Therefore, information security has become an essential task for a wireless communication system due to the proliferation of privacy-sensitive wireless services (see, e.g., Shiu et al. [1], Liu et al. [2], Poor and Schaefer [3], Wang et al. [4], and the references therein).
Traditionally, information security relies on cryptography mechanisms implemented in media access control (MAC) and the upper-layer communication protocol [5], [6]. However, due to the emergence of new encryption cracking algorithms and the drastically increasing computing power of wireless devices, traditional encryption by secret keys is challenged [7], [8], [9]. As an alternative, physical layer security can also guarantee confidentiality of the messages transmitted from one transmitter to the legitimate receiver. Remarkably, the theoretical foundation for physical-layer security has been established in the seminal works including [10], [11], [12], [13].
The concept of “wiretap channel”, proposed by Wyner in [11], is the one of the basic physical layer models that capture the nature of communication security. Moreover, the notion of secrecy rate was introduced as a popular metric to quantify the achievable maximum information rate from the source to the destination. Subsequent works extended the work of Wyner to non-degraded discrete broadcast channels [12] and applied them to the basic Gaussian channel [13].
Substantial research efforts have been dedicated to the secrecy rate analysis of multiple-antenna wiretap channels from an information-theoretic perspective (see, e.g., Khisti and Wornell [14], Li and Ma [15], Liu and Shamai [16], Weingarten et al. [17], Li and Petropulu [18], Oggier and Hassibi [19], Fakoorian and Swindlehurst [20], Khisti and Wornell [21], Huang and Swindlehurst [22] and the references therein). In [14], the authors derived a closed-form optimal transmit covariance matrix for multiple-input-single-output (MISO) secrecy channels with multiple-antenna eavesdroppers. In [15], the secrecy rate optimization for MISO secrecy channels with multiple multiple-antenna eavesdroppers was recast as a semidefinite programming (SDP) problem. However, the proposed design therein cannot be extended to a multiple-input-multiple-output (MIMO) Gaussian wiretap channel because of the non-convexity associated with this channel. The authors in [16] characterized the MIMO secrecy capacity using an information-theoretic approach motivated by Weingarten in [17]. In [18], a fixed-point based secrecy rate maximization algorithm was proposed for the MIMO wiretap channel to compute the optimal input covariances under the average power constraint. However, the convergence of the proposed algorithm is not guaranteed. In [19], Khisti and Wornell obtained a closed-form expression of the secrecy capacity for MIMO wiretap channels in the case of high signal-to-noise-ratio (SNR). To gain more insights, the authors in [20] revealed the relationship between the rank of the optimal transmit covariance matrix and the channel conditions at the legitimate receiver and the eavesdropper. A semi-closed-form solution is proposed for the case of full-rank covariance matrix. Moreover, when the solution is not a full rank matrix, there exists an equivalent channel with a full rank solution to achieve the same secrecy capacity. However, finding the equivalent channel is intractable. To further enhance the secrecy rates, relays and cooperative jamming have been introduced as a promising technique in wireless networks. They can protect the legitimate users against being intercepted by the eavesdroppers [23], [24], [25], [26], [27]. Moreover, transmitting artificial noise along with the useful data signals is also proposed to improve the physical layer security [28], [29], [30], [31], [32].
In this paper, we consider the secrecy rate maximization problem for MIMO Gaussian wiretap channels, where a multiple-antenna eavesdropper might overhear the communication. In [33], the authors developed an iterative algorithm based on first-order Taylor series expansion to tackle the non-convex problem. The convergence of the developed algorithm is guaranteed. However, at each iteration, the algorithm in [33] needs to solve a determinant maximization (MAXDET) problem, whose computational complexity is very high. Therefore, it is desirable to devise more efficient methods. To this purpose, we propose a simple surrogate function to minorize the non-convex objective function. With the linear minorizer, we can obtain a closed-form optimal solution at each iteration. Moreover, we employ the squared iterative methods (SQUAREM) to speed up the convergence rate of the proposed algorithm. Finally, we adapt the proposed algorithm to deal with the the transmission power minimization problem under the secrecy rate constraint.
The rest of this paper is organized as follows. In Section 2, we establish the signal model for the MIMO wiretap channels and formulate the secrecy rate maximization problem. Section 3 introduces the principle of MM and develops the proposed algorithm based on MM. Section 4 presents an accelerated scheme to speed up the convergence rate of the proposed algorithm. In Section 5, we extend the proposed algorithm to tackle the transmission power minimization problem under the secrecy rate constraint. In Section 6, we present several numerical examples to verify the proposed algorithms. Finally, conclusions are drawn in Section 7.
Notations: We use bold lowercase and capital letters to denote vectors and matrices, respectively. The transpose, the conjugate, and the conjugate transpose are denoted by ()T, ()*, and ()H, respectively. and represent the set of complex numbers, vectors, and matrices with complex-valued entries, respectively. The symbols Tr( · ), ‖ · ‖F, and | · | indicate the trace, the Frobenius norm, and the determinant of a matrix, respectively. stands for the statistical expectation for random variables. denotes that x is a complex Gaussian random vector with zero mean and covariance matrix Σ. Re(x) means the real part of a complex-valued number x. IM denotes the M × M identity matrix. Finally, if we say that D≻B (D⪰B), we mean that is positive definite (semi-definite).
Section snippets
System model and problem formulation
We consider a MIMO communication system with secrecy constraints shown in Fig. 1. In this system, when the legitimate receiver (LR) receives signals from the legitimate transmitter (LT), a passive eavesdropper (PE) is able to overhear the link between the LT and the LR. We assume that the LT and the LR have NT transmit antennas and NR receive antennas, respectively, whereas the PE is equipped with MR receive antennas. Let and represent the complex channel gains, from the LT
Algorithm design
In this section, we propose an MM-based iterative algorithm to tackle the optimization problem in (4). In the proposed algorithm, we introduce an auxiliary variable U, which satisfies . With the auxiliary variable, the positive definite constraint in (4c) can be eliminated. More importantly, we show that the objective function of (4) can be minorized by a simple function of U, resulting in that a closed-form solution can be obtained at each iteration.
To proceed, we briefly describe the
Acceleration scheme
In MM-based algorithms, the convergence speed of the algorithms depends on the tightness of the minorizer. Since the proposed minorizer is a quadratic or a linear function of U, it is much looser than the minorizer in [33]. As a result, the convergence rate of the proposed algorithm might be slow (see, e.g., the numerical examples in Section 6). To speed up the convergence, we employ the acceleration scheme in [45] called SQUAREM to accelerate the proposed algorithm (see also [39], [41], [46]
Transmission power minimization
In this section, we focus on the transmission power minimization problem under a predefined target secrecy rate constraint. Using the signal model in (1) and (2), the transmission power minimization problem under a secrecy rate constraint can be formulated as follows:with being the predefined target secrecy rate for the LR and Ps denoting the transmission power to be minimized. Due to the non-convex secrecy rate constraint (as
Numerical results
In this section, we provide some numerical examples to demonstrate the performance of the proposed algorithms. We assume that the noise variance at the legitimate receiver and the eavesdropper is . As to the termination criterion of the proposed algorithms, we set . Finally, all the analysis is carried out on a standard PC with CPU CoRe i7-7700 3.6GHz and 16 GB memory.
Conclusion
This paper addressed the power-constrained secrecy rate maximization problem for the MIMO wiretap channels in the presence of a multiple-antenna eavesdropper. We developed an efficient algorithm to tackle the non-convex optimization problem. The proposed algorithm was devised under the MM framework, in which we minorized the objective function with a simple surrogate function. Because of the efficiency to solve the surrogated problem at each iteration, the computational cost of the proposed
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This work was supported in part by the National Natural Science Foundation of China under Grant 61671453, and in part by the Young Elite Scientist Sponsorship Program under Grant 17-JCJQ-QT-041.
References (50)
- et al.
Enhancing the secrecy performance of the spatial modulation aided VLC systems with optical jamming
Signal Process.
(2019) - et al.
Reversible data hiding in encrypted binary images by pixel prediction
Signal Process.
(2019) - et al.
A novel image encryption algorithm based on polynomial combination of chaotic maps and dynamic function generation
Signal Process.
(2019) - et al.
An image encryption scheme based on hybridizing digital chaos and finite state machine
Signal Process.
(2019) - et al.
Physical layer security in wireless networks: a tutorial
IEEE Wirel. Commun.
(2011) - et al.
Physical layer security for next generation wireless networks: theories, technologies, and challenges
IEEE Commun. Surv. Tutor.
(2016) - et al.
Wireless physical layer security
Proc. Natl. Acad. Sci.
(2017) An introduction to contemporary cryptology
IEEE Proc.
(1988)- et al.
The wiretap channel with feedback: Encryption over the channel
IEEE Trans. Inf. Theory
(2008) Communication theory of secrecy systems
Bell Syst. Tech. J.
(1949)