Joint Secure AF Relaying and Artificial Noise Optimization: A Penalized Difference-of-Convex Programming Framework

Owing to the vulnerability of relay-assisted communications, improving wireless security from a physical layer signal processing perspective is attracting increasing interest. Hence, we address the problem of secure transmission in a relay-assisted network, where a pair of legitimate user equipments (UEs) communicate with the aid of a multiple-input multiple output (MIMO) relay in the presence of multiple eavesdroppers (eves). Assuming imperfect knowledge of the eves' channels, we jointly optimize the power of the source UE, the amplify-and-forward relaying matrix, and the covariance of the artificial noise transmitted by the relay, in order to maximize the received signal-to-interference-plus-noise ratio at the destination, while imposing a set of robust secrecy constraints. To tackle the resultant non-convex optimization problem with tractable complexity, a new penalized difference-of-convex (DC) algorithm is proposed, which is specifically designed for solving a class of non-convex semidefinite programs. We show how this penalized DC framework can be invoked for solving our robust secure relaying problem with proven convergence. In addition, to benchmark the proposed algorithm, we subsequently propose a semidefinite relaxation-based exhaustive search approach, which yields an upper bound of the secure relaying problem, however, with significantly higher complexity. Our simulation results show that the proposed solution is capable of ensuring the secrecy of the relay-aided transmission and significantly improving the robustness toward the eves' channel uncertainties as compared with the non-robust counterparts. It is also demonstrated the penalized DC-based method advocated yields a performance close to the upper bound.