Secure Communication With Cache-Aided Relaying and Finite-Alphabet Inputs

This paper investigates the secure transmission problem in a cache-aided relaying network, where we assume that the source sends data to the destination via multiple cache-aided relays and the communication from the relays to the destination is wiretapped by an eavesdropper. Unlike the ideal assumption of Gaussian input signals, this paper considers practical finite-alphabet inputs, such as phase shift keying (PSK) and quadrature amplitude modulation (QAM). Our objective is to maximize the achievable secrecy rate by optimizing the beamforming vectors and power control at the source and relays as well as caching coefficient under constraints of transmit power, quality of service, and network congestion. For the formulated nonconvex multi-variable problem, we use the M/G/1 queueing theory to analyze the network congestion performance and reveal that the congestion decreases with increasing caching coefficient and increasing source-to-relay link rate but increases with increasing relay-to-destination link rate. Then, it is shown that the formulated problem can be solved by dividing it into two optimization subproblems. Subsequently, an auxiliary-variable based alternating optimization (AVAO) framework is developed, where the beamforming at each relay is designed by introducing the Moreau-Yosida regularization (MYR) technique and rank-one decomposition procedure (RDP) and the corresponding power control is derived by applying the optimality condition. Also, the complexity and convergence of the proposed framework are analyzed. Simulation results verify the efficacy of the proposed framework. Specifically, the algorithm based on finite-alphabet inputs proposed in this paper exhibits superior performance compared with the algorithm based on Gaussian inputs. Moreover, the secrecy rate increases with the congestion probability tolerance, the caching coefficient threshold, and the number of relays.