Consider a two-way communication scenario where two single-antenna nodes,operating under full-duplex mode, exchange information to one another throughthe aid of a (full-duplex) multi-antenna relay, and there is anothersingle-antenna node who intends to eavesdrop. The relay employs artificialnoise (AN) to interfere the eavesdropper's channel, and amplify-forward (AF)Alamouti-based rank-two beamforming to establish the two-way communicationlinks of the legitimate nodes. Our problem is to optimize the rank-twobeamformer and AN covariance for sum secrecy rate maximization (SSRM). ThisSSRM problem is nonconvex, and we develop an efficient solution approach usingsemidefinite relaxation (SDR) and minorization-maximization (MM). We prove thatSDR is tight for the SSRM problem and thus introduces no loss. Also, weconsider an inexact MM method where an approximately but computationally cheapMM solution update is used in place of the exact update in conventional MM. Weshow that this inexact MM method guarantees convergence to a stationarysolution to the SSRM problem. The effectiveness of our proposed approach isfurther demonstrated by an energy-harvesting scenario extension, and byextensive simulation results.
展开▼