On Linear Precoding Strategies for Secrecy Rate Maximization in Multiuser Multiantenna Wireless Networks

Revived interest in physical layer security has led to a cascade of information theoretic results for various system topologies under different constraints. In the present paper, we provide practically oriented solutions to the problem of maximizing achievable secrecy rates in an environment consisting of multiple legitimate and eavesdropping radio nodes. By assuming "genie" aided perfect channel state information (CSI) feedback for both types of nodes, we first study two scenarios of interest. When independent messages are intended for all legitimate users (called "broadcast" mode), provably convergent second-order cone programming (SOCP)-based iterative procedure is used for designing secrecy rate maximizing beamformers. In the same manner, when a common message is intended only for legitimate nodes (dubbed "multicast" mode), SOCP-based design is proposed for obtaining linear precoders that maximize the achievable secrecy rate. Subsequently, we leverage the analysis to the more real-world scenario, where the CSI of the malicious nodes has to be somehow estimated and that of the legitimate users is corrupted with unavoidable errors. For this case, we devise provably convergent iterative semidefinite programming (SDP) procedures that maximize the achievable secrecy rates for both the beamforming-based broadcast and the linearly precoded multicast modes. Finally, numerical results are reported that evaluate the performance of the proposed solutions as a function of different system parameters. The results presented in the paper are demonstrated to outperform the ones based on interference alignment strategies. We also ascertain superior performance of the proposed schemes in the realms of real world.