A New Precoding for the MIMO Gaussian Channel with Multi-Antenna Eavesdroppers

Abstract

The problem of maximizing the secrecy rate of multiple-input multiple-output multiple-eavesdropper (MI-MOME) channels with arbitrary numbers of antennas at each node is studied in this paper. It is shown that optimal precoding can be formed using a rotation matrix in an n t -dimensional space where n t is the number of antennas at the transmitter. Next, a gradient-descent based method is developed to find the rotation and power allocation parameters. The proposed rotation-based method can be applied to the MIMOME channel with arbitrary numbers of antennas at each node and outperforms state-of-the-art solutions. In particular, it achieves higher secrecy rates than generalized singular value decomposition (GSVD)-based precoding. Further, GSVD-based precoding is used to initialize the rotation-based precoding and this combination provides an efficient approach to find a near-optimal transmit strategy for the MIMOME channel. The rotation-based precoding can be applied to related problems in multi-antenna channels. Numerous simulation results show that the rotation-based method outperforms existing methods in terms of performance and computational complexity.