Coding-Enhanced Cooperative Jamming for Secret Communication: The MIMO Case

This paper considers a Gaussian multi-input multi-output (MIMO) wiretapchannel with a legitimate transmitter, a legitimate receiver (Bob), aneavesdropper (Eve), and a cooperative jammer. All nodes may be equipped withmultiple antennas. Traditionally, the jammer transmits Gaussian noise (GN) toenhance the security. However, using this approach, the jamming signalinterferes not only with Eve but also with Bob. In this paper, besides the GNstrategy, we assume that the jammer can also choose to use the encoded jammer(EJ) strategy, i.e., instead of GN, it transmits a codeword from an appropriatecodebook. In certain conditions, the EJ scheme enables Bob to decode thejamming codeword and thus cancel the interference, while Eve remains unable todo so even if it knows all the codebooks. We first derive an inner bound on thesystem's secrecy rate under the strong secrecy metric, and then consider themaximization this bound through precoder design in a computationally efficientmanner. In the single-input multi-output (SIMO) case, we prove that althoughnon-convex, the power control problems can be optimally solved for both GN andEJ schemes. In the MIMO case, we propose to solve the problems using the matrixsimultaneous diagonalization (SD) technique, which requires quite a lowcomputational complexity. Simulation results show that by introducing acooperative jammer with coding capability, and allowing it to switch betweenthe GN and EJ schemes, a dramatic increase in the secrecy rate can be achieved.In addition, the proposed algorithms can significantly outperform the currentstate of the art benchmarks in terms of both secrecy rate and computation time.