Power Allocation for AN-Aided Beamforming Design in MISO Wiretap Channels with Finite-Alphabet Signaling

In this paper, we address physical layer security in multiple-input single-output (MISO) wiretap channels in the presence of a passive eavesdropper. Artificial noise (AN) is used to provide masked beamforming for degrading the eavesdropper's channel. Existing works based on Gaussian input assumption may lead to substantial secrecy rate loss when considering practical finite-alphabet input. In order to maximize ergodic secrecy rate, a power allocation scheme between the information-bearing signal and AN is proposed. Despite the fact that there is no closed- form expression of mutual information under the constraint of finite-alphabet input, we exploit the relationship between mutual information and minimum mean square error (MMSE) to transform the original problem into a zero-searching problem, which can be solved via gradient search algorithm. Numerical simulations demonstrate that the proposed scheme offers a very good approximation to the optimal performance.