A novel artificial noise aided security scheme to resist blind source separation attacks

In this paper, we discuss the physical layer security from a new angle of view and propose a novel approach to resist some attacks in signal processing perspective. The scenario considered in this paper is that the eavesdropper has the similar channel performance compared to the legitimate receiver. We design the optimal artificial noise (AN) to resist the attacks of the eavesdropper who uses the blind source separation (BSS) technology to reconstruct the secret information. For speech signals, the optimal AN is obtained by minimizing the maximum of the correlation coefficients between the source signal and the received signals at Eve and the correlation coefficients between the source signal and separated results of BSS. For binary phase shift keying (BPSK) signals, we maximize the minimum bit error rates (BERs) of the separated signals and the obtained signals at Eve. Moreover, we consider the AN design from the point of breaking the BSS conditions, and propose a method by changing the correlation coefficient randomly. The simulation results show that the AN we proposed has better performance than that of the white Gaussian AN to resist the BSS attacks for both speech signals and the BPSK signals.