Robust Artificial Noise-aided Secure Communication against ICA-based Attacks
Yun Wen, Makoto Yoshida · 2020
Physical layer security (PLS) technologies have attracted much attention in recent years for their potential to provide information-theoretically secure communications. Artificial Noise (AN)-aided transmission, which can realize secure communications against the external passive eavesdropping, is considered one of the most practicable PLS technologies. In this paper, we reveal that the conventional AN scheme has a severe vulnerability when the eavesdropper adopts Independent Component Analysis (ICA)-based attack, by which the AN-aided transmission will be cracked in a high probability even with most transmission power allocated to AN signals. In order to ensure secure communications, we propose a robust AN scheme focused on the ICA's ability of blind source separation. In our proposed scheme, AN signals are generated with the same modulation scheme and a certain correlation to information signals, thus to introduce irremovable correlation to degrade the ICA's separation performance. Furthermore, we derive the optimal correlation factors from the relation between an information-AN correlation and the ICA's performance. Numerical simulation results with the comparison to the conventional AN scheme show that, our proposed scheme can ensure secure communications by degrading eavesdropper's success probability of attack from 96% to lower than 1%.