A machine learning-based physical layer authentication with phase impairments

In this paper, we propose a machine learning (ML) based physical layer authentication (PLA) using the physical features of I/Q imbalance, phase noise and carrier frequency offset (CFO) impairments. By examining the phase information in the presence of these impairments, the proposed PLA method is implemented. The system model includes one legal single-antenna transmitter using orthogonal frequency-division multiplexing (OFDM) modulation, one legal multiple-antennas receiver and one external attacker. The comprehensive studies are conducted for three cases phase noise and CFO utilization, I/Q imbalance utilization, and all three impairments utilization. Our simulations show that the PLA accuracy for the mentioned these cases is more than 98% for single antenna at the receiver. The accuracy can be even improved by using more received antennas. Our results highlight that the PLA accuracy is also affected by the number of OFDM subcarriers and the received signal-to-noise-ratio.