Robust channel comparison metrics for underwater physical layer authentication

Physical layer security enables authentication and privacy in communication systems by harvesting the randomness from wireless channel realizations. For this, it leverages unique channel features that legitimate parties (Alice and Bob) can reliably observe while the same features remain mostly secret to attackers (Eve). In underwater acoustic networks, the minimal correlation of acoustic channels in space provides a promising context for PLS. Our research focuses on implementing authentication via PLS using channel crafting. More specifically, we craft an artificial channel starting from a secret seed, and then precode the transmitted signal by convolving it with the crafted channel. The parameters of the channel impulse response (CIR), namely the delay and amplitude of each arrival, serve as the source of the seed for the crafting process. The receiver performs authentication by comparing the CIR extracted from the incoming signal with an expected CIR. Therefore, we need a mismatch metric that measures tDhe matching of two channels, indicating that such channels have been generated by the two ends of the same link. Based on this metric, a decision algorithm can perform authentication. In this paper, we present the implementation of four mismatch metrics for CIRs, which account for differences that originate from measurement errors, node movement, noise, etc. These disturbances make the delay and amplitude values of the arrivals fluctuate, even to the point of disappearing. Such differences may lead to increased false alarms and missed detections, and therefore a convenient CIR mismatch metric should be robust against them. Specifically, we propose to reward strong channel similarities more than we penalize large differences.