Securing ISAC Against Pilot Spoofing Attack: A Deep Plug-and-Play Countermeasure

In this letter, we propose a deep PNP prior approach for designing a robust DFRC precoder that combats the PSA for the first time. Specifically, our precoder design is formulated as a robust optimization problem that maximizes the sensing power while minimizing the eavesdropping SINR simultaneously, even without the need for knowledge of the victim’s and eavesdropper’s CSI. To address this challenging design problem, a deep learning network is exploited to minimize the information leakage, which is subsequently integrated and plugged in as a modular component into an iterative precoder optimization algorithm, under a communication constraint. Simulation results confirm the secrecy performance of our proposed deep prior induced DFRC precoder design approach.