CNN-Based Physical Layer Authentication Method for Underwater Acoustic Sensor Networks

As Underwater Acoustic Sensor Networks (UASNs) find increasing utility in security, monitoring, and exploration applications, robust node authentication becomes crucial. We propose a novel physical-layer cooperative authentication method using convolutional neural networks (CNNs) that leverages spatial dependency of underwater acoustic channels. Our two-stage framework first allows trusted nodes to collect signals and learn unique channel characteristics from authorized transmitters through CNN-based analysis of channel impulse responses (CIRs). In the online stage, trained models authenticate incoming signals in real-time, with a central sink node combining inputs from multiple trusted nodes to enhance resilience against localized attacks. Bellhop simulations show that the proposed CNN-based method accurately detects malicious packets, outperforming SVM approach that relies on standard authentication features.