Deep Learning-based Channel Prediction for Wireless Physical Layer Security

Future sixth-generation (6G) communication systems requirements highlight wireless security as one of the most sensitive challenges, with physical-layer security (PLS) being considered as one key candidate technology. Regarding the necessity of reliable and up-to-date channel state information (CSI) Deep Learning (DL) algorithms have led to enormous breakthroughs in wireless communications, more precisely, with Recurrent Neural Networks (RNN) approaches aiming to provide improved network performances. For that, channel prediction can be an effective strategy for acquiring the CSI even in channel aging and mobility scenarios. Therefore, based on the predicted CSI parameters, we propose using them as an authentication mechanism and tracking the wireless channel to prevent network attacks from impersonating legitimate users’ communication. Our accuracy results show that by monitoring the evolution of predictions and estimated channel samples for different signal-to-noise ratios (SNR), it is statistically possible to detect illegitimate spoofing signals, through alarms, and complement them with upper-layer authentication protocols to improve security performance in wireless networks.