Ensemble Learning-Enhanced Radio Frequency Fingerprinting

Radio frequency fingerprint identification (RFFI) is a promising physical layer authentication technique that uses minute hardware imperfections as device fingerprints. However, various interferences, especially noise, can distort the RF fingerprints embedded in transmission signals, thereby limiting the overall system performance. This paper investigates performance improvements in RFFI systems through multi-receiver training, ensemble learning, and stochastic weight averaging (SWA), with a particular focus on noise immunity. Our experimental results show three main results: First, multi-receiver training strategies significantly improve system robustness, achieving up to 24% accuracy improvement. Second, ensemble learning with multiple models can improve prediction stability. Third, SWA can achieve comparable performance to ensemble learning with reduced complexity. These findings provide practical insights for improving RFFI system performance in challenging noise environments, contributing to more reliable wireless device authentication.