A lightweight radio frequency fingerprint enhancement and recognition method

Dawei Luo, Xiaonan Luo

Keywords: Deep learning, radio frequency fingerprint recognition, IoT security, wireless network security

Abstract: This study addresses the challenges of source identification in increasingly complex electromagnetic environments driven by the rapid development of wireless communication technologies. With the proliferation of Internet of Things (IoT) devices, traditional identification techniques face significant limitations. To overcome these challenges, a Lightweight Radio Frequency Fingerprint Enhancement (LRFFE) framework is proposed, which innovatively integrates a Time–Spatial–Channel (TSC) triple modeling module. The framework performs comprehensive feature extraction from IQ signals through parallel branches of temporal, spatial, and channel modeling. Experiments conducted on the public ORACLE dataset demonstrate that LRFFE achieves an identification accuracy of 99.16% and a single inference time of 0.062 seconds, significantly outperforming existing mainstream methods. Under a 10 dB signal-to-noise ratio (SNR) condition, the proposed model maintains an accuracy of 97.5%, indicating excellent anti-interference capability. Ablation experiments further verify the effectiveness of each component within the TSC module, showing that the collaborative operation of temporal, spatial, and channel modeling branches enables the model to maintain a very low confusion rate even among devices with highly similar features. This research provides an efficient and reliable solution for individual radiation source identification in complex electromagnetic environments and is particularly suitable for resource-constrained embedded deployment scenarios.

Submission Number: 46