Data-and-Channel-Independent Radio Frequency Fingerprint Extraction for LTE-V2X

Radio frequency fingerprint (RFF), derived from transmitter hardware impairments, has become one of the most secure solutions for physical layer authentication due to its copy-proof and forgery-proof. Focusing on the challenging vehicle-to-everything (V2X) scenarios, this paper conducts the first exploration of out-of-band RFF extraction based on the analysis of RFF distribution in broadband systems, and proposes a data- and channel-independent RFF extraction method. Specifically, inspired by voiceprint extraction, we decouple the pure RFF from the wireless signals containing data and channel response by task-oriented filtering as well as homomorphic processing. The method exploits the fact that the RFF is long-time invariant compared to the rapidly changing channel. By comparing the extracted RFF of the same device in both wired and wireless scenarios, the channel independence of the proposed scheme is demonstrated. Further, the feasibility and interpretability are given by Wasserstein distances and feature contribution quantification using SHapley additive exPlanation (SHAP), respectively. Extensive experiments involving different wireless environments and moving speeds, showing that the proposed scheme could achieve test accuracy of 97.30% for thirteen V2X devices of the same model from the same manufacturer, and has good generalization ability to out-of-library devices and other deep learning models.