Channel-Robust RF Fingerprint Authentication for LTE-V2X via Channel-Adversarial Siamese Networks

Radio Frequency Fingerprinting (RFF) has emerged as a promising technique for physical-layer authentication, utilizing hardware-specific characteristics for secure identification. However, in L TE-based Vehicle-to-Everything (V2X) systems, the high-speed mobility of vehicles and varying multipath channels introduce significant challenges to RFF reliability. This paper proposes a robust RFF extraction and authentication framework tailored for L TE V2X applications. In the extraction phase, Demodulation Reference Signal (DMRS)-based Channel State Information (CSI) is transformed in the logarithmic domain to enhance learning of device-specific features while minimizing channel interference. For authentication, we introduce a Channel-Adversarial Siamese Network (CASNet), which employs gradient reversal layers to map signal fingerprints from different channels into a unified feature space, improving recognition accuracy under diverse channel conditions. Experimental results demonstrate that CASNet significantly outperforms conventional methods, achieving high accuracy and stability across both stationary and mobile scenarios, proving effective for LTE V2X authentication in real-world environments.