Research on a Joint Blind Equalization Method for Device Fingerprint Extraction at Low SNR

The channel characteristics severely affect the accuracy of device fingerprint extraction. To address this, we propose a joint blind equalization algorithm for low Signal-to-Noise Ratio (SNR) environments, aiming to weaken channel interference and extract purer Radio Frequency Fingerprints (RFF). First, we enhance the Modified Constant Modulus Algorithm (MCMA) by incorporating Simulated Annealing (SA), forming MCMASA to reduce steady-state errors and avoid local optima. Then, to further improve noise resistance under low SNR, we combine MCMASA with an adaptive step-size Decision-Directed LMS (ASSDDLMS), forming the Joint Blind Equalization Algorithm (JBEA). Experiments show that JBEA can reaches -10 dB at SNR $=8 \mathrm{~dB}$. Applied before RFF extraction, the algorithm improves device identification accuracy from $54.2 \%$ to $80 \%$, demonstrating strong effectiveness in mitigating channel impact and enhancing authentication performance.