A Generalizable Channel-Resilient Algorithm in Radio Frequency Fingerprint Identification

The identification of radio frequency fingerprint (RFF) extracted from wireless signals subjected to channel interference faces a formidable challenge. Traditional equalization-based channel removal methods require known pilot signals and may potentially eliminate RFF features while suppressing channel effects. To address this challenge, the channel divide-and-conquer normalization (CDCN) algorithm is proposed to remove effects of arbitrary channels while preserving RFF features in the sense of cosine similarity. Through cross-scenario experiments on 60 off-the-shelf ZigBee devices in eight different scenarios, the CDCN algorithm with region of interest (ROI) achieves high accuracy of up to 99.98% on ResNet-9 and is verified to be generalized and practical.