RF Fingerprint Extraction Method Based on Channel Reciprocity

Broadband signals carry rich spectral information but are highly susceptible to channel interference, posing challenges for RF fingerprint extraction. The extracted fingerprints often retain undesirable channel variations, which degrade device identification accuracy. To address this issue, this paper proposes an RF fingerprint extraction method based on channel reciprocity, where synchronized channel state information (CSI) from both communicating parties at the same time is utilized to remove channel effects. RF fingerprints are extracted in time, frequency, and logarithmic domains, and their performance is systematically evaluated. To validate the proposed method, experiments are conducted using Wi-Fi signals, with ESP32 chips serving as access points (APs) and stations (STAs). A UDP-based backhaul mechanism is employed to obtain channel-reciprocal CSI pairs at the AP side. CSI data is collected in both static and mobile indoor environments, followed by RF fingerprint extraction and classification. Experimental results indicate that logarithmic-domain fingerprints achieve the highest classification accuracy and cross-scenario robustness, reaching 96.8% accuracy under SNRs of 10 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim$</tex> 20 dB.