Revisiting Backscatter Frequency Drifts for Fingerprinting RFIDs: A Perspective of Frequency Resolution

Physical-layer identification is to exploit inherent randomness introduced during manufacturing to endow a unique fingerprint to a physical entity. A classic fingerprint, called backscatter frequency drift (BFD), was explored a decade ago for the physical-layer identification of RFID tags. The BFD is defined as the offset between the frequency of the backscatter signal actually received from a tag and the requested backscatter link frequency (BLF). As a context-free fingerprint, the BFD is being seriously underestimated due to its terrible performance in tag classification or identification (e.g., accuracy < 30%). In this work, we revisit BFD from the perspective of frequency resolution to pinpoint the reason behind its underperformance. Namely, the low accuracy is not because BFD is insufficiently unique in nature but rather due to the low-resolution measurement of the frequency of a backscatter signal, which is mainly constrained by the current air interface protocol. This challenge is analogous to the recognition of human fingerprints via low-resolution and blurry imaging devices. To address this issue, we propose a practical solution to improve the frequency resolution from kHz to sub-Hz, without any modification of hardware or protocols. The results demonstrate the distinguishability of high-resolution BFD is significantly increased to 99.4% and the identification accuracy is raised to 94% when in the face of 7,135 RFID tags of nine models.