B-AUT: A Universal Architecture for Batch RFID Tags Authentication

Abstract

RFID tags authentication is always a critical but challenging problem because only checking the EPC is vulnerable to counterfeiting attacks. Past works explore the unique backscat-ter signal features induced by tags' manufacturing imperfection as fingerprints, but fail to support simultaneous authentication for a batch of tags in practice, which is vital for large-scale RFID applications (e.g., warehouse inventory). In this paper, we present a universal architecture, namely B-AUT, to simultaneously authenticate multiple tags even with the same EPC and pinpoint them, which is fully compatible with Gen2 standard and applicable to almost all tags' hardware fingerprints proposed in existing works. The workflow of B-AUT is threefold based on our novel algorithms. First, the extracted fuzzy fingerprint and EPC are jointly exploited to cluster raw data. Second, we extract the tags' fine-grained fingerprints for genuineness validation and obtain the invalid clusters. Third, we harness localization methods to match the invalid cluster to dubious tags and further conduct small-scale re-validation to pinpoint the counterfeit tags. We have implemented a prototype of B-AUT and evaluated it in extreme cases. Experiment results demonstrate that B-AUT can maintain nearly the same authentication accuracy as that of separate authentication and reduce the time overhead by 43.3%. Moreover, the pinpointing accuracy can reach as high as 92.8%, regardless of tags' total quantities or tag models.