Anti-Clone: A Lightweight Approach for RFID Cloning Attacks Detection

Abstract

Millions of radio frequency identification (RFID) tags are pervasively used all around the globe to identify a wide variety of objects inexpensively. However, the tag cannot use energy-hungry cryptography due to the limit of size and production costs, and it is vulnerable to cloning attacks. A cloning attack fabricates one or more replicas of a genuine tag, which behave the same as the genuine tag and can deceive the reader to obtain legitimate authorization, leading to potential economic loss or reputation damage. Among the existing solutions, the methods based on radio frequency (RF) fingerprints are attractive because they can detect cloning attacks and identify the clone tags. They leverage the unique imperfections in the tag’s wireless circuitry to achieve largescale RFID clone detection. However, training a high-precision detection model requires a large amount of data and high-performance hardware devices. And some methods require professional instruments such as oscilloscopes to collect fine-grained RF signals. For these reasons, we propose a lightweight clone detection method Anti-Clone. We combine convolutional neural networks (CNN) with transfer learning to combat data-constrained learning tasks. Extensive experiments on commercial off-the-shelf (COTS) RFID devices demonstrate that Anti-Clone is more lightweight than the existing methods without sacrificing detection accuracy. The detection accuracy reaches 98.4%, and the detection time is less than 5 $\textrm{s}$.