Identification Technology of RKE System Using Multi-dimension RF Fingerprints

Abstract

Recently, remote keyless entry system (RKE system) has been gradually replacing traditional way to unlock car doors for convenience. However, it has been shown that RKE system is vulnerable to cyber attacks including relay attack, amplification attack and cryptographic attack. In order to solve this dilemma, RF fingerprints method was applied to identify car key fobs in this paper. Power spectrum of preamble signal envelope was proposed to extract features while carrier frequency offset and least mean square-based adaptive filter were also used as auxiliary ones. Multi-dimension RF fingerprints were presented in this paper based on three features mentioned above to increase identification accuracy. Support vector machine(SVM) was chosen with 10-fold cross-validation to train classifier model. Corresponding to current research on keyless entry car theft, the classification results in this paper show that signals from various key fobs can be classified with 99.3% accuracy when using Rf fingerprints extracted from multiple features, with false acceptance rate (FAR) of 0.7% and false rejection rate (FRR) of 0.7% under Multiple Discriminant Analysis, Maximum Likelihood (MDA/ML) classifier.