Radio Frequency Fingerprinting based on Circulant Singular Spectrum Analysis

Radio Frequency fingerprinting is a technique to identify wireless devices on the basis of their intrinsic physical features, which can be extracted by signals generated during transmission. In recent times, a number of studies have explored this identification approach using a variety of techniques, features and machine learning algorithms. This paper proposes a new technique based on the adoption of Circulant Singular Spectrum Analysis (CSSA), which is a recent extension of Singular Spectrum Analysis (SSA). To the knowledge of the author, this is the first time in literature that CSSA is applied to the problem of Radio Frequency fingerprinting. The proposed technique is applied to a recently published data set with signals extracted from 16 Bluetooth wireless devices. The experimental results show that the CSSA based approach outperforms significantly, in terms of accuracy, the original SSA and the other approaches based on the time domain and frequency domain features usually adopted in literature.