Channel-Independent Fingerprints for Open-Set Wi-Fi Device Identification Using Wigner-Ville Distribution

Abstract

Radio frequency fingerprint (RFF) is applied for physical-layer authentication and transmitter identification in wireless communications. However, the characteristics of the wireless channel dramatically distorts RFF, and affects the identification accuracy. In this paper, we propose the channel-independent extraction of RFFs. Initially, the signal is represented using the Wigner-Ville Distribution (WVD). Subsequently, to mitigate channel distortions and isolate device-specific features, we perform division across adjacent columnar values. The linear discriminant analysis (LDA) is then applied to condense the RFF features, creating distinct device representations. Finally, the device identification is conducted using K-nearest neighbours (KNN). The proposed scheme is experimentally demonstrated using 10 wireless network interface controllers (WNICs), where an excellent unauthorized device recognition accuracy of 98.59% is achieved, with a false alarm rate of 1.23% and a miss alarm rate of 0.24%. Notably, the high recognition accuracy can be mainly attributed to the channel-independent RFF extraction using WVD, which provides an effective solution to enhance physical-layer security in wireless communications, even for dynamic channel environments.