Electromagnetic Signal Intelligent Identification Based on Radio Frequency Fingerprints

Abstract

Due to the open nature of WIFI connection, it is exposing its private information to the attackers. Traditional WIFI security methods are no longer able to meet the current security needs, and more and more wireless-side physical layer security solutions provide solutions, among which RF fingerprinting is an endogenous security technology with potential. Constructing an effective and accurate method to identify WIFI devices that steal information is a difficulty that today's society needs to face. The main problem is not only that the recognition accuracy is difficult to improve but also the problem of data shortage. In this paper, we first construct a large-scale WIFI real-world measurement dataset. Next, we use PSD and bispectrum features, as well as complex ResNet schemes for WIFI device identification experiments, and compare and analyze them from multiple perspectives. The experimental results show that the proposed algorithm can achieve up to 97% recognition accuracy among 100 devices. Moreover, when the SNR is 0 dB, the complex ResNet method can still achieve 78% recognition accuracy among 100 devices. Finally, this paper summarizes the experimental analysis of the measured dataset and discusses the open issues related to this area.