A Channel Decoupling RFF Extraction System Utilizing Bilateral Reciprocal Channel Information

Abstract

So far, Radio frequency fingerprint (RFF) has proved an effective device identification medium and has been applied to various types of communication systems. Multi-path and time-varying wireless channels, however, always exert a bad impact on the robustness of RFF-relevant methods. For solving this problem, this letter proposes a channel robust RFF extraction system which can decouple RFF from channel characteristics utilizing information of bilateral reciprocal channels. Firstly, a signal acquisition system that can interact with the device to be identified is implemented to obtain uplink and downlink reciprocal channel information. Then we design a Channel Decoupling RFF Extraction Network (CDRFF-Net) which is trained by virtue of plenty of reciprocal Channel State Information (CSI) pairs and can obtain channel robust RFF features. We implemented the proposed method with IEEE 802.11 communication system as a case study. Extensive experiments were carried out using 10 Wi-Fi devices of the same model in different channel environment where both static and moving scenarios were included. The results show that the proposed system presents high accuracy and high robustness towards complex channel environment. The average identification accuracy using one single data frame can achieve 98.2% in static scenarios and can keep as high as 82.8% even when transmitter and receiver are both moving.