Radio Frequency Fingerprint Feature Extraction Based on I/Q Data Distribution Features

Abstract

With the development of cognitive radio, more and more radio frequency devices can dynamically switch radio frequency parameters. The reliability of the traditional method of marking the radio signal source in the communication band is greatly reduced. This paper proposes a radio frequency fingerprint feature extraction method based on the accumulated distance of I/Q data components. The RF raw I/Q sample data is collected through the distributed electromagnetic spectrum detection network. The cumulative I/Q distance is extracted from the I/Q sample data collected by multiple detection nodes. The weight distribution of the I/Q components is calculated. Based on I/Q sample data estimation cumulative I/Q distance and reception signal strength parameters, the received signal strength (RSS)-I/Q distance characteristic curve and the corresponding dataset are constructed. The radio frequency fingerprint characteristic curve of the target radio device is established. The optimized Sigmod mathematical model is used as a target function. The RSS-I/Q distance data is set as a sample. The normalized dataset is fitted according to the target function. The parameter values of the optimized sigmod mathematical model are obtained by estimating the fitted curve. The estimated value is taken as the eigenvalue matrix, which is the radio frequency fingerprint characteristic parameter of the target radio equipment. The RF fingerprint feature extraction method proposed in this paper can comprehensively analyze the subtle features of radio equipment from the perspective of signal source. The experiments show that the RF fingerprint feature parameters extracted from the same signal source in different frequency bands, different environments (cement ground, asphalt road, grass, riverside, indoor, etc.), and different temperature and humidity parameters are invariant within a certain error range. The radio frequency fingerprint feature parameters extracted in different signal sources have good distinction in the same frequency band, the same environment, and the same temperature and humidity.