Radio Frequency Fingerprint Acquisition and Identification for Small Sample DMR Signals Under Blind Synchronization

Radio frequency Fingerprint (RFF) is an emerging security technology that identifies devices by utilizing the RF impairments of hardware devices, with the acquisition of fingerprint signals serving as the foundation of fingerprint identification. Under the condition of blind synchronization, this paper proposes a fingerprint signal acquisition method based on time-frequency energy spectrum detection and a fingerprint identification method based on wavelet transform for small sample signals. The datasets of transient fingerprints and steady-state fingerprints are constructed through data preprocessing operations such as frequency offset compensation and power normalization. Utilizing signals collected from Digital Mobile Radio (DMR) terminals, this paper investigates the influence of sampling rate, classification model and other factors on the identification performance under the Over-the-Air (OTA) acquisition and direct RF acquisition modes. The results indicate that under the OTA acquisition with a sampling rate of 25 kHz, the random forest classification algorithm can achieve 99% identification accuracy for transient fingerprints, whereas further increasing the sampling rate offers no substantial improvement in RFF identification accuracy.