Radio Frequency Fingerprint Extraction Scheme Based on Nonlinear Characteristics in Low SNR Environment

Radio frequency fingerprint (RFF) technique serves as a reliable solution to identity authentication by exploiting the inherent transmitter imperfection. Among various RFF features, nonlinear characteristics have been widely investigated due to their significant recognition performance. In this paper, a feasible RFF extraction and classification scheme is proposed based on the joint nonlinear effect of power amplifier (PA) and root raised cosine filter (RRC). The nonlinearity features are extracted from received signals processed by symbol stacking method to suppress the interference of channel noise. Subsequently, these features are sent to a hybrid classifier. In order to validate the proposed scheme, five PAs are generated by Volterra model for experiment. The classification accuracy of different transmitters can reach 90% at a SNR of -6dB using the nonlinear coefficients as features, which demonstrates the feasibility of our scheme.