Channel-Resilient RF Fingerprint Identification Based on Nonlinear Features With Memory Effect

Eliminating the influence of channel is crucial for radio frequency fingerprint (RFF) identification. The nonlinearity of power amplifiers (PAs) can be extracted independent of channel fading. However, when PAs exhibit memory effect, the separation of nonlinear features from channel fading becomes challenging. In this paper, based on memory polynomial PA model, an RFF nonlinear features extraction method is proposed, which includes three features containing only PA coefficients and other features mixing the PA coefficients with the training symbols. The experimental results show that when training with data received from one location and testing with data received from three other locations, the average identification accuracy can reach up to 92.92% using twenty-two IEEE 802.11 devices.