Channel-Resilient and Low-Power Radio Frequency Fingerprint Identification for VDES-SAT

Radio frequency fingerprint identification (RFFI) promises to deliver a security authentication paradigm for the VHF data exchange system satellite component (VDES-SAT) with high openness, but most previous works have not integrated the elements of scalability, robustness, and power consumption thus limiting their deployment in reality. In this paper, we propose a channel-resilient and low-power RFFI scheme for VDES-SAT. Specifically, data augmentation enables the robustness of cross-channel RFFI using modest datasets. Given the high latency of satellite links, the channel prediction module is embedded in the data augmentation thus allowing the model to represent the current channel. Furthermore, since frequent authentication can pose a burden on the system with resource constraints, the long short-term memory (LSTM) for channel prediction and convolutional neural network (CNN) for feature extraction are mapped to an adaptive clock-driven/event-driven spiking neural network (SNN) framework. Numerical results show that the proposed RFFI system can yield identification accuracy of over 95.93% at the signal-to-noise ratios (SNRs) above 3 dB on the VDES dataset. The latency can be reduced by 60% compared to artificial neural networks (ANNs) of the same structure and the power consumption is theoretically reduced by more than 70% at 45nm CMOS process.