RFF-NEU: RF Fingerprinting Neural Enhancement Unit for Robust Physical-layer Identification

Radio frequency fingerprinting (RFF) has been widely used to identify wireless devices by leveraging the unique systematic inconsistency imposed on the wireless signal. Recently neural networks have proven to be remarkably effective in RFF identification. However, since RFF is vulnerable to environmental influences, the accuracy of RFF identification can be disrupted tremendously from the wireless channel. Unlike previous studies that sought to improve the capability of RFF identification model, this article proposes a neural network assisted RFF enhancement system which introduces a RFF Neural Enhancement Unit (RFFNEU) at the transmitter, with which the endogenous features of RF signals are more recognizable through various noise channels. To counter its impact on the receiver side, a unique Time Delay Neural Network based demodulation module is constructed to provide a consistent receiving scheme for transmitters with RFFNEUs. The embeddable characteristics of RFF-NEU and TDNN demodulation modules enable the proposed system to be easily integrated into existing communication systems.