Variational AutoEncoder-Based Radio Frequency Fingerprinting for GPS Anti-spoofing

This paper addresses the spoofing detection issue for Global Positioning System (GPS) signals based on radio frequency fingerprinting (RFF). Different from existing works, this paper utilizes a new RF feature that represents the combination of various fundamental RF features like IQ imbalance and phase errors within the modulation domain. This new feature is modeled as a two-dimensional (2D) grayscale image generated by displaying the IQ sequences of the preambles of k GPS subframes after the despreading (removing carrier and PRN) process in the tracking phase onto the IQ plane. Using this RF feature, we propose a new GPS spoofing detection approach based on the Variational AutoEncoder (VAE), which takes a grayscale image as input and returns a decision indicating if the signals that generate this grayscale originate from a genuine GPS satellite or a spoofing satellite. The dataset for training and testing was collected by a universal software radio peripheral (USRP) in approximately 60 days from April to November 2023. The spoofing signals were generated by another USRP using the open-source software gps-sdr-sim. The experimental results showed that the proposed method achieves a detection accuracy of more than 99% for spoofing signals with a false positive rate of less than 2% for genuine signals.