A realtime spoofing detection system based on FPGA for GNSS Time

Abstract The Global Navigation Satellite System (GNSS) faces significant threats from spoofing attacks, which can severely compromise positioning and timing accuracy, particularly in systems that require high-precision time synchronization. To identify the deception attacks, a real-time spoofing detection system based on Field Programmable Gate Array (FPGA) is proposed and implemented for GNSS time in this paper. Six parameter values are observed as input features for identification, including pseudorange, carrier-to-noise density ratio C/N0, Doppler shift, positioning error, crystal oscillator temperature, and operational duration. A Backpropagation (BP) neural network model is proposed for identifying signal states and is compared with several widely used and recently proposed models, including SVM, FT-Transformer, and TebNet. Experimental results demonstrate that on the test dataset, the BP model achieved a classification accuracy of 92.59%, the FPR and FNR rates of
approximately 6.57% and 7.14%, respectively, outperforming the SVM, FT-Transformer, and TebNet models across all metrics. Furthermore, the BP model consumed significantly less training time and computational resources, which is quite adaptable to be implemented on the hardwares of FPGA. Consequently, the BP model was deployed on an FPGA-based clock system to evaluate its signal-detection capability under actual spoofing interference conditions. Experimental results show that this clock system achieved 91.46% accuracy in spoofing signal detection while significantly improving detection speed.