New Method Detects Gradual GNSS Spoofing in Autonomous Driving.

Accurate Global Navigation Satellite System (GNSS) localization is crucial for autonomous driving safety, but it's vulnerable to spoofing attacks. Gradual and subtle attacks are particularly challenging to detect because individual GNSS readings may seem plausible while the discrepancy between GNSS-derived displacement and onboard motion slowly grows. To address this, researchers developed a causal high-order liquid evidence framework. This method first creates a physics-guided residual by comparing GNSS-implied movement with vehicle motion data. It then processes separate evidence streams for the residual's level and its first and second-order variations, using adaptive liquid encoders. These temporal states are hierarchically coupled to predict spoofing using only current and past observations. Experiments on real-world datasets showed the method achieved superior F1-scores, detecting spoofing transitions rapidly. The code and datasets are publicly available.