Leveraging V2V Technology to Mitigate GPS Spoofing Attacks in Autonomous Driving

We present a V2V-enabled defense against GPS spoofing that combines Density-Based Spatial Clustering of Applications with Noise (DBSCAN) gating, fuzzy trust weighting, and a Root Cause Analysis (RCA)-hybrid method using a compact Bayesian Belief Network (BBN) to estimate neighbor trust and prune likely attackers before fusion. On real traffic trajectories with injected spoofing, plain averaging offers modest gains, fuzzy improves further, and the RCA-hybrid achieves the largest localization error reduction (up to 50–60%) while remaining real-time. We also evaluate robustness to V2V impairments (10% drops, 50ms delay), perform sensitivity analysis over key hyperparameters, and provide model interpretability using SHapley Additive exPlanations (SHAP) on observable residual features. Overall, the framework yields interpretable, communication-robust trust fusion for practical connected and autonomous vehicle (CAV) localization under spoofing.