A Privacy Preserving Federated Learning Framework for Securing Consumer Autonomous Vehicles against AI-Enabled GPS Spoofing Attacks

Consumer Autonomous Vehicles (AVs) increasingly rely on Global Positioning System (GPS) signals for navigation and coordination. Due to this strong dependency, consumer AVs become vulnerable to AI-enabled GPS spoofing attacks. By simulating authentic GPS signals, attackers can deceive vehicle navigation systems, causing severe operational and safety risks. Conventional GPS spoofing detection approaches are predominantly centralized and suffer from privacy exposure of sensitive location traces, high communication overhead, and single points of failure. This paper proposes a Federated Learning (FL)-based intrusion detection framework that retains raw GPS data locally at each vehicle while enabling collaborative model training. A lightweight Logistic Regression (LR) model is employed to ensure real-time deployability on resource-constrained, consumer-grade vehicular platforms. The proposed framework is evaluated using a real-world GPS spoofing dataset collected from practical testbed experiments. The system achieves an overall detection accuracy of 97.25%, with high precision and recall across different spoofing scenarios. Further analysis demonstrates robustness under limited adversarial participation during federated training. The proposed solution contributes to the secure deployment of AI-powered consumer AVs and reduces the risk of localized spoofing attacks propagating into system-wide failures.