Real-Time GNSS Spoofing Detection for Autonomous Vehicles: An Attention-Based Autoencoder Approach

With the rapid evolution of autonomous vehicles (AVs), ensuring reliable navigation has become paramount, especially against threats like Global Navigation Satellite Systems (GNSS) spoofing. This paper presents an attention-based Autoencoder approach for real-time GNSS spoofing detection in AVs. The proposed method leverages data from multiple sensors, including IMU, GNSS, and LiDAR, fully utilizing the redundancy and correlations among them. By integrating a multi-head attention mechanism into the Autoencoder, the model can thoroughly capture and analyze the complex relationships within sensor data, enhancing its capability to promptly and accurately identify spoofing attacks. Field experiments demonstrate the effectiveness of the proposed method in achieving a high detection rate and short detection time, highlighting its potential for practical deployment in AV applications.