Abstract
With the rapid development of the low-altitude economy, the application of UAVs has become increasingly widespread. However, the Global Navigation Satellite System (GNSS), which serves as a critical navigation technology for UAVs, is susceptible to spoofing attacks, severely impacting flight safety and mission accuracy. Many existing detection methods rely on additional equipment or multi-UAV cooperation for attack detection. This paper proposes an attitude angles-based GNSS spoofing detection method, AttDet, which leverages machine learning algorithm to model the change of UAV attitude and GNSS data. The method conducts feature analysis by examining the close dependency between GNSS data and attitude angle calculations, identifying attitude angles as key data for spoofing detection and extracting their statistical characteristics as feature data. Subsequently, flight experiments are designed to collect real-world data, and various machine learning algorithms are employed for training to select the optimal classifier, which is then deployed on the UAV. The system implements data acquisition and data pre-processing on the UAV, enabling online detection of GNSS spoofing attacks. Based on the collected real and spoofed data, the detection rate reaches 98.86%, with an equal error rate (EER) of 1.15%. Experimental evaluation and comparison demonstrate that this method outperforms existing detection approaches.