GPS Spoofing Attack Detection for UAVs Based on Flight State Recognition and Trajectory Prediction

To address the security threat posed by Global Positioning System (GPS) spoofing attacks to unmanned aerial vehicles (UAVs), this paper proposes a detection method based on flight state recognition and trajectory prediction. The method first utilizes an autoencoder for feature extraction and flight state classification from the UAV’s sensor data. Building on this, specific multi-head attention-enhanced Sequence-to-Sequence (Seq2Seq) models are employed to perform high-precision trajectory prediction for different flight states, such as climbing, level flight, hovering, and descending. Finally, the deviation between the predicted trajectory and the actual observed values is calculated, and an Isolation Forest algorithm is used to perform anomaly detection on the deviation feature, thereby determining the presence of a GPS spoofing attack. Simulation results demonstrate that the proposed method can effectively improve the accuracy of trajectory prediction and spoofing attack detection.