Navigation Spoofing and Jamming Signals Identification of UAV Based on Federated Learning

Precise navigation signals are essential for unmanned aerial vehicles (UAVs) to achieve accurate positioning and task execution. However, global positioning system (GPS) signals are highly vulnerable to spoofing and jamming attacks, posing serious threats to flight safety. Therefore, this paper proposes an intelligent detection algorithm based on federated learning (FL) for identifying spoofing and jamming signals in unmanned aerial vehicle (UAV) navigation. Firstly, the algorithm innovatively integrates the local feature extraction ability of the temporal convolutional network (TCN) and the global dependency modeling advantage of the Transformer network, constructing a TCN-Transformer model to capture the features of navigation data efficiently. Secondly, a distributed construction network is adopted, where weight aggregation and updates are performed via federated learning, enabling global optimization without data sharing. Furthermore, an accuracy-weighted aggregation strategy is introduced, dynamically assigning weights based on the detection performance of each client mode. This encourages contributions of high-quality data and computational resources, thereby enhancing the overall model performance. Simulation results demonstrate that the proposed method outperforms traditional algorithms in detecting navigation spoofing and jamming, offering superior detection accuracy and robustness.