Abstract:

The openness of UAV Command and Control （C2） Link makes it vulnerable to non-granted attacks， leading to the risk of UAV loss of control， crash and even malicious attacks on third parties. Focusing the risk of Denial of Service （DoS） attacks in C2 link and considering the lack of actual detection dataset conditions， this paper proposes an attack detection method based on the Bidirectional Temporal Convolutional Network （BiTCN）， capable of multi-source feature fusion capability. This method constructs a detection dataset by integrating information features from both information features of network data and physical data. Timestamp alignment and forward padding are employed to solve the asynchronous problem of network and physical data streams. The experiment was conducted using a complete dataset and datasets with data missing rates of 5%， 15%， 30%， 40%， 50%. The BiTCN model was utilized to capture contextual information through a bidirectional mechanism， enabling feature extraction and classification to detect DoS attacks. The proposed method was validated on a real drone attack dataset， and the results showed that： Compared with detection models based solely on network data or physical data， the proposed method achieved higher accuracy （97.8%）， recall （95.9%）， F1 score （97.8%）， and AUC （0.997） than single-dimensional data detection models. Compared with traditional FNN， 1D-CNN， LSTM， and GRU detection models， the proposed method maintains high detection accuracy even under 40% data missing conditions.

Key words: UAV C2 links, DoS attack detection, bidirectional temporal convolutional networks, fusion of cyber and physical data, attack detection method