ADS-B Anomaly Detection Algorithm Based on LSTM-ED and SVDD

With the rapid development of radio broadcasting technology and information technology, the possibility of ADS-B surveillance system suffering from radio frequency interference and network attack is increasing. Firstly, the types of attacks that ADS-B system is vulnerable to are analyzed, and the abnormal data of the corresponding state are simulated. Then, in order to detect ADS-B data attacks accurately, an anomaly detection algorithm based on Long Short-Term Memory Encoder–Decoder (LSTM-ED) and hypersphere classifier is proposed. Using the normal historical data to train the LSTM-ED model so that the reconstructed values can be obtained. Then, the difference values between the actual values and the reconstructed values are put into SVDD (support vector data description) for training, and a hypersphere classifier that can detect abnormal data is obtained. It is judged whether there is an abnormality according to the set threshold. Experiments show that the LSTM-ED-SVDD model can detect abnormal ADS-B data generated by various attacks and has high accuracy.