关键词: 广播式自动相关监视(ADS-B), 异常检测, 神经网络, 双向门控循环单元(BiGRU), 支持向量数据描述(SVDD)

Abstract: As a new generation ATM monitoring technology, ADS-B is vulnerable to cyber attack because it broadcasts data in a plaintext format. To solve the security issues of ADS-B, this paper considers the time correlation of ADS-B data. Firstly, the BiGRU (Bidirectional Gated Recurrent Unit) is used to predict the ADS-B data, obtaining the predicted value. Then, the difference of the predicted and the actual values is put into SVDD (Support Vector Data Description) and a hypersphere classifier which can be trained to detect the ADS-B anomalous data. In addition, a suitable sliding window is selected to ensure the accuracy of anomaly detection and reduce the training time of the BiGRU neural network. The experimental results show that the BiGRU-SVDD can detect the ADS-B anomalous data from random position deviation attack, height deviation attack, DOS attack, replay attack, and data deletion attack. Moreover, compared with other machine learning and deep learning methods, the BiGRU-SVDD anomaly detection model has better accuracy and adaptability.

Key words: ADS-B, anomaly detection, neural networks, BiGRU, Support Vector Data Description (SVDD)