GAN-LSTM-Based ADS-B Attack Detection in the Context of Air Traffic Control

Abstract

Automatic Dependent Surveillance-Broadcasting (ADS-B) is a surveillance technology strongly promoted by the International Civil Aviation Organization. It has been widely used in commercial and general aviation, to provide support for the normal operation of air traffic control (ATC) in commercial aviation. However, the openness of the ADS-B protocol makes it extremely vulnerable to cyber attacks. Previous research did not specifically consider the application scenarios of ATC in commercial air transport, and there is a problem of low attack detection rates. This paper focuses on ADS-B attack detection under the background of ATC. We combine the flight plan with ADS-B information to construct an airspace flight image stream and process the image stream using a generative adversarial network-long short-term memory (GAN-LSTM) model to predict future images. Then, we identify abnormal images based on the normalized cross correlation and mark anomalous targets. Our method can quickly locate anomalous targets in the controlled airspace. Based on real flight data, abnormal data for various malicious attacks were forged. We evaluated the detection performance of the method through a confusion matrix and several performance indices. The final experimental results showed that our detection scheme exhibited good detection performances for various attacks, with an average detection accuracy of 92.3%, a false positive rate of 11.9%, and a false negative rate of 6.2%. This approach guarantees the information security of ADS-B, thereby improving the operational security of ATC.