An Anomaly Detection Model for ADS-B Systems Based on Improved GAN and LSTM Networks
Xinying Guo, Chunhua Zhu, Jing Yang, Yan Xiao · 2021 IEEE 21st International Conference on Communication Technology (ICCT) · 2021
Automatic Dependent Surveillance-Broadcast (ADS-B) is an important part of the next generation air transportation systems, but its protocol does not provide relevant authentication and data encryption, so it is extremely vulnerable to various spoofing attacks. Therefore, it is necessary to develop a set of ADS-B data anomaly detection models. According to the strong temporal dependence and fast update rate of ADS-B data, we initiatively introduce a novel anomaly detection model based on Improved Generative Adversarial Network and long short-term memory networks (IGAN-LSTM). In the model, IGAN mainly improves the generator$G$architecture of GAN through upgrading the commonly used encoder architecture to encoder-decoder-encoder architecture, which captures the training data distribution within both data space and latent space. IGAN then utilizes encoding losses based on latent spatial features to determine whether a data sample is anomalous or not. At the same time, to model ADS-B data with the temporal dependence, we use the LSTM networks as the main architecture of the model, which is able to make use of long-term temporal dependence and avoid the vanishing gradient problem during training. Experimental results show that the proposed model has a higher area under the receiver operating characteristic curve and the precision recall curve than the other baseline methods from the literature, and reveal that the model has better performance of anomaly detection of ADS-B data.