Vehicle Anomaly Detection by Attention-Enhanced Temporal Convolutional Network

Intelligent transportation system (ITS) is the general trend in the field of transportation. As an important development direction to promote the realization of ITS, connected and automated vehicles (CAVs) have attracted extensive attention from many scholars. However, most CAVs system are vulnerable to various spoofing attacks. To solve this problem, an AttentionEnhanced Temporal Convolutional Network (TCN) for anomaly detection of vehicle data is proposed in this paper. Firstly, the Squeeze-and-Excitation Networks (SE-net) is used to automatically obtain the importance degree of each feature channel, and then according to this importance degree, the useful features are promoted and the features that are not useful for the current task are suppressed. Then, the multi-layer TCN model with attention branch is used to fully extract the data features, and the abnormal detection results are obtained. To verify the proposed model, we conducted experiments on the SPMD dataset. The experimental results show that Attention-Enhanced TCN has good detection performance for vehicle abnormal data, which is superior to the current state-of-the-arts.