Anomaly Detection for Automated Vehicles Integrating Continuous Wavelet Transform and Convolutional Neural Network

Connected and automated vehicles (CAVs) involving massive advanced sensors electronic control units (ECUs) bring intelligentization to the transportation system conveniences human mobility. Unfortunately, these face security threats due complexity connectivity. Especially, existing in-vehicle network protocols (e.g., controller area network) lack consideration, which is vulnerable malicious attacks puts people at large-scale severe risks. In this paper, we propose a novel anomaly detection model that integrates continuous wavelet transform (CWT) convolutional neural (CNN) for an network. By transforming sensor signals in different segments, adopt CWT calculate coefficients vehicle state image construction so exploits both time frequency domain characteristics of raw data, can demonstrate more hidden patterns events improve accuracy follow-up process. Our constructs two-dimensional scalogram (CWTS) utilizes it as input into our optimized CNN. The proposed able provide local transient detect deviations caused by behaviors, effective coping with various anomalies. experiments show superior performance under scenarios. Compared related works, average F1 score are improved 2.51% 2.46%.