Exploiting the Potential Anomaly Detection in Automobile Safety Data with Multi-type Neural Network

As an internal network widely used in automobiles, the automotive CAN bus network lacks effective security protection mechanisms and is vulnerable to network hackers, posing a serious threat to the safety of vehicles and drivers. The automotive intrusion detection system provides effective protection for the security of the automotive CAN network. To address the shortcomings of current intrusion detection algorithms, such as long application time and incomplete detection types, GIDPS and TIDPS models are proposed to perform supervised multi-classification experiments on vehicle intrusion data. Then, the above model is migrated to the ROAD dataset for verification, and the advantages of the new model in terms of time and accuracy compared with the old model are analysed based on the results. The proposed GIDPS and TIDPS models achieve better results than previous models in terms of synthesis. The new models provides a certain reference value for improving the level of automotive network security. They could be applied to domestic or cross-border automotive markets.