Search

Intelligent Connected Vehicles (ICVs), have developed rapidly towards information networking, and they are destined to become an important carrier of future travel. However, they also face some challenges like cyber-security. Controller Area Network (CAN), as a typical communication medium of ICVs, is very vulnerable to cyber-attacks since they lack identity authentication and message encryption. Therefore, how to pinpoint the source of abnormal messages, also namely Electronic Control Unit (ECU) source mapping scheme, is of significance to the in-vehicle communication. However, most existing studies that utilize learning algorithms consider little on the stability of training features. To this end, this paper develops a hybrid-feature-extraction based ECU mapping scheme derived from physical features of network signals. In detail, more stable bit-block groups and mode value of high-level voltages are extracted as training features. On this basis, single-layer Softmax classifier is formulated to accurately identify ECU sources. Finally, the proposed scheme is experimentally validated with eight ECUs in a real-world vehicle. Results show that the average identification accuracy of all ECUs is close to 98%, which has been improved by more than 6% compared with other typical methods. The present study effectively addresses the deficiency of identity authentication in CAN.

•Real-time energy management strategy of EVs is proposed to allocate battery energy.•Nnonlinear model predictive control is deployed to realize the real-time control.•The modified particle swarm algorithm is proposed to solve the multi-objective optimization problem.•Test results of multiple driving cycles have verified the superiority of the proposed energy management strategy.