Intelligent vehicle knowledge representation and anomaly detection using neural knowledge DNA

Abstract

Abstract The knowledge expression and anomaly detection of the intelligent vehicle are studied. Firstly, an Intelligent Vehicle Knowledge Management Framework (IVKMF) is proposed to describe the whole knowledge of Intelligent vehicle in a unified structure. Then, the detection rules are divided into simple rules based on features or statistical characteristics and complex sequence rules based on neural networks. The former can effectively detect the specific CAN command and flooding, replay attacks, the latter uses neural networks to learn the characteristics of CAN commands, which can effectively detect the complex attacks. The simple detection rules are standardized by SOEKS knowledge expression and the complex detection rules storage in Neural Knowledge DNA framework. With this unified knowledge expression, the detection rules can be shared and inherited easily. A secure gateway for intelligent vehicle is also designed. The gateway is placed between the external network and the vehicle bus network and it prevent all suspicious external data. The simulations of real car data prove the feasibility of the methods.