Research On Abnormal Detection Of Intelligent Bus CAN Bus

With the continuous breakthrough of automotive technology,the degree of automobile intelligence will continue to improve,and intelligent cars will gradually be infiltrated by Internet technology,which has greatly enhanced the driving experience.However,the intelligent configuration in the car,such as the remote diagnosis system,is connected to the Internet so that the key data in the car(battery charge state,vehicle location,etc.)are exposed on the Internet.Most of the critical data are transmitted through the CAN bus in the vehicle,but at the beginning of the design of the CAN bus,the concept of network security is relatively weak,and the security factors are not fully considered,which makes it extremely easy to attack,thus stealing vehicle information and even manipulating the vehicle,resulting in the failure of key components in the vehicle,resulting in serious security consequences.Therefore,the research of CAN bus anomaly detection is of great significance to the network security of intelligent connected vehicles.Aiming at the security threat of intelligent bus CAN bus and the problems existing in current anomaly detection technology,such as low detection accuracy,false alarm,high computing consumption and so on,this paper puts forward new methods of CAN bus anomaly detection according to the data flow characteristics and data domain characteristics of CAN bus.The feasibility and effectiveness of the scheme are verified by experiments.In this paper,the research on abnormal detection of intelligent bus CAN bus is carried out as follows:(1)the characteristics of CAN bus protocol are analyzed in detail,the weak points of CAN bus are analyzed from the design point of view,and the network architecture of intelligent bus is analyzed,and the attack process and possible attack methods are described in detail.According to the concept,technical characteristics and application fields of anomaly detection,the difficulties and challenges of CAN bus anomaly detection technology are analyzed in detail,and the CAN bus anomaly detection flow of intelligent bus is designed.(2)According to the traffic characteristics of intelligent bus CAN bus,the influence of attack behavior on traffic is analyzed,and a traffic anomaly detection method based on KNN time series is proposed.By inserting three kinds of abnormal ID into the original traffic data,the feasibility of this method for message injection attack detection is verified by simulation.(3)according to the characteristics of message data domain of intelligent bus CAN bus,the influence of attack on data domain is analyzed,and a data anomaly detection method based on OCSVM is proposed.The cross-validation training is carried out on the 10-hour driving data divided by attributes,and the optimal model parameters are obtained.(4)the simulation experiment model is built by using CANoe software.A message injection attack is implemented on a real vehicle to verify the performance of the proposed traffic anomaly detection method under multivariable control.Tampering the data from instrument and motor node,it is verified that the proposed data anomaly detection method has a better detection effect than other existing CAN bus data anomaly detection methods(HMM,FURIA,Hamming distance).