Research On Key Technologies For Intrusion Detection In Intelligent Networked Automotive CAN Bus

With the rapid development of Internet technology,various advanced technologies are gradually being used in automobiles,and smart connected vehicles are becoming more widely known to the public.However,while smart connected vehicles offer numerous convenient features and improved driving experiences,they also bring about increased security risks.The use of advanced technologies has led to more external interfaces being connected to the vehicle network,and each external interface,while providing diverse functions,also becomes a potential avenue for attacks on vehicles.This makes smart connected vehicles more vulnerable to hacking,resulting in risks such as personal information leakage of the driver and loss of vehicle control,leading to unprecedented security risks.As the communication bridge of the in-vehicle network,the Controller Area Network(CAN)bus has become the de facto standard for in-vehicle networks and is widely used in various vehicles.At the same time,almost all attacks on vehicles involve the CAN bus,and the ultimate target of attacks is the CAN bus,highlighting its importance.However,due to the lack of consideration for security issues during the initial design of the CAN bus,it is vulnerable to exploitation by intruders.Therefore,researching relevant defense technologies against CAN bus attacks in smart connected vehicles is of significant importance.This thesis proposes two intrusion detection methods for protecting the security of the CAN bus and implements an intrusion detection system based on these two methods.The main research work in this thesis can be divided into three parts:(1)CAN bus intrusion detection method based on ResNet-LSTM.To address the vulnerability of the CAN bus and better capture the temporal features in the CAN bus dataset,this thesis designs a CAN bus intrusion detection method called "HybrIDS".This method combines the advantages of ResNet neural network and LSTM neural network,allowing the model to better extract temporal features from CAN data frames and thus improve the model’s performance.To validate the performance of the model,experiments are conducted on different datasets and compared with four existing models.Experimental results show that the proposed intrusion detection model outperforms the compared models in terms of performance.(2)Propose a CAN bus intrusion detection method based on biological features.In previous research on CAN bus intrusion detection,researchers have not considered the driver’s biological features of driving behavior.This thesis demonstrates through experiments that the driver’s driving behavior can also serve as a learning feature for the model in the CAN bus data.The introduction of this feature can help improve the performance of the model.To further enhance the practicality of the CAN bus intrusion detection method,this thesis uses a new CAN bus dataset with driver’s biological features,which can better reflect the actual usage of the vehicle.Through experimental analysis,it is shown that this method can fully utilize the driver’s biological features in the dataset and still maintain good detection performance under stricter detection constraints.This thesis is the first to incorporate the driver’s biological features into CAN bus intrusion detection in existing research.(3)Design and implement a CAN bus intrusion detection system.Based on the research on the two intrusion detection methods mentioned above,this thesis develops a CAN bus intrusion detection system.The system allows developers to freely select models and datasets for training,and it has a model training parameter library to guide developers in optimizing the models.Moreover,the system can visually display the results of experiments and the evaluation metrics of the models.