Design And Implementation Of Intrusion Detection System For Connected Cars Integrating Host-level And Network-level Detection

In recent years the concepts of driverless cars,intelligent vehicles and intelligent transportation have sprung up.A vehicle has changed from a traditional relatively closed embedded system into the new one that interacts with the external environment more fre-quently and besides the communication between vehicles and people,it contains the ex-change between vehicles and vehicles,even vehicles and infrastructures through wireless network or mobile network.While the Internet of vehicles brings people convenience,it also brings latent dangers about network security.Most of vehicle terminals lack effec-tive cyber security defenses.Attacks against cars,such as stealing personal information and remote control,will pose a serious threat to personal life and property.An intrusion detection system is an effective way of cyber defense.To solve the security problem of this network,this paper proposes an intrusion detection system based on neural network integrating host-level and network-level intrusion detection.The main contributions of this work are summarized as follows:1.Design of network-level intrusion detection.The Internet of vehicles can be di-vided into the network based on TCP/IP protocol for external communication and the one based on CAN bus in which each electronic control unit of the vehicles communicates with each other.Based on convolutional neural network and auto-encoders,a complete network-level intrusion detection system is designed from three parts:data acquisition,data preprocessing and detection engine.2.Design of host-level intrusion detection.On the vehicle terminal based on Linux system,the system call executed by the process reflects the interactive behavior between the program and the operating system.With the system call sequence of the process as the data source and the word embedding technology for preprocessing,the host-level intrusion detection engine based on convolutional neural network is completely designed.3.Implementation of intrusion detection system for Internet of vehicles.The intru-sion detection system includes data acquisition module,preprocessing module,training module,detection engine module.They are deployed on the Linux-based vehicle termi-nal.Besides,the model training part adopts the way of offline training and the server completes the training part and transmits the detection model to the vehicle terminal.To evaluate the system,we carry it out on public intrusion detection datasets,the experi-mental results show that the system can effectively protect information security of vehicle terminals.