Vehicle intrusion detection and prediction model training methods and devices, and storage medium

Abstract

The invention discloses vehicle intrusion detection and prediction model training methods and devices, and a storage medium. The vehicle intrusion detection method can comprise the steps: controller area network (CAN) messages on a CAN bus in the vehicle operation process are monitored; message fragments are generated according to the monitored CAN messages, each message fragment comprises the corresponding N continuous CAN messages, and N is a positive integer larger than one; and aiming at the message fragment generated each time, the next CAN message after the message fragment is predictedthrough a prediction model trained in advance, and according to the predicted next CAN message after the message fragment, whether the vehicle intrusion behavior occurs or not is determined. By applying the vehicle intrusion detection and prediction model training methods and devices, and the storage medium, the accuracy of the detection result can be improved.

Description

[0001] 【Technical field】 [0002] The invention relates to computer application technology, in particular to a vehicle intrusion detection and prediction model training method, device and storage medium. [0003] 【Background technique】 [0004] CAN is the abbreviation of Controller Area Network (CAN, Controller Area Network). The CAN bus protocol has become the standard bus for automotive computer control systems and embedded industrial control area networks. [0005] Through the CAN bus, it can control the engine, gearbox, anti-lock brake system (ABS, Antilock Brake System) and other body safety modules, and share the rotation speed, vehicle speed, oil temperature, etc. to the whole vehicle to realize intelligent control of the vehicle, such as high-speed Automatically lock the door when the airbag pops up, and automatically open the door when the airbag pops up. [0006] Vehicle intrusion usually refers to the behavior of attacking the vehicle by invading the CAN bus, such a...

AI Technical Summary

Problems solved by technology

[0007] The current common vehicle intrusion detection methods are based on rules and signatures, that is, in the form of blacklists, which are very easy to bypass and miss, resulting in relatively low accuracy of detection results and high cost of maintaining rules. In addition, CAN The format of the message is usually not public, and each car company is different. The intrusion detection method based on rules and signatures is very poor in scalability, and needs to be customized for different car companies

Images