A multi-attack intrusion detection model based on Mosaic coded convolutional neural network and centralized encoding.

Area of Science:

• Cybersecurity
• Automotive Engineering
• Network Security

Background:

• Internet of Vehicles (IoV) development increases cyberattack risks to automotive control area networks (CAN).
• Existing intrusion detection systems often fail to address simultaneous, multi-type attacks on CAN buses.
• There is a critical need for robust methods to detect combined intrusions in real-world scenarios.

Purpose of the Study:

• To propose a novel method for detecting multiple simultaneous intrusions on vehicle CAN buses.
• To address the limitations of current methods that focus on single attack types.
• To enhance the security of connected vehicles against complex cyber threats.

Main Methods:

• Developed a Mosaic coded convolutional neural network (CNN) for intrusion detection.
• Converted one-dimensional CAN IDs into a two-dimensional data grid for enhanced feature extraction.
• Utilized a centralized coding method to improve the model's discrimination capability.
• Trained and tested the model using four distinct attack types and all their combinations.

Main Results:

• The proposed method successfully detected all combinations of intrusion types.
• The Mosaic coded CNN effectively extracted data and temporal characteristics from CAN IDs.
• The model demonstrated very high and stable performance in identifying multi-intrusions.
• Centralized coding significantly increased the model's discrimination capability.

Conclusions:

• The developed single model is capable of detecting various combinations of simultaneous CAN bus attacks.
• This approach offers a significant advancement in automotive cybersecurity for the Internet of Vehicles.
• The method provides a robust solution for real-time, multi-intrusion detection in connected vehicles.