Project Abstract

The Control Area Network (CAN) is a widely used communication protocol for in-vehicle communication. In CAN node is referred to as any device that is linked to the bus and can send and receive messages. Each of these nodes is equipped with its controller and transceiver which enables it to interact with the bus. However, since CAN messages are transmitted to multiple nodes on the bus, they do not contain information about their source and destination addresses. This makes it easy for attackers to inject malicious messages. In this project, we have taken can-train-and-test dataset. This dataset is pre-processed, labelled, and organized. This CAN dataset is by four different vehicles produced by two different manufacturers. For each vehicle, there are samples of normal traffic and attack traffic. The can-train-and-test dataset provides nine distinct types of attacks, including denial of service, gear spoofing, and standstill. This data serves as a valuable resource for developing and testing security solutions for in-vehicle networks. The project aims to leverage machine learning techniques to effectively detect anomalies in CAN traffic data, specifically targeting the attack types. By exploring and comparing different algorithms, the project seeks to identify the most effective approach for CAN anomaly detection.

Keywords: Control area network (CAN), Machine Learning, Anomaly Detection

 Conference Details

Session: Presentation Stream 23 at Presentation Slot 9

Location: College 127 at Wednesday 8th 09:00 – 12:30

Markers: Eike Neumann, Julian Hough

Course: MSc Cyber Security, Masters PG

Future Plans: I’m looking for work