Cyber Attack Detection On Self-Driving Cars Using Machine Learning Techniques

Autonomous cars are smart vehicles that are capable of sensing their surroundings and navigating accordingly without or less human involvement. Such cars are vulnerable to different cyber-attacks and hijacks. The paper aims to use machine learning approaches to prevent cyber-attacks. The work aims to locate the false messages from the steering angle sensors or incorrect steering angle that is injected artificially into the Controller Area Network (CAN) bus of the car. The two approaches used to classify these false data are K-Nearest Neighbor (KNN) and Decision Tree algorithms. The paper uses a driving dataset, purposely inserts wrong values, pre-processing, and then classifies it. The results reveal that the used approaches detect incorrect values with an accuracy of up to 81.48% using KNN and 77.99% using Decision tree algorithms.