Detection of Cyber Attacks in Self-Driving Vehicles By Machine Learning

The paper describes three machine learning models for detecting cyber attacks in self-driving vehicles. The aim is recognition of abnormal data, inserted into the CAN bus of self-driving car. Two of the models are neural networks and the third one is implemented by logistic regression. The models are simple and at the same time the results show that some of them have high accuracy. The models are trained and tested on a widely accessible hardware. Experimental results from training and validation of the models are presented and discussed.