Development of an artificial intelligence supported intrusion detection system for detecting cyber attacks on autonomous vehicles

The development of technology increases the comfort and welfare of societies in every area of life. This development is encountered in various areas, especially transportation. In this direction, the widespread use of autonomous vehicles has accelerated with the integration of smart systems into automobiles. The fact that autonomous vehicles have their own systems and networks has also paved the way for cyber attacks that can be carried out against autonomous vehicles. Such attacks generally have three purposes. These are to infiltrate the system and take control of system components, to cause the system to work slowly by putting a load on the system network and to cause the system to crash. The serious consequences of cyber attacks against autonomous vehicles have necessitated the adoption of security measures that will protect autonomous vehicles against cyber attacks. In this thesis, the detection of cyber attacks on the in-vehicle networks of smart vehicles was carried out using machine learning models. Cyber attacks were carried out by creating a simulation environment and a data set was created. Then, the data set was classified with XGBoost, LightGBM, Random Forest and Decision Tree algorithms and performance comparisons were made. As a result of the experiments, XGBoost gave the best accuracy rate with %86.22, while the Decision Trees algorithm gave the lowest accuracy rate with %80.7. It is thought that the thesis study will contribute to the efforts of smart vehicle security experts to prevent cyber threats and increase awareness about smart vehicle cyber security.