Enhancing Security in Autonomous Vehicles: A Machine Learning-Based Intrusion Detection Approach

Autonomous vehicles (AV) have transformed the transportation industry by driving themselves and making decisions without human input. Enhancing the safety, reliability, as well as cybersecurity of AV is of utmost importance, and this research showcases notable accomplishments in intrusion detection systems (IDS) for such vehicles. Using several machine learning (ML) as well as deep learning (DL) algorithms, the study’s goal is to establish improved detection rates with guaranteed reliability. As such, the study’s objective is to develop a reliable IDS for autonomous vehicles, using XGBoost for feature selection and a combination of ML/DL algorithms which ensure high detection accuracy, low false alarms, and real-time responsiveness. These techniques successfully manage unbalanced data, improving the detection accuracy of intruded as well as normal cases. This study verified the IDS performance using two real-world datasets, including CICIDS 2017 and the CAN BUS intrusion dataset. This work evaluates the recall, specificity, and sensitivity of this study using conventional accuracy and precision measures, as well as F1score. The outcomes ensure the safety as well as reliability of AV, potentially leading to the establishment of a robust IDS. This achievement aims to protect road users and pedestrians from the effects of potential cyber threats as well as passengers.