Intrusion Detection System for Smart Vehicles Using Machine Learning Algorithms

As Electric Vehicles (EV s) become more prevalent, the need to protect their complicated and associated systems from cyber threats which vital necessity. It is critical to ensure the safety of an electric vehicle to prevent damage to human life or assets. Controller Area Network (CAN) is a message-broadcast protocol i.e., work on data link layer most deployed on EV s to facilitate communication between different control components called Electronic Control Units or ECUs. An ECU controls the functioning of different electronic or mechanical components and communicates with other ECUs on the network by sending messages over CAN bus. Communication over CAN is unencrypted and unauthenticated and thus prone to malicious attacks. This work presents an Intrusion Detection System (IDS) for electric vehicles. It uses different machine learning algorithms designed specifically for EV s to identify adversary attacks. Thus, through this work, we aim to strengthen the security of an EV by detecting and mitigating potential intrusions or attacks. The proposed system keeps check on network traffic, communication protocols and helps in exchanging data between various parts of the EV such as ECU s, also with charging stations and other attached devices. Key components of the IDS include a data collection mechanism, feature extraction, machine learning algorithms, and a decision making module. The data collection mechanism gathers relevant information from different sensors, control units, and communication channels. We employ feature extraction techniques to transform collected data into meaningful features for analysis. We further compare and contrast the different machine learning techniques to provide useful insights into their functioning.