MUMBAI, India, April 17 -- Intellectual Property India has published a patent application (202641043106 A) filed by Pokkunuri Pardha Saradhi; A V Prabu; and Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andhra Pradesh, on April 4, for 'machine learning-based intrusion detection for automotive can bus networks using real vehicle datasets.'
Inventor(s) include Pokkunuri Pardha Saradhi; A V Prabu; Drprabakaran; Mr Sasikumar Chinnadurai; Redboina Nag Aditya; K Ranga Nitheesh Kumar Reddy; Praneeth Gujjeti; and Kuchipudi Yashwanth.
The application for the patent was published on April 17, under issue no. 16/2026.
According to the abstract released by the Intellectual Property India: "With the increased interconnectivity of the present-day vehicles, there are considerable challenges regarding the security of in-vehicle communication networks. The Controller Area Network (CAN) protocol is a standard form of communication between Electronic Control Units (ECUs) in terms of its reliability and efficiency; however, there are no security features associated with the CAN Protocol, including authentication or message encryption. Therefore, CAN networks are subject to many cyber-attack types (e.g., message injection, spoofing, and denial of service). This paper propose a machine learning-based Intrusion Detection Framework for detecting malicious CAN messages and identifying anomalous traffic patterns in automotive networks. The intrusion detection framework is developed based on the publicly available "Car-Hacking" and "CAN-Train-and-Test" port CAN datasets, all of which are real-world CAN data from actual vehicle manufacturers and contain several different types of attack scenarios (e.g., denial of service, fuzzy attacks, RPM spoofing, and gear spoofing). For this project, the use of machine learning for the Intrusion Detection Framework has been based on preprocessing the raw CAN frames by converting the hexadecimal payload data into numerical features, and using Machine Learning techniques to identify attacks. Many different models have been developed for the Intrusion Detection Framework, with the optimized gradient boosting techniques providing improved accuracy at detecting anomalies in CAN traffic. The experimental results indicate that an Intrusion Detection Framework will be able to effectively differentiate between normal and malicious CAN messages with approximately 70% detection accuracy across all types of attacks.Therefore, this work provides evidence that machine learning techniques improve automotive cybersecurity and sets the groundwork for the development of future real time intrusion detection systems for connected vehicles."
Disclaimer: Curated by HT Syndication.
