Paper Contents
Abstract
The increasing integration of Electronic Control Units within modern vehicles has amplified concerns about cybersecurity threats in CAN protocols, which inherently lack encryption and authentication. As a response, Intrusion Detection Systems have emerged as a vital defense mechanism to secure in-vehicle networks. This research proposes a hybrid IDS model that leverages advanced feature engineering techniques and ensemble learning methods to enhance intrusion detection performance. A detailed review and preprocessing of a real-world CAN dataset comprising over 817,000 records enabled the extraction of relevant temporal and payload-based features. The proposed system employs Gradient Boosting and XGBoost classifiers to detect diverse attack types such as Denial of Service ,Gear Spoofing , RPM Spoofing , and Fuzzy attacks. Experimental results demonstrate high detection capability , with XGBoost achieving up to 99.27% accuracy , 99.60% F1-score, and 99.97% recall on the test set. These findings validate the efficacy of the proposed hybrid approach in securing in-vehicle networks with minimal false negatives. Future improvements may include deeper neural models and real
Copyright
Copyright © 2025 Poonam . This is an open access article distributed under the Creative Commons Attribution License.
