A Hybrid Data-driven Model for Intrusion Detection in VANET
Hind Bangui, Mouzhi Ge, Barbora Bühnová
Abstract
Nowadays, VANET (Vehicular Ad-hoc NETwork) has gained increasing attention from many researchers with its various applications, such as enhancing traffic safety by collecting and disseminating traffic event information. This increased interest in VANET has necessitated greater scrutiny of machine learning (ML) methods used for improving the security capabilities of intrusion detection systems (IDSs), such as the need to solve computationally intensive ML problems due to the increased vehicular data. Therefore, in this paper, we propose a hybrid ML model to enhance the performance of IDSs by dealing with the explosive growth in computing power and the need for detecting malicious incidents timely. The proposed approach mainly uses the advantages of Random Forest to detect known network intrusions. Besides, there is a post-detection phase to detect possible novel intruders by using the advantages of coresets and clustering algorithms. Our approach is evaluated over a very recent IDS dataset named CICIDS2017. The preliminary results show that the proposed hybrid model can increase the utility of IDSs.