PROJECT TITLE :

Misbehavior Detection for Position Falsification Attacks in VANETs Using Machine Learning

ABSTRACT:

Vehicles are able to communicate with one another and with infrastructures through the use of an advanced technology known as Cooperative Intelligent Transport Systems (C-ITS). This technology was developed to improve road safety and the efficiency of traffic flow through the use of Vehicular Ad Hoc Networks (VANETs). The possibility of attacks on this kind of network, which could put the passengers' safety in jeopardy, is one of the primary motivating factors behind one of the primary concerns in C-ITS, which is the security of VANETs. Intrusion Detection Systems, also known as IDS, play a significant part in the protection of the vehicular network by identifying vehicles that are acting inappropriately. In a centralized intrusion detection system, the majority of the published works use the same familiar characteristics. In this paper, we propose a Machine Learning (ML) mechanism that takes advantage of three new features that are primarily related to the sender position. This allows us to enhance the performances of IDS for position falsification attacks, which is the main focus of the paper. In addition, it provides a comparison of two distinct machine learning methods for classification, namely k-Nearest Neighbor (kNN) and Random Forest (RF), both of which are used to identify malicious vehicles based on these features. kNN and RF are referred to as kNN and RF respectively. In addition, a process known as Ensemble Learning (EL), which improves detection performance by combining multiple machine learning techniques (kNN and RF in this instance), is carried out. During construction of an IDS, which enables vehicles to detect inappropriate behavior in a decentralized manner, the detection mechanism itself is trained in a centralized location. The findings indicate that the proposed mechanism provides results that are superior, on average, to the best approaches that have been used in the past in terms of classification performance indicators and the amount of computational time required.