PROJECT TITLE :

Machine Learning Based Misbehaviour Detection in VANET Using Consecutive BSM Approach

ABSTRACT:

In the context of an Intelligent Transportation System, the term "vehicular ad-hoc network" (abbreviated as "VANET") refers to a developing method of vehicle-to-vehicle communication known as "vehicular ad-hoc network" (VANET). This method is essential for lowering the number (ITS). The integrity of messages sent over VANET and the authentication of vehicles are both typically handled with cryptographic methods due to the fact that VANET communication is susceptible to a variety of attacks. However, the use of cryptographic techniques by themselves might not be enough to provide adequate protection against attacks from within. Numerous VANET safety applications rely on the periodic broadcast of basic safety messages (BSMs) from vehicles in the surrounding area. These messages contain essential status information about a vehicle, such as its position, speed, and heading, and they are sent out by vehicles in the surrounding area. The injection of erroneous position information into a BSM by an adversary, in the form of a misbehaving vehicle, can have serious repercussions, including the creation of traffic congestion or even collisions. For this reason, it is absolutely necessary to accurately detect and identify the attackers in order to guarantee the network's security. Detecting position falsification attacks with the help of machine learning (ML) algorithms is the topic of this research paper, which presents a novel data-centric approach to the problem. The information for training and testing in the proposed method is combined from two separate BSMs, in contrast to the techniques that are currently in use. Simulations conducted with the Vehicular Reference Misbehavior (VeReMi) dataset show that the proposed model performs noticeably better than any of the existing approaches when it comes to identifying a wide variety of different kinds of assaults.