Ensemble Based Machine Learning Classifier for Sybil Attack Detection in Vehicular AD-HOC Networks (VANETS)

In VANETs (Vehicular Ad-hoc Networks), the duality of promise and challenge is present, where main risks in security, such as Sybil Attack, seem formidable. In order to achieve this, adversaries manipulate such identities to disrupt the nodes thus creating a major risk to transportation systems and result in traffic jams. Rather, a new collaborative framework that incorporates autoencoders for feature mapping and selection is presented to address this disorder. In this paper, an Ensemble based Machine Learning Classifiers are combined with Hard and Soft Majority Voting mechanism to detect Sybil attacks. This is a framework that is based voting that is majority and it integrates these multiple classifiers which include k Nearest Neighbour, Decision Tree, and Support Vector Machines (SVM) that are all executing in parallel. By using autoencoders framework, it can extract and base particularly the VANET data, hence accuracy and reliability of detection improved. The use of the Majority Voting (Hard and Soft) as defense mechanism is then employed to provide a diverse defense mechanism against the Sybil attacks. This is done by several tests and analysis leading to an $\mathbf{9 8. 5 6 \%}$ Sybil attacks depending and establishing it as a reliable security option for VANETs.