Misbehavior Detection Based on Ensemble Learning in VANET

Abstract

Detection of misbehaviors in Vehicular Ad Hoc Networks (VANETs) using machine learning methods has not been investigated extensively. In VANET, an illegitimate vehicle may transmit inaccurate messages to trigger an un- avoidable situation. In this paper, we present an ensemble based machine learning approach to classify misbehaviors in VANET. The performance of classifiers used for classification depends on the induction algorithms. We exploit the strengths of different classifiers using an ensemble method that combines the results of individual classifiers into one final result in order to achieve higher detection accuracy. Proposed security framework to classify different types of misbehaviors is implemented using WEKA. Features of nodes participating in VANET are extracted by performing experiments in NCTUns-5.0 simulator with different simulation scenarios (varying the number of legitimate and misbehaving nodes). We evaluate ensemble method using five different base inducers (Naive Bayes, IBK, RF, J48, Adaboost(J48)). We also show that ensemble based approach is more efficient in classifying multiple misbehaviors present in VANET as compared to base classifiers used for classification.