90 - 107
Ensemble Miscellaneous Classifiers Based Misbehavior Detection Model for Vehicular Ad-Hoc Network Security
Authors: S. Sumithra, R. Vadivel
Number of views: 113
Vehicular Ad-Hoc Network is an emerging technology, mainly developed for road safety applications, entertainment applications, and effective traffic conditions. VANET applications work based on the accurate mobility information shared among the vehicles. Sometimes attackers manipulate the mobility information shared by the adjacent vehicle or neighboring vehicle, which results in terrible consequences. To deal with the illusion-based type of attacks, researchers have proposed enormous solutions. Unfortunately, those solutions could not deal with the dynamic vehicle conditions and variable cyber malfunctions, which reduces the misbehavior detection accurateness and increases the false-positive rate. In this paper, the dynamic vehicle context is taken into account to propose a two solutions such as Miscellaneous VANET Classifiers based Misbehavior Detection Model (MVC-MDM) and Ensemble Miscellaneous VANET Classifiers based Misbehavior Detection Model (EMVC-MDM). This model is constructed based on the Mobility Data Gathering phase, Mobility Context Feature Extraction phase, Mobility Context Feature Level Fixing phase, Hampel Filter based Context Reference Building phase, Constructing Miscellaneous VANET Classifiers based Misbehavior Detection model and Ensemble Miscellaneous VANET Classifiers based Misbehavior Detection phase. Vehicle context is prepared using the data-centric features and the behavior-based features of the vehicles. The Nonparametric Hampel filter and Kalman filter are used to building the context reference model. These filters discover the temporal and spatial correlation of the uniformity in the current mobility information. Vehicle features are extracted locally according to the stability, likelihood, and performance of the vehicles' mobility information. A random forest based learning algorithm is used to train and test the classifiers. The proposed MVC-MDM and EMVC-MDM has been simulated in various context scenarios and the presence of misbehaving vehicles. NGSIM dataset has been used for extensive simulation. The results prove that the effectiveness and the reliability of the proposed MVC-MDM and EMVC-MDM are higher than the existing misbehavior detection systems.