ManetSVM: Dynamic anomaly detection using one-class support vector machine in MANETs

Abstract

The main goal of one-class classification is to classify one class from remaining feature space. One-class SVM is a kernel based approach which is very fast and precise and therefore is used in different fields such as image processing, protein classification and anomaly detection for statistical learning. There are some approaches suggested for anomaly detection in MANETs that most of them are static and use a predefined model. Due to the dynamic characteristics of MANETs, they cannot be applied to these networks well. In this paper we have proposed a one-class SVM for dynamic anomaly detection in mobile ad-hoc networks with AODV routing protocol, called ManetSVM. The efficiency of ManetSVM for detection of flooding, blackhole, neighbour, rushing, and wormhole attacks has been evaluated. Simulation results show that ManetSVM is able to achieve a better balance between Detection Rate and False alarm Rate in comparison with other dynamic anomaly detection approaches.