Multi-class Intrusion Detection System for MANETs

Abstract

As MANETs change their topology dynamically, intrusion detection in these networks is a challenging task.These networks are more liable to the security attacks because of the properties such as node mobility, lack of concentration points where aggregated traffic can be analyzed, intermittent wireless communications and limited band width.We present a multiclass intrusion detection system that addresses these challenges.In this paper we propose a neural network method based on MLP (multi-layer perceptron) for detecting normal and attacked behavior of the system.The method was tested for Black Hole and Gray Hole attacks.We have implemented these attacks using NS2 simulator.The method successfully detected these attacks.We compared the results with KNN (K-Nearest Neighborhood) which is another classifier used for classification.Finally, Re sampling methods were also applied to assess the performance of classifier.This paper presents a graphical representation of the results.