Analysis of Defuzzification Methods for Network Intrusion Detection

Abstract

Fuzzy logic is appropriated for the intrusion detection problem because many quantitative features are involved in intrusion detection. Fuzzy logic system can handle simultaneously the numerical data and linguistic knowledge. The concept of linguistic variables is used to model the state of the system which is imprecise and uncertain. The purpose of this paper is to analyze the behavior of the intrusion detection on the KDD dataset using the five defuzzification methods. The result shows that the centroid and bisector methods can detect intrusion better than the other methods for intrusion detection. The experiments and evaluations of this paper were performed with the KDD Cup 99 intrusion detection dataset. Simulation results are demonstrated by using MATLAB.