Anomaly based intrusion detection system for networks using k-means clustering, discretization, and naïve Bayes classification: A hybrid approach [articol]

Abstract

The role of Intrusion Detection System (IDS) is to discover, assess and account unauthorized access, illegal activities and security issues in Network. The aim of this research is to model an Anomaly-Based Intrusion Detection System as a Semi-supervised machine learning method involving K - Means Clustering and Naïve Bayesian Classification. This proposed method focuses on reducing data loss during clustering and improving the exactness of classification. This is achieved by adding an intermediate pre-processing technique named Proportional K - Interval Discretization (PKID). The proposed combination of K - Means Clustering, PKID, and Naïve Bayesian Classification is evaluated against KDD Cup 99 Dataset. The results show that the proposed scheme gives an average Accuracy rate of 99.32%, Detection rate of 99.45% and False alarm rate of 0.128 during training and testing phase.