Network Intrusion Detection Using a HNB Binary Classifier

Abstract

Utilization of the data mining techniques in intrusion detection systems is common for the classification of the network events as either normal events or attack events. Naïve Bayes (NB) method is a simple, efficient and popular data mining method that is built on conditional independence of attributes assumption. Hidden Naïve Bayes (HNB) is an extended form of NB that keeps the NB's simplicity and efficiency while relaxing its independence assumption. Our experimental research claims that the HNB binary classifier model can be applied to intrusion detection problem. Experiment results using classic KDD 1999 Cup intrusion detection dataset indicate that HNB binary classifier has better performance in terms of detection accuracy compared to the traditional NB classifier.