An analysis of supervised tree based classifiers for intrusion detection system

Abstract

Due to increase in intrusion incidents over internet, many network intrusion detection systems are developed to prevent network attacks. Data mining, pattern recognition and classification methods are used to classify network events as a normal or anomalous one. This paper is aimed at evaluating different tree based classification algorithms that classify network events in intrusion detection systems. Experiments are conducted on NSL-KDD 99 dataset. Dimensionality of the attribute of the dataset is reduced. The results show that RandomTree model holds the highest degree of accuracy and reduced false alarm rate. RandomTree model is evaluated with other leading intrusion detection models to determine its better predictive accuracy.