Description

The menace of illegal access to data resources is a growing concern of researchers in the field of computer science. A significant amount of effort is required to monitor the activities in a computer network with a view to detect any attempt for intrusion. From this perspective, the main motivation behind this research is to design an efficient intrusion detection system using some novel data mining approaches that have the capability to detect intrusions with high detection rate with low false positive rate. In this work, we take multiple supports Apriori algorithm with various interestiness measures to obtain the most significant rules in detecting network intrusions. Further, we propose some novel ensemble of classifiers in order to enhance the detection rate of network attacks. Some unsupervised clustering algorithms have been proposed to further increase the detection rate of new or unseen attacks that fall under rare attacks categories. Finally, certain hybrid data mining approaches have been employed in order to design an efficient anomaly based network intrusion detection system that can achieve high detection rate and low false positive rate.