(An) intrusion detection system using multiple binary-class stacking ensemble method

Abstract

Today, internet is closely related to our daily life. Internet creates benefits using huge amount of information. But using Internet contains some risks of network attacks, thus, detecting network intrusion becomes a significant issue all over the world. Intrusion Detection System is widely employed for security purpose to detect network attacks. In this thesis, we suggest an intrusion detection system using multiple binary-class stacking ensemble method. Previous studies showed that ensemble of weak classifiers can enhance performance of intrusion detection system. In multi-class problem, certain classifier can accurately classify only specific classes, not entire classes. Thus, we employed classifier which is specialized in only one specific class. Also, we use stacking, one of meta-classification methods, ensemble method to combine base classifiers output predictions. Simulation results show that our proposed framework improves detection rate without increasing false positive rate.