Improving performance of intrusion detection system by applying a new machine learning strategy

Abstract

The most acute problem for misuse detection method is its inability to detect new kinds of attacks. A better detection method, which uses a new learning strategy, is proposed to solve this problem. A Concept Hierarchy Generation for attack Labels (CHGL) applying relevant feature subset codes clustering, makes common machine learning algorithms learn attack profiles on high concept levels. And that will enable the system detect more attack instances. Experimental results show the advantage of this new method.