A New High Performances Intrusion Detection System based on Dempster-Shafer Theory

Abstract

Intrusion detection systems (IDS) have become a very important element in securing any network infrastructure. Malicious attacks have resulted in negative impacts on network as before, increasing the need for an effective approach to detect and identify such attacks more effectively. In context of improving performance, this paper presents a new adaptive intrusion detection system based on Dempster Shafer theory. It is a hybrid and multi-levels model. Each level includes two classifiers Naïve Bayes and Support Vector Machine known for their performance in classification. The decision at each level is performed using fuzzy logic and the combination rule of Dempster. The experiments were carried out with KDD’99 data sets. The experimental results show that our approach greatly improves the detection rate with low false alarm rates compared to some recent existing works.