Skyline computation for improving naïve Bayesian classifier in intrusion detection system

Abstract

Intrusion detection systems (IDSs) are critical to network security. However, there are some common defects with the existing IDSs, namely, low detection rate of rare attacks and high number of false alarms. Many have suggested solving these defects by integrating different IDSs techniques, but the effectiveness has not been justified. This paper puts forward a two-layer hybrid IDS based on Skyline operator and Naïve Bayesian classifier. First, the most suitable classifier was identified through Skyline computation based on three criteria, namely, accuracy, detection rate and false alarm rate. Then, the results were integrated by the Naïve Bayesian classifier into the final decision. To verify its effectiveness, the proposed IDS was tested on the famous KDD dataset. The results show that our system greatly improves the detection rate of rare attack, while decreasing false alarms rate, from the levels of the previous techniques.