A New Semi-unsupervised Intrusion Detection Method Based on Improved DBSCAN

Abstract

In order to improve the efficiency of the existing intrusion detection systems, this paper proposed a new semi-unsupervised intrusion detection model based on improved DBSCAN algorithm, called IIDBG, and it was applied to detection engine. In IIDBG, distance calculation formula and clusters merger process were improved based on the DBSCAN and existing the improved algorithm IDBC. The experiments demonstrate that our method outperforms the existing clustering methods in terms of accuracy and detecting unknown intrusions.