Intrusion Detection by Ellipsoid Boundary

This paper presents a novel approach to describe the normal behavior of computer networks (as used in IDS) based on Support Vector Data Description (SVDD). In the proposed method we find a minimal hyper-ellipse around the normal objects in the input space. Hyper-ellipse can be expanded in high dimensional space (ESVDD) or to be used as a preprocessing in SVDD method (PESVDD) to obtain better results for IDS. KDD-cup99 has been used as data set for test of the proposed method. The overall experiments show prominence of our work in comparison with similar previous works.