Application of kernel spherical k-means for intrusion detection systems

Operational system can be threatened by malicious network activities from intruders or hackers. Consequently, security of a system is indeed become an important subject to tackle this matter. Intrusion Detection System (IDS) is a system which can prevent network traffic and observe suspicious activities in network systems. Therefore, IDS can solve multiple privacy concerns. This paper will propose new method called Kernel Spherical K-Means (KSPKM) that has been modified from Spherical K-Means (SPKM) algorithm by using RBF and polynomial kernel. For our empirical study, we will be using the dataset from KDD Cup 1999 then classified types of attacks into five classes. In the end, we will see which one will produce better results in terms of classification accuracy. We found out that KSPKM succeed to improve clustering accuracy with the highest rate being 98,31% compared to SPKM.