International Journal of Computer Science Issues

More than a traditional journal...

All most all existing intrusion detection systems focus on low level attacks, and only generate isolated alerts. They cant find logical relations among alerts. In addition, IDS accuracy is low;a lot of alerts are false alerts. To reduce this problem we propose a hybrid approach which is the combination of K-Medoids clustering and Nave-Bayes classification. The proposed approach applies clustering on all data into the corresponding group and after that applies a classifier for classification purpose.The proposed work will explore Nave-Bayes Classification and K-Medoid methods for intrusion detection and how it is useful for IDS. Nave Bayes Classification can be mined to find the abstract correlation among different security features. In this, we are presenting implementation results on existing intrusion detection system and K-Medoid clustering technique with Nave Bayes classification for intrusion detection system. An experiment is carried out to evaluate the performance of the proposed approach using our own created dataset. Result shows that the proposed approach performs better in term of accuracy,execution time, CPU utilization and memory consumption with reasonable false alarm rate.