Reducing false alarm using hybrid Intrusion Detection based on X-Means clustering and Random Forest classification

• Main Authors: Juma, Sundus, Muda, Zaiton, Yassin, Warusia
• Format: Article
• Published: Asian Research Publishing Network (A R P N) 2014
• Online Access: http://psasir.upm.edu.my/id/eprint/35184/; http://www.jatit.org/volumes/sixtyeighth2.php
• Institution: Universiti Putra Malaysia

In recent times, Intrusion Detection systems (IDSs) incarnate the high network security. Anomaly-based intrusion detection techniques, that utilize algorithms of machine learning, have the capability to recognize unpredicted malicious. Unluckily, an essential provocation of this method is to maximize accuracy, detection whereas minimize false alarm rate. This paper proposed a hybrid machine learning approach based on X-Means clustering and Random Forest classification called XM-RF in order to aforementioned drawbacks. X-Means clustering is utilized to gather whole data into congruent cluster based on their behaviour whereas Random Forest classifier is utilized to rearrange the misclassified clustered data to apropos group. The ISCX 2012 Intrusion Detection Evaluation is used as a model dataset. The experimental result pose that the proposed approach obtains better than other techniques, with the accuracy, detection and false alarm rates of 99.96%, 99.99%, and 0.2%, respectively.