Network programming and mining classifier for intrusion detection using probability classification

Abstract

In conventional network security simply relies on mathematical algorithms and low counter measures to taken to prevent intrusion detection system, although most of this approaches in terms of theoretically challenged to implement. Therefore, a variety of algorithms have been committed to this challenge. Instead of generating large number of rules the evolution optimization techniques like Genetic Network Programming (GNP) can be used. The GNP is based on directed graph, In this paper the security issues related to deploy a data mining-based IDS in a real time environment is focused upon. We generalize the problem of GNP with association rule mining and propose a fuzzy weighted association rule mining with GNP framework suitable for both continuous and discrete attributes. Our proposal follows an Apriori algorithm based fuzzy WAR and GNP and avoids pre and post processing thus eliminating the extra steps during rules generation. This method can sufficient to evaluate misuse and anomaly detection. Experiments on KDD99Cup and DARPA98 data show the high detection rate and accuracy compared with other conventional method.