Detecting Network Anomalies in Mixed-Attribute Data Sets

Abstract

Detecting network anomalies is important part of intrusion detection systems that have been developed with great successes on homogeneous data. There have been successes with mixed-attribute data using various techniques, however, few of them exist for using mixed-attribute data without further manipulation or consideration of dependencies among the different types of attributes. We propose in this paper a fusion of decision tree and Gaussian mixture model (GMM) to detect anomalies in mixed-attribute data sets. Evaluation experiments were performed on the popular KDDCup 1999 data set using C4.5 decision tree, GMM and the fusion of C4.5 and GMM.