Abstract: With the rapid development of Internet in recent years, computer systems arc facing increased number of security threats. Various soft computing based approaches have been proposed to detect computer network attacks. A method for attack feature extraction based on extension matrix theory was given in this paper. 13y constructing extension matrix on positive and negative examples, the integer programming model for its optimal feature subset selection was built, which will be solved by simple genetic algorithm. Finally optimal rules for detection of specific attack were generated. Experimental results show the achievement of high correct detection rates and acceptable low false positive rates based on benchmark KDI)Cup99 data sets.

Key words: Extension matrix, Feature subset selection, Uenetic algorithm, Intrusion detection