INTRUSIONDETECTION SYSTEM BASED ON FEATURE SELECTIONAND SUPPORT
Chunhua Gu · 2006
Support VectorMachine(SVM)hasbeenapplied to Intrusion Detection System(IDS)foritsabilities to perform classification andregression. Butforlarge-scale network intrusion detection problem, sincesolving a support vector machine isatypical quadratic optimization problem, whichisinfluenced bythedimension and quantity ofexamples, manyproblems arise. KDDCUP'99 wasusedastheexperiment dataset inthis paper. A feature selection technology based onFisher score waspresented andusedtoconstruct a reduced feature subset of KDDCUP'99dataset. SVM wasusedasa classifier. Experiment wasrun.Theexperiment results show, using Fisher score combined withSVM toselect theimportant features isaneffective method toreduce thedimension of theexample feature space, andtheclassification accuracy hasnotdramatically decreased comparing totheoriginal feature space.