Masquerade Detection SystemBasedon Correlation EigenMatrixand Support Vector Machine

This article presents amasquerade detection system based onCorrelation Eigen Matrix andsupport vector machine (SVM). Thesystem first creates aprofile defining a normaluser's behavior byCorrelation EigenMatrix, andthencompares thesimilarity ofa current behavior withthecreated profile todecide whethertheinputinstance isvaliduseror masquerader. Inorder toavoid overfitting andreduce thecomputational burden, userbehavior principal features areextracted bythePCAmethod SVMisused todistinguish valid userorMasquerader foruser behavior after training procedure hasbeencompleted bylearning. Intheexperiments forperformance evaluation thesystem achieved acorrect detection rate equal to82.6%andafalse detection rateequal to 3.000, whichisconsistent withthebest results reports intheliterature forthesamedatasetandtesting paradigm.