DataMining aKeystroke Dynamics Based Biometrics Database UsingRoughSets

Softwarebasedbiometrics, utilising keystroke dynamicshasbeenproposedas a cost effective meansofenhancing computer access security. Keystroke dynamics hasbeensuccessfully employed as a meansofidentifying legitimate/illegitimate login attempts basedonthetyping style ofthelogin entry. In this paper, wecollected keystroke dynamics datainthe formofdigraphs froma series ofusersentering a specific login ID.We wished todetermine ifthere were anyparticular patterns inthetyping styles thatwould indicate whether alogin attempt waslegitimate ornot using roughsets. Ouranalysis produced asensitivity of 96%,specificity of93% andan overall accuracy of 95%. Theresults ofthisstudyindicate thattyping speedandthefirst fewandthelast fewcharacters ofthe login IDwerethemostimportant indicators ofwhether thelogin attempt waslegitimate ornot.