A Novel Approach for Keyboard Dynamics Authentication based on Fusion of Stochastic Classifiers
Abbas Rezaei, S. Mirzakochaki · 2012
Summary Computer security is one of most important issues around the world. Most computer systems are using passwords for their own authentication or verification mechanisms. A robust and efficacious approach for classification of 24 persons who their typing patterns were collected introduced. A linear discriminate classifier (LDC), quadratic discriminant classifier (QDC) and knearest neighbor (K-NN) are utilized to classify users keystroke patterns. After that a set of mentioned ensemble methods are adopted to reduce the error rate and increase the reliability of biometric authentication system. Promising results have been achieved. The best mean FAR, FRR and EER parameters are achieved for singular classifiers as 19.20%, 0.81% and 1.39% respectively. The state of the art performance results mean FAR, FRR and EER parameters are achieved for the ensemble classifiers as 0.00%, 0.00% and 1.15% respectively.