Soumen Roy et. al. / International Journal of New Technologies in Science and Engineering Vol. 2, Issue. 4, October 2015, ISSN 2349-0780 Available online @ www.ijntse.com 64 Performance Perspective of Different Classifiers on Different Keystroke Datasets Soumen Roy #1 , Utpal Roy *2 , D. D. Sinha #3 #1,3 Department of Computer Science and Engineering University of Calcutta, 92 APC Road, Calcutta -700 009, INDIA. *2 Department of Computer & System Sciences, Visva-Bharati, santiniketan -731235, INDIA 1 soumen.roy_2007@yahoo.co.in 2 roy.utpal@gmail.com 3 devadatta.sinha@gmail.com Abstract- According to SplashData (who gathered data from millions of stolen passwords posted online), the top three passwords in the year 2013 are “123456,” “password” and “12345678”. So we can say most of the people are uninspired while choosing a healthy password because we, as people are still very lazy. It increases the probability of guessing attacks. To minimize these attacks, we pick up some words for password from relatively small dictionary and decorate it by adding extra texts or combine capital, small letter with some symbols. It increases the complexity of password which is very difficult to remember and we forget to distinguish this type of healthy passwords for different access control systems. To solve this problem, here, in this paper, we investigated fixed-text user authentication through keystroke dynamics. Here our typing style is also accounted with the pressed password and user ID. It is established that our typing style is a behavioral biometric characteristic relates the issue of human identification or authentication. But the accuracy level of this technique is not much accepted in practice. In order to realize this technique demands higher level of security with accepted level of error 0.000000…1. Hence, it is highly needed to identify the keystroke dynamics features or combination of features and analyses the accuracy with different classification techniques. Keywords: Keystroke Dynamics, EER, Behavioral Biometric, Canberra , Chebyshev, Czekanowski, Gower, Intersection, Kulczynski, Lorentzian, Minkowski, Motyka, Ruzicka, Soergel, Sorensen, Wavehedges, Manhattan Distance, Euclidean Distance, Mahanobolis Distance, Z Score, KMean, SVM, NaiveBaysian, ROC Curve. I. INTRODUCTION Passwords or PINs are used to recognition the user in knowledge-based user authentication technique. But it is unsafe while all around the areas are covered by video cameras or spy cameras. Basically in Bank is a public place, where pressing password slowly can be traceable by bank customers. If we pick up some words from our small dictionary for password then an attacker may collect our personal information and can check one by one until the actual result is obtained. So there is a probability of Brute force attack or shoulder surfing attack or dictionary attacks. Token-based user authentication scheme uses some physical items called tokens such as smart cards, user ID card, driving licence and some PIN. Here, something we have that is token and something we know that is PIN which can create