International Journal of Innovative Research and Development

Soumen Roy ¹, Utpal Roy ², D. D. Sinha ¹

Affiliations
1 Department of Computer Science and Engineering, University of Calcutta, West Bengal, IN
2 Department of Computer & System Sciences, Visva-Bharati, Santiniketan, West Bengal, IN

Abstract

User authentication or identification is the big challenges in E-Business. In this paper, we have implemented a typing biometric technique which increases the security level up to 98.1% without changing existing authentication technique. Habitual typing speed pattern is a behavioural biometric characteristic in Biometric Science can be effectively implemented to classify the users. This typing speed pattern is promising as biometric characteristics which cannot be lost or stolen in addition with inexpensive to collect. Many statistical, distance-based and machine learning algorithms are proposed on habitual typing pattern and many have obtained impressive results, but in practice, the accuracy level is not much promising, it demands higher level of security and reliability. In our experiment, we have collected press and release time of 12096 keystrokes using Java Applet programming form 12 individuals during 12 months in 4 sessions for 1440 samples then we analysed that data using R statistical programming language and obtained average Equal Error Rate (EER) of 21 different data-mining and distance-based classification algorithms and compared their performance in accuracy to search the suitable algorithms on typing patterns. But in evaluation process, a classifier’s average Equal Error Rate (EER) widely jumped from 1.9% to 63%. The question may arise, which classifier is suitable on typing speed patterns, which pattern of string is suitable. To get the answer, we have started our experiment and created our own rhythmic keystroke dynamics database of different pattern of strings and executed various classification algorithms on it, so, we can compare their performance soundly. 

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.

Full Text