A comparative performance study of neural network paradigms for identifying computer users

Abstract

This paper presents a neural network system for identifying computer users. A comparative evaluation study of three neural networks paradigms as applied to the identification of computer users using keystroke intervals when typing a well known phrase is made. The input vectors were made up of the time intervals between successive keystrokes created by users while typing characters. Each input vector was classified into one of several classes, thereby identifying the user who typed the phrase. We investigated and compared the performance of the neural network paradigms as applied to this problem. These paradigms are: Adaptive Resonance Theory-2, (ART-2), Back Propagation, and Counterpropagation. The identification technique presented here is accurate, practical and novel.