Evaluation of classifiers in a pressure and latency-based typing biometric system

Abstract

System authentication in present time relies on validation by some sort of a password or Personal Identification Number (PIN). However, if an intruder discovers this password or PIN, the user's account can be easily compromised. Biometric systems rely on user authentication based on some physical or behavioral attribute. Typing biometrics is a behavioral biometric authentication system that seeks to identify users based on typing behavior and style, similar to the way that a signature identifies a person based on handwriting. In this paper, Microsoft's newly prototyped Pressure Sensitive Keyboard (PSK) has been used to explore pressure and latency based typing biometrics. Statistical and neural network classifiers are used for user identification on testing samples and compared to evaluate their efficiency.