User authentication with keystroke dynamics in long-text data

Keystroke dynamics is a form of behavioral biometrics that can be used for continuous authentication of users while working at a terminal. In this paper, we extend the use of support vector machine (SVM) for continuous authentication with long-text data, from one-time password based authentication using short text. In result, we show we can authenticate legitimate users and reject impostors with negligible error (close to 0% equal error rate) by setting a one-class SVM for each user using a dataset of 34 users in a controlled environment. Our results show that by standardizing the input and setting the correct kernel scale, one-class SVM can be utilized as a tool to continuously authenticate users, and recognize keystroke dynamics with a high accuracy.