Iterative Keystroke Continuous Authentication: A Time Series Based Approach

Abstract

Keyboard typing patterns are a form of behavioural biometric that can be usefully employed for the purpose of user authentication. The technique has been extensively investigated with respect to the typing of fixed texts such as passwords and pin numbers, so-called static authentication. The typical approach is to compare a current “typing sample” with a typing template expressed in terms of a feature vector comprised of keystroke dynamics. The feature vector approach has also been applied in the context of continuous authentication where features are extracted from free typing. However, the use of feature vectors for keystroke continuous authentication entails a number of disadvantages, mostly associated with the size of the feature vectors and their generation, which need to capture a large number of features to be effective; thus making the technique unsuitable for iterative (real-time) authentication as would be required in the case of, for example, online assessments. To address this issue, a mechanism whereby iterative real-time keystroke continuous authentication can be achieved is proposed, by considering typing behaviour as a form of time series, that avoids the disadvantages associated with the feature vector approach. The reported experimental results show a significantly improved performance using the proposed method in comparison with the feature vector based technique.