Soft biometrics for keystroke dynamics: Profiling individuals while typing passwords

This paper presents a new profiling approach of individuals based on soft biometrics for keystroke dynamics. Soft biometric traits are unique representation of a person, which can be in a form of physical, behavioural or biological human characteristics that differentiate between him/her into a group people (e.g. gender, age, height, colour, race etc.). Keystroke dynamics is a behavioural biometric modality to recognise how a person types on a keyboard. In this paper, we consider the following soft traits: the hand category (i.e. if the user types with one or two hands), the gender category, the age category and the handedness category. For this purpose, we collected a new database. Two cases are studied: static passwords and free text. By combining machine learning and fusion process, the results are promising.