Improving User Privacy and the Accuracy of User Identification in Behavioral Biometrics

Humans exhibit their personality and their behavior through their daily actions. Moreover, these actions also show how behaviors differ between different scenarios or contexts. However, Human behavior is a complex issue as it results from the interaction of various internal and external factors such as personality, culture, education, social roles and social context, life experiences, among many others. This implies that a specific user may show different behaviors for a similar circumstance if one or more of these factors change. In past work we have addressed the development of behavior-based user identification based on keystroke and mouse dynamics. However, user states such as stress or fatigue significantly change interaction patterns, risking the accuracy of the identification. In this paper we address the effects of these variables on keystroke and mouse dynamics. We also show how, despite these effects, user identification can be successfully carried out, especially if task-specific information is considered.