Analyzing user typing behaviour in different positions using keystroke dynamics for mobile phones

Abstract

Hold time and latency are the most frequently used features for developing a user typing profile for authenticating mobile phone users. In this paper, we combine these features along with motion features, i.e., the X, Y and Z coordinates at which the user holds the mobile while typing and take the input from the user when he/she is sitting, walking and relaxing. We then address three questions (1) whether there is any change in the typing pattern of the user when his typing position (relaxing, sitting or walking) changes?; (2) in which position the user has the unique typing pattern?; (3) which is the best feature vector to consider for authentication? The results showed that the user typing pattern differs in different positions and also with the change in phone orientation. A user has the unique pattern when he/she is in relaxing mode and holding the mobile in landscape orientation and the best feature vector to build a user profile is by combining hold time, latency and motion features.

Subject Classification:

• 86A17

Keywords:

• Keystroke Dynamics
• Mobile Phones
• Authentication
• Typing Pattern
• Random Forest