Smartphone context-aware authentication using gait, touch and keystroke data

Biometric modalities based on human behavior are now seen as a solution to many of the tradition problems of authentication methods such as the PIN lock mechanism. In this work, we consider context-aware authentication based on three behavioral biometric modalities: gait, touch and keystroke. With respect to the gait modality, this work shows that through a combination of supervised learning methods and statistical tests, it is possible to use information on how a phone is held (in a hand or pocket) to improve classification performance drastically. We propose a two-stage authentication framework that identifies the location of the phone before performing authentication, and show its benefits based on a dataset of 30 users. With respect to the touch modality, this work shows experimentally the impact of contexts beyond phone orientation on touch-based authentication. This work presents experimental results on how touch-based authentication is impacted by users who are listening to music while swiping, users who are not listening to music while swiping, users who are swiping over images, and users who are not swiping over images. Experiments are carried out with a data-set which we collected for this purpose from 34 subjects. Moreover, this work also provides design considerations towards a touch-based context-aware system and shows how the presence of music (which was found to have significance) can be incorporated. With respect to the keystroke modality, this work examines the impact of the contexts: listening to music while typing, not listening to music while typing, typing while looking at images, and typing while not looking at images. Experiments are carried out with a data-set which we collected for this purpose from 27 subjects. This work also provides design considerations towards a keystroke-based context-aware system and shows how the presence of music (which was found to have significance) can be incorporated