Eathentication

Abstract

Nowadays, authentication systems are a milestone for the security of modern societies. In particular, researchers proposed several effective authentication mechanisms for mobile devices. Unfortunately, most of these still require the user to interact lately with the smartphone screen, which is often undesirable in many setting where the user can not take the phone (e.g., at an airport's gate, in the crowded subway, while driving). In several of these scenarios, users are anyway wearing earphones. In this paper, we propose Eathentication: a novel user-friendly authentication method based on an assessment of the ear channel movement during chewing. Eathentication exploits proximity led sensors mounted on earphones to measures the movement of ear channel. We conducted our experiments collecting data from 23 participants, during three chewing sessions. During the experiment, the subject performed the test, moving the jaw and chewing different types of food. We trained different Machine Learning models on single participant performing intra-subject and inter-subject prediction. Results show that Eathentication can effectively authenticate people based on their chewing behaviour on the investigated stimuli. For the better classifier, our method achieved a False Acceptance Rate of 0.041± 0.016 and a False Rejection Rate of 0.128± 0.043.