AirAuth: Towards Attack-Resilient Biometric Authentication Using In-Air Gestures

Abstract

AirAuth is a biometric, gesture-based authentication system based on
in-air gesture input. We describe the operations necessary to sample
enrollment gestures and to perform matching for authentication, using
data from a short range depth sensor. We present the results of two
initial user studies. A first study was conducted to crowd source a
simple gesture set for use in further evaluations. The results of our
second study indicate that AirAuth achieves a very high Equal Error Rate
(EER-)based accuracy of 96.6 % for simple gesture set and 100 % for
user-specific gestures. Future work will encompass the evaluation of
possible attack scenarios and obtaining qualitative user feedback on
usability advantages of gesture-based authentication.