Smart kiosk with gait-based continuous authentication

Abstract

The authors propose to develop a smart kiosk that plays the role of an identity selector activated implicitly when a user is approaching that kiosk. The identity of a user is recognized implicitly in background by a mobile/wearable device based on his or her gait features. Upon arriving at a smart kiosk, the authentication process is performed automatically with the current available user identity in his or her portable device. To realize our system, we propose a new secure authentication scheme compatible with gait-based continuous authentication that can resist against known attacks, including three-factor attacks. Furthermore, we also propose a method to recognize users from their moving patterns using multiple SVM classifiers. Experiments with a dataset with 38 people show that this method can achieve the accuracy up to 92.028%.