Gaze-based Presentation Attack Detection for Users Wearing Tinted Glasses

Abstract

Biometric authentication is vulnerable to presentation (spoofing) attacks. It is important to address the security vulnerability of spoofing attacks where an attacker uses an artefact presented at the sensor to subvert the system. Gaze-tracking has been proposed for such attack detection. In this paper, we explore the sensitivity of a gaze-based approach to spoofing detection in the presence of eye-glasses that may impact detection performance. In particular, we investigate the use of partially tinted glasses such as may be used in hazardous environments or outdoors in mobile application scenarios The attack scenarios considered in this work include the use of projected photos, 2D and 3D masks. A gaze-based spoofing detection system has been extensively evaluated using data captured from volunteers performing genuine attempts (with and without wearing such tinted glasses) as well as spoofing attempts using various artefacts. The results of the evaluations indicate that the presence of tinted glasses has a small impact on the accuracy of attack detection, thereby making the use of such gaze-based features possible for a wider range of applications.