The value of posture, build and dynamics in gesture-based user authentication

摘要: User authentication based on biometrics such as fingerprint, iris, face, speech or gait has been around for many years. Recently, intentional user gestures have shown to be a promising modality authentication. However, it is unclear how much of the performance can attributed pure biometric information that no control over, individual limb lengths, and gesture dynamics, fully control. A related question is: How easy copy these dynamics? In this paper, we propose framework decompose into three components: initial posture, proportions, dynamics. We then study impact each component various combinations gesture-based using dataset 36 users performing 3 varying complexity. also spoof attacks same show, somewhat surprisingly, amateurs are unable with sufficient accuracy so significantly degrade overall even when they trained closest to. While training certainly improves an attacker's ability seems unique proportions (which cannot altered) posture attackers fail pay attention to), more than make up loss due compromised dynamics always renewed).