3D Synthetic On-Air Signatures for Biometric Attack

Abstract

The importance of authentication is paramount when it comes to the protection of resources, encompassing both physical corporate facilities and digital assets. Among the different authentication methodologies, 3D on-air signatures represent a novel behavioral biometric that is highly challenging to imitate, even when observed from a distance. However, the emergence of robotic arms capable of mimicking human movements poses a potential security vulnerability for these emerging schemes. This paper focuses on examining the possibility of generating 3D fingertip movements that accurately mimic a signature, with the intention of circumventing a biometric system. We utilize the iDeLog3D algorithm to produce artificial movements of the fingertips, based on the 3D Sigma-Lognormal model and the Kinematic Theory of Rapid Movements. We then replicate a 3D on-air database using this model and proceed to compare the performance of a DTW-based signature verifier using both real and synthetic signatures. During the experiments, it has been demonstrated that synthetic signatures are effective in bypassing a signature verifier that is exclusively trained using real signatures. This underscores the importance of incorporating an additional layer of security, such as a bot detector, into these types of biometric schemes.