This is published under the terms of CC BY-NC.

Downloaded: 32 times

Abstract

Biometric authentication systems are facing increasing threats from artificial intelligence-generated content. Previous research has revealed the vulnerability of 2D face authentication systems to master face attacks, which use GAN-based models to create facial samples capable of matching multiple registered user templates in the database. However, the effectiveness of such attacks in 3D scenarios has not been thoroughly investigated.

In this paper, we present a systematic approach to generate master faces that can compromise both 2D and 3D face recognition systems. It uses a latent variable evolution algorithm with a 3D face morphable model. Notably, our approach achieves, for the first time, controllable and morphable master face attacks on face authentication systems. We explore the effect of facial reenactment and face morphing on enhancing the efficacy of master face attacks and reducing the time required for master face generation. Comprehensive simulations of simultaneous master face attacks based on white-box, gray-box, and black-box scenarios demonstrated that our approach achieves superior attack success rates and has advanced flexibility compared with existing methods, highlighting the importance of defending against master face attacks.