A survey on 3D mask presentation attack detection and countermeasures
A survey on 3D mask presentation attack detection and countermeasures
复制标题
DOI：
10.1016/j.patcog.2019.107032
复制
发表时间：
2020-02-01
影响因子：
8
通讯作者：
Xu, Zhengquan
中科院分区：
文献类型：
--
作者：
Jia, Shan;Guo, Guodong;Xu, Zhengquan
尽管在人脸识别方面取得了令人印象深刻的进展，但当前的系统容易受到呈现攻击，呈现攻击通过呈现人脸伪像来破坏人脸识别系统。已经开发了几种技术来自动检测不同的呈现攻击，主要是针对2D照片打印和视频回放攻击。然而，随着3D建模和打印技术的发展，3D面具已经成为攻击人脸识别系统的一种更有效的方法。在过去的十年中，已经提出了各种检测方法的3D面具攻击，但还没有调查总结的进展。我们提出了一个全面的概述，最先进的方法在3D掩模欺骗和反欺骗，包括现有的数据库和对策。此外，我们定量地比较了不同掩码欺骗检测方法在共同基础上的性能（即，使用相同的数据库和评估度量）。几个二维演示攻击检测方法的有效性也评估了两个3D面具欺骗数据库，以显示它们是否适用于3D面具攻击。最后，我们提出了一些见解和总结开放的问题，以解决在未来。(C)2019爱思唯尔有限公司版权所有。
Despite the impressive progress in face recognition, current systems are vulnerable to presentation attacks, which subvert the face recognition systems by presenting a face artifact. Several techniques have been developed to automatically detect different presentation attacks, mostly for 2D photo print and video replay attacks. However, with the development of 3D modeling and printing technologies, 3D mask has become a more effective way to attack the face recognition systems. Over the last decade, various detection methods for 3D mask attacks have been proposed, but there is no survey yet to summarize the advances. We present a comprehensive overview of the state-of-the-art approaches in 3D mask spoofing and anti-spoofing, including existing databases and countermeasures. In addition, we quantitatively compare the performance of different mask spoofing detection methods on a common ground (i.e., using the same database and evaluation metric). The effectiveness of several 2D presentation attack detection methods is also evaluated on two 3D mask spoofing databases to show whether they are applicable or not for 3D mask attacks. Finally, we present some insights and summarize open issues to address in the future. (C) 2019 Elsevier Ltd. All rights reserved.