APSIPA Transactions on Signal and Information Processing > Vol 13 > Issue 1

A Survey on Deep Learning-based Face Anti-Spoofing

Pei-Kai Huang, National Tsing Hua University, Taiwan, Jun-Xiong Chong, National Tsing Hua University, Taiwan, Ming-Tsung Hsu, National Tsing Hua University, Taiwan, Fang-Yu Hsu, National Tsing Hua University, Taiwan, Cheng-Hsuan Chiang, National Tsing Hua University, Taiwan, Tzu-Hsien Chen, National Tsing Hua University, Taiwan, Chiou-Ting Hsu, National Tsing Hua University, Taiwan, cthsu@gapp.nthu.edu.tw
 
Suggested Citation
Pei-Kai Huang, Jun-Xiong Chong, Ming-Tsung Hsu, Fang-Yu Hsu, Cheng-Hsuan Chiang, Tzu-Hsien Chen and Chiou-Ting Hsu (2024), "A Survey on Deep Learning-based Face Anti-Spoofing", APSIPA Transactions on Signal and Information Processing: Vol. 13: No. 1, e34. http://dx.doi.org/10.1561/116.20240053

Publication Date: 09 Dec 2024
© 2024 P. Huang, J. Chong, M. Hsu, F. Hsu, C. Chiang, T. Chen, and C. Hsu
 
Subjects
Security and privacy,  Artificial intelligence methods in security and privacy,  Classification and prediction,  Deep learning
 
Keywords
Face anti-spoofingpresentation attack3D mask attackliveness featureone-class detection
 

Share

Open Access

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

Downloaded: 136 times

In this article:
Introduction 
Common Challenges in FAS 
Two-class Face Anti-spoofing Methods 
One-class Face Anti-spoofing Methods 
Face Anti-spoofing Datasets and Evaluation Metrics 
Conclusion and Research Directions 
References 

Abstract

Face anti-spoofing (FAS) aims to distinguish live images and facial spoof attacks to defend facial recognition systems. Thanks to advancements in deep learning, recent deep learning-based FAS methods have shown promising potential, especially in effectively addressing newly developed attacks. In this survey, we first provide an overview of common challenges in FAS and then recap recent advances in deep learning-based FAS. In particular, these anti-spoofing methods generally fall into two main categories, i.e., two-class FAS and one-class FAS. Recent two-class FAS methods have employed a wide range of techniques in developing FAS models, including auxiliary supervision, local descriptor-enhanced feature learning, disentangled feature learning, meta learning, adversarial learning, data augmentation, long-range dependency learning, and multimodal learning. Meanwhile, recent one-class FAS methods have utilized reconstruction supervision, statistical learning, and generative feature learning to learn liveness features solely from live images. In this survey, we also provide an overview of publicly available FAS datasets. Finally, we summarize recent FAS development and highlight some potential future research directions.

DOI:10.1561/116.20240053