FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning

TL;DR

This paper introduces FSFM, a self-supervised facial representation learning framework that enhances the generalization of face security tasks such as deepfake detection and anti-spoofing, outperforming existing methods.

Contribution

Proposes a novel self-supervised pretraining approach combining masked image modeling and instance discrimination for robust facial representations.

Findings

Outperforms supervised pretraining and existing self-supervised methods.

Effective across multiple face security tasks and datasets.

Enhances transferability and generalization in face security applications.

Abstract

This work asks: with abundant, unlabeled real faces, how to learn a robust and transferable facial representation that boosts various face security tasks with respect to generalization performance? We make the first attempt and propose a self-supervised pretraining framework to learn fundamental representations of real face images, FSFM, that leverages the synergy between masked image modeling (MIM) and instance discrimination (ID). We explore various facial masking strategies for MIM and present a simple yet powerful CRFR-P masking, which explicitly forces the model to capture meaningful intra-region consistency and challenging inter-region coherency. Furthermore, we devise the ID network that naturally couples with MIM to establish underlying local-to-global correspondence via tailored self-distillation. These three learning objectives, namely 3C, empower encoding both local features and global semantics of real faces. After pretraining, a vanilla ViT serves as a universal vision foundation model for downstream face security tasks: cross-dataset deepfake detection, cross-domain face anti-spoofing, and unseen diffusion facial forgery detection. Extensive experiments on 10 public datasets demonstrate that our model transfers better than supervised pretraining, visual and facial self-supervised learning arts, and even outperforms task-specialized SOTA methods.