Unified Face Attack Detection via Fine-Grained Semantic Guidance
For facial recognition security, this work improves detection of diverse attacks by integrating fine-grained semantic guidance, though it is incremental as it builds on an existing dataset and combines known techniques.
The paper addresses face attack detection by enriching the MS-UFAD dataset with fine-grained textual descriptions of forgery cues and proposing DAF-Net to leverage them, achieving superior generalization and performance over vision-only and coarse-grained methods.
The growing applications of facial recognition systems are accompanied by increasingly diverse security threats. Existing datasets lack detailed textual descriptions of forgery cues, leading most prior methods to treat face attack detection primarily as a visual recognition task. In this paper, building upon the large-scale MS-UFAD dataset which contains over 8 million attack images, we enrich each image with a fine-grained textual description of forgery cues. Furthermore, we propose a Dual Alignment Forgery Network(DAF-Net) to better leverage these textual information. Extensive experiments demonstrate that our approach extracts more generalizable and semantically meaningful forgery representations from attack images, outperforming both vision-only methods and approaches based on coarse-grained descriptions.