Evidence-based Decision Modeling for Synthetic Face Detection with Uncertainty-driven Active Learning

2026-05-10

AI Summary

This work addresses the limitations of existing synthetic face detection methods, which often exhibit overconfidence due to reliance on Softmax, struggle with out-of-distribution samples, and require extensive labeled data. To overcome these challenges, the authors propose an evidential deep learning framework that models class-specific evidence using Dirichlet distributions, thereby explicitly quantifying predictive uncertainty. Notably, this is the first approach to integrate uncertainty estimation with active learning for synthetic face detection, enabling efficient selection of high-informative samples from unlabeled data for annotation. Evaluated across multiple benchmarks, the method achieves a 15% accuracy improvement over current state-of-the-art models, significantly enhancing generalization, robustness, and interpretability while substantially reducing labeling costs.

Abstract

With the rapid development of deep generative models, forged facial images are massively exploited for illegal activities. Although existing synthetic face detection methods have achieved significant progress, they suffer from the inherent limitation of overconfidence due to their reliance on the Softmax activation function. Thus, these methods often lead to unreliable predictions when encountering unknown Out-of-Distribution (OOD) images, and cannot ascertain the model's uncertainty in its prediction. Meanwhile, most existing methods require massive high-quality annotated data, which greatly limits their practicability across diverse scenarios. To address these limitations, we propose EMSFD (Evidence-based decision Modeling for Synthetic Face Detection with uncertainty-driven active learning), an approach designed to enhance detection reliability and generalizability. Specifically, EMSFD models class evidence using the Dirichlet distribution and explicitly incorporates model uncertainty into the prediction process. Furthermore, during training, the estimated uncertainty is exploited to prioritize more informative samples from the unlabeled pool for annotation, thereby reducing labeling cost and improving model generalization. Extensive experimental evaluations demonstrate that our method enhances the interpretability of synthetic face detection. Meanwhile, our method yields a 15\% increase in accuracy compared to existing state-of-the-art (SOTA) baselines, which demonstrates the superior detection performance and generalizability of our approach. Our code is available at: https://github.com/hzx111621/EMSFD.

Problem

• synthetic face detection
• model uncertainty
• Out-of-Distribution
• active learning
• evidence-based modeling

Innovation

• evidence-based modeling
• uncertainty quantification
• active learning
• synthetic face detection
• Dirichlet distribution