DiffusionFF: Face Forgery Detection via Diffusion-based Artifact Localization

2025-08-03

AI Summary

To address the limited interpretability and coarse localization accuracy in deepfake detection, this paper proposes a diffusion-based artifact localization framework. The method leverages denoising diffusion probabilistic models to generate high-fidelity DSSIM structural discrepancy maps, precisely capturing subtle manipulation traces. It further fuses high-level semantic features extracted from a pre-trained face detector to enable multimodal feature co-modeling. Compared to existing approaches, our framework maintains high detection accuracy while significantly improving fine-grained localization of forged regions. Extensive experiments demonstrate state-of-the-art performance under both cross-dataset and in-dataset evaluation protocols. Moreover, the framework produces intuitive, visually grounded explanations—enhancing model interpretability and fostering user trust without compromising detection reliability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The rapid evolution of deepfake generation techniques demands robust and accurate face forgery detection algorithms. While determining whether an image has been manipulated remains essential, the ability to precisely localize forgery artifacts has become increasingly important for improving model explainability and fostering user trust. To address this challenge, we propose DiffusionFF, a novel framework that enhances face forgery detection through diffusion-based artifact localization. Our method utilizes a denoising diffusion model to generate high-quality Structural Dissimilarity (DSSIM) maps, which effectively capture subtle traces of manipulation. These DSSIM maps are then fused with high-level semantic features extracted by a pretrained forgery detector, leading to significant improvements in detection accuracy. Extensive experiments on both cross-dataset and intra-dataset benchmarks demonstrate that DiffusionFF not only achieves superior detection performance but also offers precise and fine-grained artifact localization, highlighting its overall effectiveness.

Problem

• Detect manipulated face images accurately
• Localize forgery artifacts for model explainability
• Improve detection via diffusion-based DSSIM maps

Innovation

• Uses denoising diffusion for DSSIM maps
• Fuses DSSIM with semantic features
• Enhances detection and artifact localization