Masked Conditional Diffusion Model for Enhancing Deepfake Detection

Tiewen Chen,Shanmin Yang,Shu Hu,Zhenghan Fang,Ying Fu,Xi Wu,Xin Wang

2024-02-01

Abstract:Recent studies on deepfake detection have achieved promising results when training and testing faces are from the same dataset. However, their results severely degrade when confronted with forged samples that the model has not yet seen during training. In this paper, deepfake data to help detect deepfakes. this paper present we put a new insight into diffusion model-based data augmentation, and propose a Masked Conditional Diffusion Model (MCDM) for enhancing deepfake detection. It generates a variety of forged faces from a masked pristine one, encouraging the deepfake detection model to learn generic and robust representations without overfitting to special artifacts. Extensive experiments demonstrate that forgery images generated with our method are of high quality and helpful to improve the performance of deepfake detection models.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper aims to address the issue of data augmentation in deepfake detection. Specifically, it proposes a new method called the Masked Conditional Diffusion Model (MCDM) to generate diverse fake face images, thereby improving the generalization ability and robustness of deepfake detection models. Existing deepfake detection methods experience significant performance degradation when faced with unseen datasets. MCDM, by partially masking the original images and using a conditional diffusion model for completion, generates high-quality and consistent fake images, thus helping to train more general and powerful deepfake detection models.

Experimental results show that compared to existing methods, MCDM generates higher quality images and demonstrates better generalization ability on multiple challenging datasets. Additionally, ablation studies validate the effectiveness of random mask conditions and feature-level reconstruction loss, proving the importance of these components in enhancing the performance of detection models.