---
license: apache-2.0
---

The dataset accompanying the paper "Robustness and Generalizability of Deepfake Detection: A Study with Diffusion Models".

[Website] [paper] [GitHub].

Introduction

Welcome to the DeepFakeFace (DFF) dataset! Here we present a meticulously curated collection of artificial celebrity faces, crafted using cutting-edge diffusion models. Our aim is to tackle the rising challenge posed by deepfakes in today's digital landscape.

Here are some example images in our dataset:

Our proposed DeepFakeFace(DFF) dataset is generated by various diffusion models, aiming to protect the privacy of celebrities. There are four zip files in our dataset and each file contains 30,000 images. We maintain the same directory structure as the IMDB-WIKI dataset where real images are selected.

• inpainting.zip is generated by the Stable Diffusion Inpainting model.
• insight.zip is generated by the InsightFace toolbox.
• text2img.zip is generated by Stable Diffusion V1.5
• wiki.zip contains original real images selected from the IMDB-WIKI dataset.

DeepFake Dataset Compare

We compare our dataset with previous datasets here:

Experimental Results

Performance of RECCE across different generators, measured in terms of Acc (%), AUC (%), and EER (%):

Robustness evaluation in terms of ACC(%), AUC (%) and EER(%):

Cite

Please cite our paper if you use our codes or our dataset in your own work:

@misc{song2023robustness,
      title={Robustness and Generalizability of Deepfake Detection: A Study with Diffusion Models}, 
      author={Haixu Song and Shiyu Huang and Yinpeng Dong and Wei-Wei Tu},
      year={2023},
      eprint={2309.02218},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}