DDL: A Large-Scale Dataset for Deepfake Detection and Localization in Diversified Real-World Scenarios
Authors: Changtao Miao, Yi Zhang, Weize Gao, Zhiya Tan, Weiwei Feng, Man Luo, Jianshu Li, Ajian Liu, Yunfeng Diao, Qi Chu, Tao Gong, Zhe Li, Weibin Yao, Joey Tianyi Zhou
Organizations: AntGroup · A⋆STAR Centre for Frontier AI Research · Institute of Automation, Chinese Academy of Sciences · Hefei University of Technology · Anhui Province Key Laboratory of Digital Security
Recent advances in AIGC have exacerbated the misuse of malicious deepfake content, making the development of reliable deepfake detection methods an essential means to address this challenge. Although existing deepfake detection models demonstrate outstanding performance in detection metrics, most methods only provide simple binary classification results, lacking interpretability. Recent studies have attempted to enhance the interpretability of classification results by providing spatial manipulation masks or temporal forgery segments. However, due to the limitations of forgery datasets, the practical effectiveness of these methods remains suboptimal. The primary reason lies in the fact that most existing deepfake datasets contain only binary labels, with limited variety in forgery scenarios, insufficient diversity in deepfake types, and relatively small data scales, making them inadequate for complex real-world scenarios. To address this predicament, we construct a novel large-scale deepfake detection and localization (DDL) dataset containing 1.4M+ forged samples and encompassing 80 distinct deepfake methods. The DDL design incorporates four key innovations: (1) Comprehensive Deepfake Methods (covering 7 different generation architectures and a total of 80 methods), (2) Varied Manipulation Modes (incorporating 7 classic and 3 novel forgery modes), (3) Diverse Forgery Scenarios and Modalities (including 3 scenarios and 3 modalities), and (4) Fine-grained Forgery Annotations (providing 1.18M+ precise spatial masks and 0.23M+ precise temporal segments). Through these improvements, our DDL not only provides a more challenging benchmark for complex real-world forgeries but also offers crucial support for building next-generation deepfake detection, localization, and interpretability methods.
Figures & tables
Figure 1. Overview of our DDL dataset. It shows the DDL’s advantages in comprehensive deepfake methods, varied manipulation modes, diverse forgery scenarios and modalities, and fine-grained forgery annotations.
Datasets
Publication
Tasks
Modality
Scenarios
Latest Deepfake
Methods
#Fake
FF++ ( Rossler et al., 2019 )
ICCV’ 19
Cla
V
SF
NeuralTextures (2019)
4
4K
Celeb-DF ( Li et al., 2020 )
CVPR’ 20
Cla
V
SF
Unknown
1
5K+
DF-1.0 ( Jiang et al., 2021 )
CVPR’ 20
Cla
V
SF
DF-VAE (2020)
1
10K
DFDC ( Dolhansky et al., 2020 )
Arxiv’ 20
Cla
V
SF
StyleGAN (2018)
8
0.1M+
FFIW ( Zhou et al., 2021 )
CVPR’ 21
Cla/SL
V
MF
FSGAN (2019)
3
10K
OpenForensics ( Le et al., 2021 )
ICCV’ 21
SL
I
MF
InterFaceGAN (2020)
2
0.1M
Table 1. Comparison of existing deepfake datasets. Our DDL surpasses others in terms of the diversity of deepfake methods and the scale of fake samples. Cla: Binary classification. SL: Spatial forgery localization. TL: Temporal forgery localization. A: Audio modality. I: Image modality. V: Video modality. SF: Single-Face scenario. MF: Multi-Face scenario. AV: Audio-Visual scenario. Methods: The total number of deepfake methods in the dataset. #Fake: Number of forged samples, each deepfake image, video, or audio is considered as one sample.
Table 2. Statistics of original data sources in the DDL dataset.
Subsets
Train
Valid
Test
Total
Real
Fake
Real
Fake
Real
Fake
Real
Fake
DDL-I
156K
799K
39K
199K
58K
182K
253K
1180K
DDL-AV
68K
134K
5K
9K
26K
88K
99K
231K
Table 3. Statistics of DDL datasets.
Datasets
FS
FR
FFS
FE
HFF
DE
RE
IS
AVAM
AVFS
DDL (ours)
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
DF40
✓
✓
✓
✓
×
×
×
×
×
×
AV-DF1M
×
×
×
×
×
✓
✓
✓
×
×
ForgeryNet
✓
✓
×
✓
×
×
×
×
×
×
Table 4. Statistics of forgery modes in deepfake datasets.
Figure 2. Dataset statistical analysis. (a) Number of deepfake methods under different generation model architectures. "Others" denotes the total of VAE, NeRF, Normalizing Flow, and Autoregressive. (b) Proportion of test set samples originating from the same source as the training set.
Models
Valid Set
Test Set
DDL-I Dataset
ACC
AUC
F1
IoU
ACC
AUC
F1
IoU
UnivFD ( Ojha et al., 2023 )
83.64
71.50
–
–
76.12
70.14
–
–
DualNet ( Xi et al., 2023 )
80.97
83.21
–
–
74.65
65.96
–
–
SPSL ( Liu et al., 2021 )
81.34
79.27
–
–
72.79
65.65
–
–
IML-ViT ( Ma et al., 2023 )
90.39
95.07
73.45
69.28
80.35
86.52
45.77
41.93
MSCCNet ( Miao et al., 2025b )
99.83
99.95
98.14
96.79
78.17
88.88
66.39
49.69
Table 5. Benchmark Experiments on DDL datasets.
Figure 3. Robustness experiments of DDL-I (top) and DDL-AV (bottom) datasets. Clean represents the raw test set.
This paper presents a systematic study of scaling laws for the deepfake detection task. Specifically, we analyze the model performance against the number of real image domains, deepfake generation methods, and training images. Since no existing dataset meets the scale requirements for this research, we construct ScaleDF, the largest dataset to date in this field, which contains over 5.8 million real images from 51 different datasets (domains) and more than 8.8 million fake images generated by 102 deepfake methods. Using ScaleDF, we observe power-law scaling similar to that shown in large language models (LLMs). Specifically, the average detection error follows a predictable power-law decay as either the number of real domains or the number of deepfake methods increases. This key observation not only allows us to forecast the number of additional real domains or deepfake methods required to reach a target performance, but also inspires us to counter the evolving deepfake technology in a data-centric manner. Beyond this, we examine the role of pre-training and data augmentations in deepfake detection under scaling, as well as the limitations of scaling itself.The ScaleDF dataset is available at https://huggingface.co/datasets/WenhaoWang/ScaleDF.
The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety. However, existing deepfake video benchmarks provide limited coverage of recent synthesis methods and generally lack reliable fine-grained textual annotations. Meanwhile, conventional detectors and multimodal large language models (MLLMs), whether operating as a single model or relying on a single analytical perspective, often fail to capture subtle forgery artifacts, limiting their generalization to emerging AI-generated methods. To address these limitations, we introduce FaceVid-Forensics-100K, a large-scale deepfake video dataset comprising 100,000 videos and spanning 33 synthesis methods across face swapping, face reenactment, and entire-face synthesis, including recent generators such as Seedance 2.0. The dataset provides fine-grained textual annotations of visual observations and verdict-consistent forensic explanations, automatically synthesized through a multi-model aggregation and conflict-resolution pipeline powered by advanced MLLMs. Building on this benchmark, we propose a multi-agent forensic reasoning framework that employs four specialized domain-expert agents to independently analyze forgery cues from four perspectives: texture, lighting, motion, and physics. A judge agent then reconciles their reports to produce a final prediction together with an explanation. Extensive evaluations on out-of-domain test sets show that, despite being composed entirely of small open-source MLLMs, our framework outperforms all methods including closed-source GPT and Gemini models and ranks first across all reported metrics on this benchmark. The project page is available at https://xavierjiezou.github.io/ARGUS/.
Xuechao Zou, Shun Zhang, Kai Li +6
Beijing Jiaotong University · Tsinghua University · Ant Group
In Deepfake Detection (DFD) tasks, researchers proposed two types of MLLM-based methods: complementary combination with small DFD detectors, or static forgery knowledge injection. The lack of professional forgery knowledge hinders the performance of these DFD-MLLMs. To solve this, we deeply considered two insightful issues: How to provide high-quality associated forgery knowledge for MLLMs? AND How to endow MLLMs with critical reasoning abilities given noisy reference information? Notably, we attempted to address above two questions with preliminary answers by leveraging the combination of Retrieval-Augmented Generation (RAG) and Reinforcement Learning (RL). Through RAG and RL techniques, we propose the VRAG-DFD framework with accurate dynamic forgery knowledge retrieval and powerful critical reasoning capabilities. Specifically, in terms of data, we constructed two datasets with RAG: Forensic Knowledge Database (FKD) for DFD knowledge annotation, and Forensic Chain-of-Thought Dataset (F-CoT), for critical CoT construction. In terms of model training, we adopt a three-stage training method (Alignment->SFT->GRPO) to gradually cultivate the critical reasoning ability of the MLLM. In terms of performance, VRAG-DFD achieved SOTA and competitive performance on DFD generalization testing.