REVIEW 14 cited by
DF40: Toward Next-Generation Deepfake Detection
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
DF40: Toward Next-Generation Deepfake Detection
read the original abstract
We propose a new comprehensive benchmark to revolutionize the current deepfake detection field to the next generation. Predominantly, existing works identify top-notch detection algorithms and models by adhering to the common practice: training detectors on one specific dataset (e.g., FF++) and testing them on other prevalent deepfake datasets. This protocol is often regarded as a "golden compass" for navigating SoTA detectors. But can these stand-out "winners" be truly applied to tackle the myriad of realistic and diverse deepfakes lurking in the real world? If not, what underlying factors contribute to this gap? In this work, we found the dataset (both train and test) can be the "primary culprit" due to: (1) forgery diversity: Deepfake techniques are commonly referred to as both face forgery and entire image synthesis. Most existing datasets only contain partial types of them, with limited forgery methods implemented; (2) forgery realism: The dominated training dataset, FF++, contains out-of-date forgery techniques from the past four years. "Honing skills" on these forgeries makes it difficult to guarantee effective detection generalization toward nowadays' SoTA deepfakes; (3) evaluation protocol: Most detection works perform evaluations on one type, which hinders the development of universal deepfake detectors. To address this dilemma, we construct a highly diverse deepfake detection dataset called DF40, which comprises 40 distinct deepfake techniques. We then conduct comprehensive evaluations using 4 standard evaluation protocols and 8 representative detection methods, resulting in over 2,000 evaluations. Through these evaluations, we provide an extensive analysis from various perspectives, leading to 7 new insightful findings. We also open up 4 valuable yet previously underexplored research questions to inspire future works. Our project page is https://github.com/YZY-stack/DF40.
Forward citations
Cited by 14 Pith papers
-
Toward Calibrated, Fair, and accurate Deepfake Detection
Face-Feature Tuning is a label-free logit remapping method that reduces FPR/TPR gaps across groups in deepfake detection while preserving overall accuracy.
-
SynCred-Bench: Benchmarking Synthetic Credibility in AI-Generated Visual Misinformation
SynCred-Bench shows that 15 MLLMs reach only 10.5% TPR, open-source detectors under 5%, commercial APIs 57.6%, and humans 63% TPR at 5% FPR when identifying AI-generated images with synthetic credibility.
-
The Regularizing Power of Language-Training Deepfake Detectors
A dual-encoder deepfake detector pairs a frozen specialist with a LoRA-tuned MLLM, trained first via binary alignment then via RL to reward explain-then-classify behavior, yielding improved cross-dataset performance a...
-
Direct Discrepancy Replay: Distribution-Discrepancy Condensation and Manifold-Consistent Replay for Continual Face Forgery Detection
A replay method for continual face forgery detection condenses real-fake distribution discrepancies into compact maps and synthesizes compatible samples from current real faces to reduce forgetting under tight memory ...
-
The DeepSpeak Dataset
DeepSpeak provides over 100 hours of consented, identity-matched real and modern deepfake audiovisual content focused on talking heads, with evaluations showing existing detectors fail to generalize without retraining.
-
Continuously Evolving Deepfake Detection: An Architecture and Public-Benchmark Evaluation of a Dynamic Detection System
A continuously refreshed, incentive-driven deepfake detector beats static detectors on in-the-wild benchmarks and improves on post-export AI-generated media.
-
Unveiling Deepfakes: A Frequency-Aware Triple Branch Network for Deepfake Detection
A frequency-aware triple-branch network with mutual information-based decoupling and fusion losses achieves state-of-the-art deepfake detection across six benchmarks.
-
Open Set Face Forgery Detection via Dual-Level Evidence Collection
DLED reformulates open-set face forgery detection as an uncertainty estimation task and uses dual-level spatial-frequency evidence collection to identify novel fake categories, claiming 20% average gains over baselines.
-
When Generative Replay Meets Evolving Deepfakes: Dual Confusion-Aware Regularization for Incremental Face Forgery Detection
Adaptively weighting direct supervision versus a relative-separation loss makes generative replay work for incremental deepfake detection.
-
Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection
Orthogonal subspace decomposition via SVD on vision foundation model features preserves high-rank pre-trained knowledge by freezing principal components and adapting residuals, reducing overfitting for better generali...
-
InfoDense: Density-Aware Regional Decisive Replay for Memory-Efficient Incremental Face Forgery Detection
InfoDense replays only density-ranked, forgery-decisive face fragments rather than full images, cutting memory use and improving incremental deepfake detection.
-
Generalizable Audio Spoofing Detection using Non-Semantic Representations
Frozen non-semantic TRILLson embeddings with a lightweight backend beat prior spoofing detectors on out-of-domain datasets while staying competitive in-domain.
-
Deepfake Detection in Social Media: A Temporal Artifact Analysis Using 3D Convolutional Neural Networks
3D CNN detector with temporal consistency regularizer reaches 92.8% accuracy on DeepfakeTIMIT and 76.4% cross-dataset on FaceForensics++ without fine-tuning.
-
Robust Deepfake Detection, NTIRE 2026 Challenge: Report
The NTIRE 2026 challenge finds that large foundation models combined with ensembles and degradation-aware training produce the most robust deepfake detectors.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.