Venus-DefakerOne2 detects diverse fake media with broad coverage
Venus-DefakerOne2: A Vision-Forensics Foundation Model with Broader Coverage
Computer Vision and Pattern Recognition
Summary
Fake images and videos come in many forms, such as digital edits or physical impersonations, and can appear in photos, videos, and documents. The authors developed Venus-DefakerOne2, a unified system that can spot different types of forgeries across various media. It learns from tens of millions of examples and uses smart tuning methods to perform well on new and tricky cases. This new system outperforms older tools at detecting fake faces, AI-made images, doctored documents, physical spoofs, and deepfake videos. It serves as a versatile tool to help tell real from fake in many forensic settings.
What this means in practice
- •For digital security teams: Use the model to detect various fake media types including images, videos, and documents in real-time security monitoring.
- •For social media platform developers: Integrate the system to identify and reduce the spread of AI-generated and manipulated visual content.
Authors
Chenfan Qu, Yachao Liang, Mingqi Fang, Jiangwei Xie, Zhenming Wang, Baoyue Zhang, Changjiang Jiang, Zheng Chang, Fanli Jin, Xuekang Zhu, Kaiwen Feng, Junchi Li, Weihong Zhang, Haoyue Wang, Chenfeng Zhang, Yinan Xia, Fangming Cui, Dong Fan, Jingjing Liu, Tianyi Zhang, Luobin Chen, Ji-Zhe Zhou, Sheng Li, Wei Wang, Bin Li, Ran He, Xinpeng Zhang, Linna Zhou, Jian Liu, Weiqiang Wang
Abstract
Real-world forgery is inherently heterogeneous: it manifests not only as digital manipulation but also as physical spoofing, and is carried not only by images but also by videos. Despite substantial progress, a unified anti-forgery foundation model covering such a broad threat spectrum remains absent. In this paper, we present Venus-DefakerOne2, which substantially extends the forensic capability boundary along both the forgery-form and media-type dimensions. Trained on tens of millions of open-source and in-house samples, it incorporates two key designs. Tide-Style-Tuning alternates LoRA and full-parameter fine-tuning to balance computational efficiency, generalization, and hard-sample fitting while mitigating catastrophic forgetting. Entropy-based Test Adaptation determines the optimal prediction modality by analyzing prediction entropy. Venus-DefakerOne2 achieves new state-of-the-art performance across a comprehensive suite of benchmarks, including DeepFake faces (96.7%), AI-generated images (93.3%), tampered documents (87.6%), manipulated natural images (89.6%), physical spoofs (93.1%), and AI-generated videos (96.0%). It significantly outperforms both specialized expert models and its predecessor, DefakerOne, establishing itself as a powerful and versatile foundation model for binary-decision forensic tasks.