论文HF Daily Papers · arXiv 2610.379186 天前

Venus-DefakerOne2视觉取证基座模型发布，物理欺骗检测达93.1%

针对具身智能面临的物理与数字混合伪造威胁，Venus-DefakerOne2基座模型在数千万样本上训练，物理欺骗检测准确率达93.1%，AI视频鉴伪96.0%。采用Tide-Style-Tuning与熵测试自适应技术，显著提升机器人感知安全边界。

AI 整理

感知与导航
评测与安全

本文由 AI 根据公开内容整理，可能有误，请以原文为准：Venus-DefakerOne2: A Vision-Forensics Foundation Model with Broader Coverage（HF Daily Papers · arXiv 2610.37918）

核心内容

针对真实世界中异构的伪造威胁，研究团队发布视觉取证基座模型Venus-DefakerOne2，将反伪造能力从纯数字图像扩展至视频流及物理欺骗场景。该模型基于数千万开源及内部样本训练，在DeepFake人脸、AI生成图像/视频、篡改文档及物理欺骗等六大基准测试中均取得SOTA性能，其中物理欺骗检测准确率达93.1%，AI生成视频检测率达96.0%，大幅超越前代DefakerOne及各类专用专家模型。

技术架构包含两项关键创新：Tide-Style-Tuning策略交替使用LoRA与全参数微调，在平衡计算效率与泛化能力的同时缓解灾难性遗忘；Entropy-based Test Adaptation机制通过分析预测熵动态确定最优预测模态。作为二分类取证任务的基础设施，该模型能有效识别针对机器人视觉系统的打印攻击、屏幕重放等物理 spoofing 手段，为具身智能感知层提供底层安全校验能力。

关键看点

物理欺骗检测：针对打印/屏幕重放等实体攻击检测准确率达93.1%，填补具身感知安全缺口
多模态鉴伪性能：AI生成视频检测率96.0%，DeepFake人脸96.7%，覆盖数字与物理全谱系威胁
Tide-Style-Tuning：交替LoRA与全参微调，兼顾硬样本拟合与通用性，有效抑制灾难性遗忘
熵测试自适应：基于预测熵分析动态选择最优推理模态，提升复杂场景下的决策鲁棒性

工程与产业观察

该模型确立了具身智能视觉安全的新基准，其93.1%的物理欺骗检测率直接回应了机器人易受实体攻击的行业痛点。Tide-Style-Tuning与熵自适应机制为端侧部署提供了高效范式，建议算法团队将其作为感知栈的安全守门员，防范对抗性物理注入攻击。

英文摘要 (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.

Hugging Face Daily Papers 热度：15 upvotes 作者：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 Zh…