ARISE: Authentic Reference-Informed Stability Enhancement for Continual Multimodal Forgery Detection
作者: Yuchen Zhang, Yaxiong Wang, Yujiao Wu, Lianwei Wu, Qi Zhang, Li Zhu, Chaofeng Dong, Le Wu, Zhedong Zheng
分类: cs.CV
发布日期: 2026-10-27
💡 一句话要点
提出ARISE以解决持续多模态伪造检测中的稳定性问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多模态伪造检测 持续学习 高斯混合模型 稳定性增强 脆弱性感知
📋 核心要点
- 现有的多模态伪造检测方法假设训练攻击与实际攻击相似,但这一假设在面对新兴伪造机制时变得脆弱。
- 本文提出ARISE,通过拟合真实数据的高斯混合模型来增强模型的稳定性,同时引入样本自适应的度量以保持灵活性。
- 实验结果显示,ARISE在多个攻击演变协议下的历史保留和最终性能均优于现有的持续学习基线方法。
📝 摘要(中文)
多模态伪造检测通常在静态假设下进行,假设训练攻击与部署后遇到的攻击相似。然而,随着生成器、编辑系统和多模态大语言模型不断引入新的伪造机制,这一假设变得脆弱。本文将问题定义为攻击增量的持续多模态伪造检测,提出ARISE(真实参考信息稳定性增强),通过拟合真实高斯混合模型来构建历史参考,并引入样本自适应的各向异性度量以保持稳定性。实验表明,ARISE在多个攻击演变协议和检测器架构上均表现出更好的历史保留和最终性能。
🔬 方法详解
问题定义:本文解决的是在持续多模态伪造检测中,如何在面对新攻击时保持对历史攻击的记忆和检测能力。现有方法在处理新兴伪造机制时容易遗忘之前的知识。
核心思路:ARISE的核心思路是利用真实数据的几何特性,通过拟合高斯混合模型来构建历史参考,从而增强模型的稳定性并保持对新攻击的适应性。
技术框架:整体架构包括三个主要模块:真实数据的高斯混合模型拟合、样本自适应的各向异性度量(GRASP)和脆弱性感知重放(VAR)。这些模块协同工作,以提升模型的稳定性和适应性。
关键创新:ARISE的关键创新在于引入了样本自适应的度量和脆弱性感知重放机制,这与传统方法不同,后者通常依赖于固定的重放策略,无法动态调整。
关键设计:在设计中,ARISE使用了高斯混合模型来表示真实数据的分布,并通过GRASP模块实现了对历史稳定性方向的强约束,同时在其他方向上保持灵活性。VAR模块则根据训练时的重放统计信息动态调整重放策略。
🖼️ 关键图片
📊 实验亮点
实验结果表明,ARISE在三个攻击演变协议下的历史保留能力显著提升,最终平均性能较代表性的持续学习基线方法提高了XX%。具体而言,在某些协议中,ARISE的准确率提升了YY%,显示出其在动态环境中的有效性。
🎯 应用场景
该研究的潜在应用领域包括数字内容的真实性验证、社交媒体内容监测以及网络安全等。通过提高伪造检测的稳定性和适应性,ARISE可以有效应对不断演变的伪造技术,具有重要的实际价值和未来影响。
📄 摘要(原文)
Multimodal forgery detection is commonly studied under a static assumption: training attacks are expected to approximate those encountered after deployment. This assumption becomes increasingly fragile as generators, editing systems, and multimodal large language models continually introduce new forgery mechanisms and forensic evidence. We formulate this problem as attack-incremental continual multimodal forgery detection, where the binary real/fake interface remains fixed while emerging attacks must be learned without forgetting previous ones. We introduce Fable-Bench, which organizes attack evolution through complementary joint-modality, text-side, and image-side protocols. We further propose ARISE (Authentic Reference-Informed Stability Enhancement), motivated by the open-ended forged class and the comparatively reusable geometry of authentic content. ARISE fits a real-only Gaussian mixture model to projected authentic replay representations as a compact historical reference. Meanwhile, the proposed Gaussian-Responsibility-Aware Stability Preservation (GRASP) derives a sample-adaptive anisotropic metric that strongly constrains historically stable representation directions while retaining plasticity elsewhere. Our Vulnerability-Aware Replay (VAR) further estimates stage-wise vulnerability from authentic-density degradation, forged-to-authentic boundary confusion, and authentic-mode occupancy drift, then assigns bounded replay emphasis to fragile stages. All adaptive signals are obtained from training-time replay statistics, while inference uses only the original multimodal backbone and binary classification head. Experiments across three attack-evolution protocols and multiple detector architectures demonstrate improved historical retention and final average performance over representative continual-learning baselines.