arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

EIB-Net:用于可泛化AI生成图像检测的熵引导信息瓶颈

EIB-Net: Entropy-Guided Information Bottleneck for Generalizable AI-Generated Image Detection

Zhida Zhang, Xinlei Ma, Jie Cao

arXiv 2609.29064首次发表:更新:

发表机构

UCAS; NLPR, CASIA; Institute of Automation, Chinese Academy of Sciences(中国科学院大学; 中国科学院自动化研究所模式识别国家重点实验室; 中国科学院自动化研究所)

机构由 AI 辅助整理,请以论文原文为准。

AI 中文总结

针对AI生成图像检测的泛化难题,提出熵引导信息瓶颈网络EIB-Net,利用图像熵选择低纹理补丁并经变分信息瓶颈提取紧凑特征,仅用2%训练数据达85.7%准确率,跨生成器泛化稳健。

AI 中文摘要

逼真AI生成图像的激增要求开发出能够跨多种生成模型进行泛化的稳健检测方法。尽管现有方法针对具有局部伪影的基于操作的伪造,但基于生成的图像(例如来自扩散模型的图像)缺乏此类痕迹,这构成了根本性挑战。我们观察到,生成模型优先考虑全局语义而牺牲局部纹理保真度,使得低纹理区域成为合成来源的关键指标。为了利用这一点,我们提出了EIB-Net,一种熵引导的信息瓶颈网络。EIB-Net引入了一种新颖的图像熵(IE)度量,用于自动选择信息量最大(最低熵)的补丁,然后使用变分信息瓶颈(VIB)对其进行处理,以学习紧凑、可泛化的特征。在DIFF、DiffusionForensics和GenImage基准上的大量实验证明了最先进的性能:EIB-Net仅使用2%的训练数据就达到了85.7%的准确率,比全图像基线高出15%以上,并保持了稳健的跨生成器泛化能力(在GenImage上平均准确率为83.5%)。此外,我们的熵引导补丁选择(EGPL)持续增强了多种骨干网络(CNN和Transformer),证明了其在数据高效检测中的实用价值。

英文摘要

The proliferation of photorealistic AI-generated images demands robust detection methods that generalize across diverse generative models. While existing approaches target manipulation-based forgeries with local artifacts, generation-based images (e.g., from diffusion models) lack such traces, posing a fundamental challenge. We observe that generative models prioritize global semantics at the expense of local texture fidelity, making low-texture regions key indicators of synthetic origin. To exploit this, we propose EIB-Net, an Entropy-guided Information Bottleneck Network. EIB-Net introduces a novel Image Entropy (IE) metric to automatically select the most informative (lowest-entropy) patch, then processes it with a Variational Information Bottleneck (VIB) to learn compact, generalizable features. Extensive experiments on DIFF, DiffusionForensics, and GenImage benchmarks demonstrate state-of-the-art performance: EIB-Net achieves 85.7\% accuracy using only 2\% of training data, outperforming full-image baselines by over 15\%, and maintains robust cross-generator generalization (83.5\% average accuracy on GenImage). Furthermore, our entropy-guided patch selection (EGPL) consistently enhances diverse backbones (CNNs and Transformers), proving its practical value for data-efficient detection.

CommentsAccept by ICME 2026

论文原文

arXiv 摘要页 · PDF 原文 · HTML 原文

↑