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~~
arXiv 2511.16136cs.CV
收藏

噪声如何助力AI生成图像检测

How Noise Benefits AI-generated Image Detection

Engineering Research Center of Digital Forensics, Ministry of Education, Nanjing University of Information Science and Technology(南京信息工程大学数字取证教育部工程研究中心);University of Macau(澳门大学);University of Siena(锡耶纳大学)

Ziqiang Li, Jiazhen Yan, Fan Wang, Kai Zeng, Zhangjie Fu

AI总结:

本文提出PiN-CLIP方法,通过生成噪声和检测网络的联合训练,提升AI生成图像检测的鲁棒性与泛化能力,在开放数据集上取得新的SOTA性能。

摘要:

生成模型的快速发展使真实与合成图像日益难以区分。尽管已有大量研究致力于检测AI生成图像,但分布外泛化仍是一个持续挑战。我们追溯这一弱点到训练过程中利用的虚假捷径,并观察到小的特征空间扰动可以缓解捷径主导。为更可控地解决此问题,我们提出正激励噪声CLIP(PiN-CLIP),在变分正激励原则下联合训练噪声生成器和检测网络。具体而言,我们通过视觉和类别语义特征的交叉注意力融合在特征空间中构造正激励噪声。在优化过程中,噪声被注入特征空间以微调视觉编码器,抑制对捷径敏感的方向,同时增强稳定的取证线索,从而提取更稳健和泛化的伪影表示。在包含42种不同生成模型生成的合成图像的开放数据集上进行了对比实验。我们的方法在平均准确率上比现有方法提高了5.4。

英文摘要:

The rapid advancement of generative models has made real and synthetic images increasingly indistinguishable. Although extensive efforts have been devoted to detecting AI-generated images, out-of-distribution generalization remains a persistent challenge. We trace this weakness to spurious shortcuts exploited during training and we also observe that small feature-space perturbations can mitigate shortcut dominance. To address this problem in a more controllable manner, we propose the Positive-Incentive Noise for CLIP (PiN-CLIP), which jointly trains a noise generator and a detection network under a variational positive-incentive principle. Specifically, we construct positive-incentive noise in the feature space via cross-attention fusion of visual and categorical semantic features. During optimization, the noise is injected into the feature space to fine-tune the visual encoder, suppressing shortcut-sensitive directions while amplifying stable forensic cues, thereby enabling the extraction of more robust and generalized artifact representations. Comparative experiments are conducted on an open-world dataset comprising synthetic images generated by 42 distinct generative models. Our method achieves new state-of-the-art performance, with notable improvements of 5.4 in average accuracy over existing approaches.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑