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 2410.08620cs.CRcs.CVcs.MM
收藏

自然语言诱导的对抗图像

Natural Language Induced Adversarial Images

Tsinghua University(清华大学);Chinese Institute for Brain Research (CIBR)(中国脑科学研究所)

Xiaopei Zhu, Peiyang Xu, Guanning Zeng, Yingpeng Dong, Xiaolin Hu

AI总结:

提出一种自然语言诱导的对抗图像攻击方法,利用text-to-image模型结合自适应遗传算法和CLIP生成具有语义一致性的对抗图像,揭示了高频语义信息对分类器的欺骗性及跨模型迁移能力。

摘要:

对抗攻击的研究对AI安全具有重要意义,因为它揭示了深度学习模型的脆弱性,并有助于构建更鲁棒的模型。图像上的对抗攻击研究最为广泛,包括基于噪声的攻击、基于图像编辑的攻击和基于潜在空间的攻击。然而,这些方法构造的对抗样本通常缺乏足够的语义信息,使得人类难以理解深度学习模型在自然条件下的失效模式。为了解决这一局限性,我们提出了一种自然语言诱导的对抗图像攻击方法。核心思想是利用text-to-image模型,根据输入提示词生成对抗图像,这些提示词经过恶意构造,可导致目标模型产生误分类。为了采用商业text-to-image模型合成更自然的对抗图像,我们提出了一种自适应遗传算法(GA)用于在不需要梯度的情况下优化离散对抗提示词,并提出了一种自适应词空间缩减方法以提高查询效率。我们进一步使用CLIP来保持生成图像的语义一致性。在实验中,我们发现一些高频语义信息(如“foggy”、“humid”、“stretching”等)很容易导致分类器出错。这种对抗性语义信息不仅存在于生成图像中,也存在于真实世界拍摄的照片中。我们还发现,一些对抗性语义信息可以迁移到未知的分类任务中。此外,我们的攻击方法可以迁移到不同的text-to-image模型(如Midjourney、DALL-E 3等)和图像分类器。我们的代码可在以下地址获取:https://github.com/zxp555/Natural-Language-Induced-Adversarial-Images。

英文摘要:

Research of adversarial attacks is important for AI security because it shows the vulnerability of deep learning models and helps to build more robust models. Adversarial attacks on images are most widely studied, which include noise-based attacks, image editing-based attacks, and latent space-based attacks. However, the adversarial examples crafted by these methods often lack sufficient semantic information, making it challenging for humans to understand the failure modes of deep learning models under natural conditions. To address this limitation, we propose a natural language induced adversarial image attack method. The core idea is to leverage a text-to-image model to generate adversarial images given input prompts, which are maliciously constructed to lead to misclassification for a target model. To adopt commercial text-to-image models for synthesizing more natural adversarial images, we propose an adaptive genetic algorithm (GA) for optimizing discrete adversarial prompts without requiring gradients and an adaptive word space reduction method for improving query efficiency. We further used CLIP to maintain the semantic consistency of the generated images. In our experiments, we found that some high-frequency semantic information such as "foggy", "humid", "stretching", etc. can easily cause classifier errors. This adversarial semantic information exists not only in generated images but also in photos captured in the real world. We also found that some adversarial semantic information can be transferred to unknown classification tasks. Furthermore, our attack method can transfer to different text-to-image models (e.g., Midjourney, DALL-E 3, etc.) and image classifiers. Our code is available at: https://github.com/zxp555/Natural-Language-Induced-Adversarial-Images.

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。

↑