English(EN) CBV: Clean-label Backdoor Attacks on Vision Language Models via Diffusion Models

研究人员揭示利用扩散模型和风格特征对AI模型进行新的隐蔽后门攻击

作者 PulseAugur 编辑部 · [2 个来源] · 2026-05-06 04:00

研究人员开发了针对先进AI模型的新型后门攻击方法，特别针对视觉语言模型（VLMs）和扩散模型（DMs）。一种方法CBV利用扩散模型通过微妙地改变图像生成过程并在语义重要区域集中修改，为VLMs创建外观自然的受污染样本。另一种方法Gungnir利用图像内的风格特征作为扩散模型的隐蔽触发器，使攻击更难被检测和绕过现有防御。 AI

影响新的攻击向量凸显了VLMs和扩散模型的漏洞，有必要在AI安全和防御机制方面取得进展。

排序理由两篇研究论文详细介绍了针对AI模型的新型后门攻击方法。

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.AI TIER_1 English(EN) · Ji Guo, Xiaolong Qin, Cencen Liu, Jielei Wang, Jierun Chen, Wenbo Jiang · 2026-05-06 04:00

CBV: Clean-label Backdoor Attacks on Vision Language Models via Diffusion Models

arXiv:2605.02202v1 Announce Type: new Abstract: Vision-Language Models (VLMs) have achieved remarkable success in tasks such as image captioning and visual question answering (VQA). However, as their applications become increasingly widespread, recent studies have revealed that V…
arXiv cs.CV TIER_1 English(EN) · Lei Zhang, Yu Pan, Bingrong Dai, Lin Wang · 2026-05-08 04:00

Gungnir: Exploiting Stylistic Features in Images for Backdoor Attacks on Diffusion Models

arXiv:2502.20650v5 Announce Type: replace Abstract: Diffusion Models (DMs) have achieved remarkable success in image generation, yet recent studies reveal their vulnerability to backdoor attacks, where adversaries manipulate outputs via covert triggers embedded in inputs. Existin…

报道来源 [2]

CBV: Clean-label Backdoor Attacks on Vision Language Models via Diffusion Models

Gungnir: Exploiting Stylistic Features in Images for Backdoor Attacks on Diffusion Models

相关实体

相关话题