PulseAugur
实时 09:26:31
English(EN) VPA-Guard: Defending and Benchmarking Image-to-Video Generation Against Visual Prompt Attacks

新的防御框架VPA-Guard解决了AI视频生成中的视觉提示攻击问题

研究人员推出了一种新颖的防御框架VPA-Guard,旨在保护图像到视频(I2V)生成模型免受视觉提示攻击。这些攻击利用箭头或草图等视觉线索来操纵模型生成有害内容。为了解决这个问题,该团队还开发了VVA-Bench,这是第一个专门用于评估I2V模型针对此类攻击安全性的基准。在VVA-Bench上的实验表明,当前最先进的模型极易受到攻击,在某些模型上的成功率高达100%。VPA-Guard利用检索增强和自演化能力,显著降低了攻击成功率和有害性得分,同时保持了模型的效用。 AI

影响 增强了AI视频生成的安全协议,可能有助于更负责任地部署多模态AI系统。

排序理由 该集群包含一篇详细介绍AI模型安全新防御机制和基准的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →

新的防御框架VPA-Guard解决了AI视频生成中的视觉提示攻击问题

报道来源 [3]

  1. Hugging Face Daily Papers TIER_1 English(EN) ·

    VPA-Guard: Defending and Benchmarking Image-to-Video Generation Against Visual Prompt Attacks

    Recent advancements in Image-to-Video (I2V) generation have transformed input images from simple appearance references into interactive control interfaces where visual cues such as arrows, sketches, and emojis orchestrate complex video dynamics with unprecedented controllability.…

  2. arXiv cs.CV TIER_1 English(EN) · Yining Sun, Haoyu Kang, Jiajun Wu, Heng Zhang, Danyang Zhang, Zhenjun Zhao, Haochen Han, Fangming Liu, Wai Kin Victor Chan, Alex Jinpeng Wang ·

    VPA-Guard: Defending and Benchmarking Image-to-Video Generation Against Visual Prompt Attacks

    arXiv:2606.25592v1 Announce Type: new Abstract: Recent advancements in Image-to-Video (I2V) generation have transformed input images from simple appearance references into interactive control interfaces where visual cues such as arrows, sketches, and emojis orchestrate complex vi…

  3. arXiv cs.CV TIER_1 English(EN) · Alex Jinpeng Wang ·

    VPA-Guard: Defending and Benchmarking Image-to-Video Generation Against Visual Prompt Attacks

    Recent advancements in Image-to-Video (I2V) generation have transformed input images from simple appearance references into interactive control interfaces where visual cues such as arrows, sketches, and emojis orchestrate complex video dynamics with unprecedented controllability.…