Universal Adversarial Attacks against Closed-Source MLLMs via Target-View Routed Meta Optimization
针对闭源多模态大语言模型的通用对抗攻击:通过目标视角路由元优化
机构 * Rapid-Rich Object Search Lab, Interdisciplinary Graduate Programme, Nanyang Technological University, Singapore(快速丰富目标搜索实验室,跨学科研究生项目,南洋理工大学,新加坡) ; School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore(电气与电子工程学院,南洋理工大学,新加坡) ; College of Computing and Data Science, Nanyang Technological University, Singapore(计算与数据科学学院,南洋理工大学,新加坡) ; DSO National Laboratories, Singapore(新加坡国防科学实验室)
专题命中 多模态训练与对齐 :MLLM(summary_cn,abstract_cn);multimodal(abstract);分类 cs.AI
AI总结 本文提出MCRMO-Attack,通过多作物聚合与注意力引导裁剪稳定监督,通过对齐性门控令牌路由提升token级可靠性,并元学习跨目标扰动先验,从而在商业MLLM上提升未见图像攻击成功率。
Comments This work has been submitted to the IEEE for possible publication