BriefGPT.xyz
Feb, 2024
多模式大型语言模型的测试期后门攻击
Test-Time Backdoor Attacks on Multimodal Large Language Models
HTML
PDF
Dong Lu, Tianyu Pang, Chao Du, Qian Liu, Xianjun Yang...
TL;DR
通过对多模态大型语言模型进行测试时背门攻击,使用对抗测试图像将背门注入文本模态中,无需获取或修改训练数据,并能在设置和激活有害效果的时间上解耦。
Abstract
backdoor attacks
are commonly executed by contaminating training data, such that a trigger can activate predetermined harmful effects during the test phase. In this work, we present AnyDoor, a
test-time
backdoor
→