Clean-label backdoor attack and defense: An examination of language model vulnerability

Shuai Zhao, Xiaoyu Xu, Luwei Xiao, Jinming Wen, Luu Anh Tuan

Abstract: Highlights•We propose a novel clean-label backdoor method using prompts as triggers.•We first explore defense algorithms against backdoor attacks that leverage LoRA.•Our attack method achieves state-of-the-art attack success rates.

External IDs:dblp:journals/eswa/ZhaoXXWT25