BriefGPT.xyz
Feb, 2024
OrderBkd: 文字背门攻击的重新定位
OrderBkd: Textual backdoor attack through repositioning
HTML
PDF
Irina Alekseevskaia, Konstantin Arkhipenko
TL;DR
借助特定词语在句子中的重新定位作为触发器,设计和应用基于词性标注的规则来选择这些词汇,在保持高攻击成功率的同时,优于现有攻击的困惑度和与清洁样本的语义相似性。
Abstract
The use of
third-party datasets
and
pre-trained machine learning models
poses a threat to
nlp systems
due to possibility of hidden
→