arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2408.09878cs.CR
收藏

通过知识蒸馏在大语言模型间转移后门

Transferring Backdoors between Large Language Models by Knowledge Distillation

Pengzhou Cheng, Zongru Wu, Tianjie Ju, Wei Du, Zhuosheng Zhang Gongshen Liu

AI总结:

本文针对大语言模型后门攻击的模型可迁移性风险,提出ATBA自适应可迁移后门攻击方法,通过TTG和ATO模块仅用干净微调即可将教师模型后门转移到小型模型,后门迁移率超80%,兼具鲁棒性与隐蔽性,为LLM安全敲响警钟。

摘要:

后门攻击一直是大语言模型(LLM)面临的严重安全漏洞。然而,以往的方法仅揭示了特定模型中的此类风险,或是展示了在预训练阶段实施攻击后的任务可迁移性。那么,后门攻击的模型可迁移性风险究竟有多高?本文聚焦于现有小型LLM是否可能通过知识蒸馏(KD),在无意识中被带毒的教师LLM植入后门知识。具体而言,我们提出了ATBA(自适应可迁移后门攻击),该方法仅需执行干净微调,就能有效将教师LLM的后门蒸馏到小型模型中。我们首先提出目标触发器生成(TTG)模块,该模块基于余弦相似度分布从词元列表中筛选出一组具有指示性的候选触发器。随后,我们利用影子模型模拟蒸馏过程,并引入自适应触发器优化(ATO)模块,通过基于梯度的贪心反馈搜索最优触发器。大量实验表明,ATBA不仅能为学生模型提供正向引导,还能隐式转移后门知识。该攻击具备鲁棒性与隐蔽性,后门迁移率超过80%,希望能引起安全领域的关注。

英文摘要:

Backdoor Attacks have been a serious vulnerability against Large Language Models (LLMs). However, previous methods only reveal such risk in specific models, or present tasks transferability after attacking the pre-trained phase. So, how risky is the model transferability of a backdoor attack? In this paper, we focus on whether existing mini-LLMs may be unconsciously instructed in backdoor knowledge by poisoned teacher LLMs through knowledge distillation (KD). Specifically, we propose ATBA, an adaptive transferable backdoor attack, which can effectively distill the backdoor of teacher LLMs into small models when only executing clean-tuning. We first propose the Target Trigger Generation (TTG) module that filters out a set of indicative trigger candidates from the token list based on cosine similarity distribution. Then, we exploit a shadow model to imitate the distilling process and introduce an Adaptive Trigger Optimization (ATO) module to realize a gradient-based greedy feedback to search optimal triggers. Extensive experiments show that ATBA generates not only positive guidance for student models but also implicitly transfers backdoor knowledge. Our attack is robust and stealthy, with over 80% backdoor transferability, and hopes the attention of security.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑