arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2408.12798cs.AI
收藏

BackdoorLLM:面向大语言模型后门攻击与防御的综合基准

BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models

Singapore Management University(新加坡管理大学);The University of Melbourne(墨尔本大学);Fudan University(复旦大学)

Yige Li, Hanxun Huang, Yunhan Zhao, Xingjun Ma, Jun Sun

AI总结:

针对大语言模型在开放式文本生成场景下后门脆弱性研究不足的问题,提出首个综合后门攻防基准BackdoorLLM,覆盖多类攻击、场景与模型,并提供防御工具包与开源资源。

摘要:

生成式大语言模型(LLM)已在广泛任务上取得了当前最优的结果,但它们仍易受后门攻击:输入中精心设计的触发器可操纵模型生成攻击者指定的输出。此前研究主要聚焦于视觉与分类场景下的后门风险,而开放式文本生成中LLM的脆弱性仍未得到充分探究。为填补这一空白,我们推出了BackdoorLLM(我们的BackdoorLLM基准在由AI安全中心(https://safe.ai/)举办的SafetyBench竞赛中荣获一等奖,相关信息见https://www.mlsafety.org/safebench/winners),这是首个用于系统评估文本生成LLM后门威胁的综合基准。BackdoorLLM提供:(i)具备标准化训练与评估流程的统一基准库;(ii)多样化的攻击模式集合,包括数据投毒、权重投毒、隐藏状态操纵与思维链劫持;(iii)涵盖8种不同攻击策略、7种真实场景与6种模型架构的200余项实验;(iv)对LLM中后门效果的影响因素与失效模式的关键洞见;(v)包含7种代表性缓解技术的防御工具包。我们的代码与数据集可在https://github.com/bboylyg/BackdoorLLM获取。我们将持续纳入新兴的攻击与防御方法,以助力提升LLM安全性与可靠性的研究。

英文摘要:

Generative large language models (LLMs) have achieved state-of-the-art results on a wide range of tasks, yet they remain susceptible to backdoor attacks: carefully crafted triggers in the input can manipulate the model to produce adversary-specified outputs. While prior research has predominantly focused on backdoor risks in vision and classification settings, the vulnerability of LLMs in open-ended text generation remains underexplored. To fill this gap, we introduce BackdoorLLM (Our BackdoorLLM benchmark was awarded First Prize in the SafetyBench competition, https://www.mlsafety.org/safebench/winners, organized by the Center for AI Safety, https://safe.ai/.), the first comprehensive benchmark for systematically evaluating backdoor threats in text-generation LLMs. BackdoorLLM provides: (i) a unified repository of benchmarks with a standardized training and evaluation pipeline; (ii) a diverse suite of attack modalities, including data poisoning, weight poisoning, hidden-state manipulation, and chain-of-thought hijacking; (iii) over 200 experiments spanning 8 distinct attack strategies, 7 real-world scenarios, and 6 model architectures; (iv) key insights into the factors that govern backdoor effectiveness and failure modes in LLMs; and (v) a defense toolkit encompassing 7 representative mitigation techniques. Our code and datasets are available at https://github.com/bboylyg/BackdoorLLM. We will continuously incorporate emerging attack and defense methodologies to support the research in advancing the safety and reliability of LLMs.

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。

↑