迫使生成模型退化：数据投毒攻击的力量
Forcing Generative Models to Degenerate Ones: The Power of Data Poisoning Attacks
AI总结：
本文系统研究针对自然语言生成任务的数据投毒攻击，提出评估成功率与隐蔽性的指标，证明仅用1%调优数据即可在微调阶段污染LLM。
摘要：
由第三方训练的大语言模型（LLM）的应用日益广泛，这引发了人们对 LLM 安全漏洞的严重担忧。已有研究表明，恶意行为者可以通过旨在生成不良输出的投毒攻击，隐蔽地利用 LLM 中的这些漏洞。尽管投毒攻击在图像领域（例如目标检测）和分类任务中受到了大量关注，但其对生成模型的影响，尤其是在自然语言生成（NLG）任务领域中的影响，仍缺乏充分理解。为弥补这一差距，我们对各种投毒技术进行了全面探索，以评估它们在一系列生成任务中的有效性。此外，我们提出了一系列指标，用于量化专门针对 NLG 任务设计的投毒攻击的成功率和隐蔽性。通过在多个 NLG 任务、LLM 和数据集上进行大量实验，我们表明，仅使用总调优数据样本的 1%，就有可能在微调阶段成功对 LLM 投毒。本文提出了首个系统性方法，用于在广泛触发器和攻击设置下理解针对 NLG 任务的投毒攻击。我们希望这些发现能够帮助 AI 安全社区针对此类威胁设计适当的防御措施。
英文摘要：
Growing applications of large language models (LLMs) trained by a third party raise serious concerns on the security vulnerability of LLMs.It has been demonstrated that malicious actors can covertly exploit these vulnerabilities in LLMs through poisoning attacks aimed at generating undesirable outputs. While poisoning attacks have received significant attention in the image domain (e.g., object detection), and classification tasks, their implications for generative models, particularly in the realm of natural language generation (NLG) tasks, remain poorly understood. To bridge this gap, we perform a comprehensive exploration of various poisoning techniques to assess their effectiveness across a range of generative tasks. Furthermore, we introduce a range of metrics designed to quantify the success and stealthiness of poisoning attacks specifically tailored to NLG tasks. Through extensive experiments on multiple NLG tasks, LLMs and datasets, we show that it is possible to successfully poison an LLM during the fine-tuning stage using as little as 1\% of the total tuning data samples. Our paper presents the first systematic approach to comprehend poisoning attacks targeting NLG tasks considering a wide range of triggers and attack settings. We hope our findings will assist the AI security community in devising appropriate defenses against such threats.
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。
