arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.02255cs.CRcs.AI
收藏

大语言模型的隐私性:一项消融研究

On the Privacy of LLMs: An Ablation Study

UDST(多哈大学);CUQ-Ulster University(卡塔尔阿尔斯特大学)

Karima Makhlouf, Lamiaa Basyoni, Syed Khaderi, Gabriel Marquez, Peter Sotomango, Mahmoud Awawdah, Sami Zhioua

AI总结:

本文通过对四种LLM代表性攻击的系统消融研究,明确不同攻击的隐私敏感性,进而给出隐私感知型LLM部署的实用建议。

摘要:

大语言模型(LLMs)面临多样的隐私威胁,但现有研究通常孤立评估各类攻击,难以明确哪些系统因素会导致隐私漏洞。本文提出统一威胁模型,并对四种代表性攻击开展系统消融研究:成员推断(MIA)、属性推断(AIA)、数据提取(DEA)和后门攻击(BA)。研究发现不同攻击存在特定敏感性:成员推断高度依赖数据集特征与攻击策略;属性推断受模型及训练特征影响;数据提取成功率随模型规模和训练数据重复次数增加而上升;后门攻击始终保持高成功率。基于上述发现,本文给出隐私感知型LLM部署的实用建议,包括优先评估检索增强系统的成员泄露风险、减少训练数据中重复的个人信息、监控属性推断风险、保护微调流程以防范后门攻击。

英文摘要:

Large language models (LLMs) face diverse privacy threats, yet existing studies typically evaluate individual attacks in isolation, offering limited guidance on which system factors drive privacy vulnerabilities. This paper presents a unified threat model and a systematic ablation study of four representative attacks: Membership Inference (MIA), Attribute Inference (AIA), Data Extraction (DEA), and Backdoor Attacks (BA). Our findings reveal attack-specific sensitivities: membership inference depends strongly on dataset characteristics and attack strategy, attribute inference is influenced by model and training characteristics, data extraction increases with model scale and training-data repetition, and backdoor attacks achieve consistently high success rates. Based on these findings, we derive practical recommendations for privacy-aware LLM deployment, including prioritizing membership leakage assessments in retrieval-augmented systems, minimizing repeated personal information in training data, monitoring attribute inference risks, and safeguarding fine-tuning pipelines against backdoors.

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。

↑