模型的语言很重要：大语言模型的比较隐私分析
The Model's Language Matters: A Comparative Privacy Analysis of LLMs
AI总结：
本研究首次定量证明语言结构影响大语言模型的隐私泄露，意大利语泄露最强，法语和西班牙语更稳健，需语言感知的隐私保护机制。
摘要：
大语言模型（LLMs）越来越多地部署在涉及敏感数据的多语言应用中，然而其规模和语言变异性带来了重大的隐私风险。本文主要针对英语进行评估，研究了语言结构如何影响在英语、西班牙语、法语和意大利语医学语料库上训练的LLMs的隐私泄露。我们量化了六个语言指标，并评估了三种攻击向量：提取、反事实记忆和成员推断。结果表明，隐私脆弱性随语言冗余度和分词粒度而增加：意大利语表现出最强的泄露，而英语显示出更高的成员可分性。相比之下，法语和西班牙语由于更高的形态复杂性而表现出更强的韧性。总体而言，我们的发现首次提供了语言在隐私泄露中起作用的定量证据，强调了在LLM部署中需要具有语言意识的隐私保护机制。
英文摘要：
Large Language Models (LLMs) are increasingly deployed across multilingual applications that handle sensitive data, yet their scale and linguistic variability introduce major privacy risks. Mostly evaluated for English, this paper investigates how language structure affects privacy leakage in LLMs trained on English, Spanish, French, and Italian medical corpora. We quantify six linguistic indicators and evaluate three attack vectors: extraction, counterfactual memorization, and membership inference. Results show that privacy vulnerability scales with linguistic redundancy and tokenization granularity: Italian exhibits the strongest leakage, while English shows higher membership separability. In contrast, French and Spanish display greater resilience due to higher morphological complexity. Overall, our findings provide the first quantitative evidence that language matters in privacy leakage, underscoring the need for language-aware privacy-preserving mechanisms in LLM deployments.
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。
