首次以个人为单位评估文本匿名化，发现90%信息遮蔽仍难防隐私泄露。
Subject-level Inference for Realistic Text Anonymization Evaluation
- 以个人而非文本片段为评估单位，更贴近真实攻击场景
- 即使90%敏感信息被遮蔽，仍有超6成个人身份可被推断
- 聚焦特定目标的匿名化反而让其他人暴露更多
当前文本匿名化评估依赖基于片段的指标，无法反映攻击者实际能推断的信息，且假设单一数据主体，忽略多主体场景。为此，我们提出SPIA（Subject-level PII Inference Assessment），首个将评估单位从文本片段转向个体的基准，包含675份法律与在线领域文档，并引入新型个体级保护度量。大量实验表明，即使超过90%的PII片段被遮蔽，个体级推理保护率仍低至33%，多数个人信息仍可通过上下文推断恢复。此外，针对特定目标的匿名化会使非目标主体暴露程度显著高于目标主体。结果表明，基于个体推理的评估对保障真实场景下文本匿名化安全至关重要。
原文摘要 · Abstract (English)
Current text anonymization evaluation relies on span-based metrics that fail to capture what an adversary could actually infer, and assumes a single data subject, ignoring multi-subject scenarios. To address these limitations, we present SPIA (Subject-level PII Inference Assessment), the first benchmark that shifts the unit of evaluation from text spans to individuals, comprising 675 documents across legal and online domains with novel subject-level protection metrics. Extensive experiments show that even when over 90% of PII spans are masked, subject-level inference protection drops as low as 33%, leaving the majority of personal information recoverable through contextual inference. Furthermore, target-subject-focused anonymization leaves non-target subjects substantially more exposed than the target subject. We show that subject-level inference-based evaluation is essential for ensuring safe text anonymization in real-world settings.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。