Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.21211 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917430454910976 |
|---|---|
| author | Oh, Myeong Seok Kim, Dong-Yun Oh, Hanseok Kang, Chaean Kang, Joeun Wang, Xiaonan Park, Hyunjung Jung, Young Cheol Kim, Hansaem |
| author_facet | Oh, Myeong Seok Kim, Dong-Yun Oh, Hanseok Kang, Chaean Kang, Joeun Wang, Xiaonan Park, Hyunjung Jung, Young Cheol Kim, Hansaem |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_21211 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Subject-level Inference for Realistic Text Anonymization Evaluation Oh, Myeong Seok Kim, Dong-Yun Oh, Hanseok Kang, Chaean Kang, Joeun Wang, Xiaonan Park, Hyunjung Jung, Young Cheol Kim, Hansaem Computation and Language 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. |
| title | Subject-level Inference for Realistic Text Anonymization Evaluation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2604.21211 |