跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.21689 · v1

A Critical Appraisal and Motivated Intruder Test of A Search Dataset Anonymised for the Purposes of the EU Digital Markets Act

首次提交 2026-11-16 15:33
版本更新 2026-11-16 15:33
雷达观测 2026-11-18 19:45
arXiv ↗PDF ↗

摘要

This study evaluated the European Commission's anonymisation measures for Alphabet's search-data sharing under Article 6(11) of the Digital Markets Act against European Data Protection Board guidance and contemporary practice, and assessed re-identification through motivated intruder testing. The measures focus on search end-users, leaving other individuals mentioned in queries insufficiently protected, and do not explicitly address inference. The accompanying controls and safeguards reflect good practice for managing residual risk. Motivated intruder testing targeted 98,501 individuals across six countries using publicly available profiles. Google-verified matches yielded a re-identification rate of 0.094%, below the study's 0.47% per-team threshold. Each verified match required about three hours (Euro 105), excluding Google verification and Gemini costs, suggesting limited cost deterrence. All country-level rates were below the threshold, but one region exceeded it. The threshold drew on conservative precedents, adjusted for data quality, verification, controls and safeguards, assuming five independent attack teams. Some sensitivity scenarios exceeded the threshold, but relied on strong assumptions, such as treating all uncertain matches as correct. Results may not generalise beyond the targeted population, and ethics requirements constrained the attacks. Personal data detection recall was assumed rather than tested: the minimum defensible recall was 0.88, with 0.976 recommended. However, missed identifiers outside the allow-list would remain in the dataset. Under the study's contextual framework, assumptions and testing limitations, identity-disclosure risk was considered insignificant. This finding does not resolve the gaps concerning inference and other individuals' information; safeguards remain necessary.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-16 15:33 · 最新观测版本

A Critical Appraisal and Motivated Intruder Test of A Search Dataset Anonymised for the Purposes of the EU Digital Markets Act

This study evaluated the European Commission's anonymisation measures for Alphabet's search-data sharing under Article 6(11) of the Digital Markets Act against European Data Protection Board guidance and contemporary practice, and assessed re-identification through motivated intruder testing. The measures focus on search end-users, leaving other individuals mentioned in queries insufficiently protected, and do not explicitly address inference. The accompanying controls and safeguards reflect good practice for managing residual risk. Motivated intruder testing targeted 98,501 individuals across six countries using publicly available profiles. Google-verified matches yielded a re-identification rate of 0.094%, below the study's 0.47% per-team threshold. Each verified match required about three hours (Euro 105), excluding Google verification and Gemini costs, suggesting limited cost deterrence. All country-level rates were below the threshold, but one region exceeded it. The threshold drew on conservative precedents, adjusted for data quality, verification, controls and safeguards, assuming five independent attack teams. Some sensitivity scenarios exceeded the threshold, but relied on strong assumptions, such as treating all uncertain matches as correct. Results may not generalise beyond the targeted population, and ethics requirements constrained the attacks. Personal data detection recall was assumed rather than tested: the minimum defensible recall was 0.88, with 0.976 recommended. However, missed identifiers outside the allow-list would remain in the dataset. Under the study's contextual framework, assumptions and testing limitations, identity-disclosure risk was considered insignificant. This finding does not resolve the gaps concerning inference and other individuals' information; safeguards remain necessary.

查看此版本解读 →查看此版本原文 ↗