Metric-Normalized Posterior Leakage (mPL): Attacker-Aligned Privacy for Joint Consumption
Gaoyi Chen, Minghao Li, Weishi Shi, Yan Huang, Yusheng Wei, Sourabh Yadav, Chenxi Qiu · 2026
Read Metric-Normalized Posterior Leakage (mPL): Attacker-Aligned Privacy for Joint Consumption by Gaoyi Chen, Minghao Li, Weishi Shi, Yan Huang, Yusheng Wei, Sourabh Yadav, Chenxi Qiu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-01137.txt.
Description
Gaoyi Chen and six co-authors introduce metric-normalized posterior leakage, a privacy measure designed for systems consumed under joint observation. Metric differential privacy scales noise to semantic distance, but per-record guarantees can miss leakage when learned aggregators combine evidence across correlated items. Their measure is attacker-aligned and distance-calibrated. It tracks the shift in posterior odds induced by releases.
For single or independent releases, bounding the new measure is equivalent to metric differential privacy, but under joint observation the older guarantee can still leave leakage high. A version bounded in probability limits how often leakage may exceed a budget, and an adaptive trust-and-verify framework perturbs, audits with a learned attacker, and adjusts parameters. A word-embedding case study shows substantial reduction in violations with low utility loss.
Catalog record
| CREATOR | Gaoyi Chen, Minghao Li, Weishi Shi, Yan Huang, Yusheng Wei, Sourabh Yadav, Chenxi Qiu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 90 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.01137