English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Synthetic data is seen as the most promising solution to shareindividual-level data while preserving privacy. Shadow modeling-basedMembership Inference Attacks (MIAs) have become the standard approach toevaluate the privacy risk of synthetic data. While very effective, they requirea large number of datasets to be created and models trained to evaluate therisk posed by a single record. The privacy risk of a dataset is thus currentlyevaluated by running MIAs on a handful of records selected using ad-hocmethods. We here propose what is, to the best of our knowledge, the firstprincipled vulnerable record identification technique for synthetic datapublishing, leveraging the distance to a record's closest neighbors. We showour method to strongly outperform previous ad-hoc methods across datasets andgenerators. We also show evidence of our method to be robust to the choice ofMIA and to specific choice of parameters. Finally, we show it to accuratelyidentify vulnerable records when synthetic data generators are madedifferentially private. The choice of vulnerable records is as important asmore accurate MIAs when evaluating the privacy of synthetic data releases,including from a legal perspective. We here propose a simple yet highlyeffective method to do so. We hope our method will enable practitioners tobetter estimate the risk posed by synthetic data publishing and researchers tofairly compare ever improving MIAs on synthetic data.

Achilles' Heels: Vulnerable Record Identification in Synthetic Data Publishing