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

Training generative models to produce synthetic data is meant to provide aprivacy-friendly approach to data release. However, we get robust guaranteesonly when models are trained to satisfy Differential Privacy (DP). Alas, thisis not the standard in industry as many companies use ad-hoc strategies toempirically evaluate privacy based on the statistical similarity betweensynthetic and real data. In this paper, we review the privacy metrics offeredby leading companies in this space and shed light on a few critical flaws inreasoning about privacy entirely via empirical evaluations. We analyze theundesirable properties of the most popular metrics and filters and demonstratetheir unreliability and inconsistency through counter-examples. We then presenta reconstruction attack, ReconSyn, which successfully recovers (i.e., leaks allattributes of) at least 78% of the low-density train records (or outliers) withonly black-box access to a single fitted generative model and the privacymetrics. Finally, we show that applying DP only to the model or usinglow-utility generators does not mitigate ReconSyn as the privacy leakagepredominantly comes from the metrics. Overall, our work serves as a warning topractitioners not to deviate from established privacy-preserving mechanisms.

On the Inadequacy of Similarity-based Privacy Metrics: Reconstruction Attacks against "Truly Anonymous Synthetic Data''