Gen-LRA is a computationally efficient no-box MIA that exploits local overfitting in tabular generative models to produce a closed-form density-ratio statistic with a provable mean-score gap between members and non-members.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Privacy Auditing Synthetic Data Release through Local Likelihood Attacks
Gen-LRA is a computationally efficient no-box MIA that exploits local overfitting in tabular generative models to produce a closed-form density-ratio statistic with a provable mean-score gap between members and non-members.