Epistemic Parity: Reproducibility as an Evaluation Metric for Differential Privacy
Summary: Introduce 'epistemic parity': measure how often peer-reviewed empirical conclusions (reproduced on ICPSR datasets) persist when rerun on DP synthetic data. Benchmark shows SOTA synthesizers often achieve high parity at practical ε but some claims remain unreproducible, motivating utility-first DP mechanisms and application-specific risk models. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Lucas Rosenblatt (New York University)
- 2. Bernease Herman (University of Washington)
- 3. Anastasia Holovenko (Ukrainian Catholic University)
- 4. Wonkwon Lee (New York University)
- 5. Joshua Loftus (London School of Economics)
- 6. Elizabeth McKinnie (Microsoft)
- 7. Taras Rumezhak (Ukrainian Catholic University)
- 8. Andrii Stadnik (Ukrainian Catholic University)
- 9. Bill Howe (University of Washington)
- 10. Julia Stoyanovich (New York University)
BibTeX Citation
@article{rosenblatt_vldb23,
title = {{Epistemic Parity: Reproducibility as an Evaluation Metric for Differential Privacy}},
author = {Rosenblatt, Lucas and Herman, Bernease and Holovenko, Anastasia and Lee, Wonkwon and Loftus, Joshua and McKinnie, Elizabeth and Rumezhak, Taras and Stadnik, Andrii and Howe, Bill and Stoyanovich, Julia},
journal = {PVLDB},
series = {{VLDB} '23},
volume = {16},
number = {11},
pages = {3178--3191},
doi = {10.14778/3611479.3611517},
url = {https://doi.org/10.14778/3611479.3611517},
year = {2023}
}Incoming Citations (Sorted by Pagerank)
Showing 0 of 0 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 6 of 6 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 123 | Revealing Information while Preserving Privacy | 2003 | PODS | 0.00031082693 |
| 1,169 | PrivBayes: Private Data Release via Bayesian Networks | 2014 | SIGMOD | 0.00011838753 |
| 2,167 | Optimizing error of high-dimensional statistical queries under differential privacy | 2018 | VLDB | 9.047803e-05 |
| 2,300 | Principled Evaluation of Differentially Private Algorithms using DPBench | 2016 | SIGMOD | 8.7833594e-05 |
| 2,476 | Data Synthesis via Differentially Private Markov Random Fields | 2021 | VLDB | 8.5258582e-05 |
| 2,841 | AIM: An Adaptive and Iterative Mechanism for Differentially Private Synthetic Data | 2022 | VLDB | 8.0637668e-05 |
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 10,254 | HeteroFedSyn: Differentially Private Tabular Data Synthesis for Heterogeneous Federated Settings | 2026 | SIGMOD |
| 2 | 7,639 | Budget Sharing for Multi-Analyst Differential Privacy | 2021 | VLDB |
| 3 | 2,300 | Principled Evaluation of Differentially Private Algorithms using DPBench | 2016 | SIGMOD |
| 4 | 7,909 | Differentially Private Data Generation with Missing Data | 2024 | VLDB |
| 5 | 7,358 | PreFair: Privately Generating Justifiably Fair Synthetic Data | 2023 | VLDB |
| 6 | 10,335 | A General Framework for Per-record Differential Privacy | 2026 | SIGMOD |
| 7 | 11,026 | PrivEval: a tool for interactive evaluation of privacy metrics in synthetic data generation | 2025 | VLDB |
| 8 | 11,130 | Continual Release of Differentially Private Synthetic Data from Longitudinal Data Collections | 2024 | PODS |
| 9 | 3,016 | Plausible Deniability for Privacy-Preserving Data Synthesis | 2017 | VLDB |
| 10 | 10,345 | Benchmarking Differentially Private Tabular Data Synthesis: [Experiments & Analysis] | 2026 | SIGMOD |