Unleash the Power of Ellipsis: Accuracy-enhanced Sparse Vector Technique with Exponential Noise
Summary: Revisits SVT privacy analysis by exploiting that SVT only releases binary exceedance bits, enabling a less-conservative DP accounting that admits exponential noise as optimal for perturbation. Introduces utility-driven threshold correction and appending strategies to counter exponential-noise bias, boosting precision/recall and improving query accuracy up to ~50% theoretically and empirically. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
Authors
- 1. Yuhan Liu (Renmin University of China)
- 2. Sheng Wang (Alibaba)
- 3. Yixuan Liu (Renmin University of China)
- 4. Feifei Li (Alibaba)
- 5. Hong Chen (Renmin University of China)
BibTeX Citation
@article{liu_vldb25,
title = {{Unleash the Power of Ellipsis: Accuracy-enhanced Sparse Vector Technique with Exponential Noise}},
author = {Liu, Yuhan and Wang, Sheng and Liu, Yixuan and Li, Feifei and Chen, Hong},
journal = {PVLDB},
series = {{VLDB} '25},
volume = {18},
number = {2},
pages = {187--199},
doi = {10.14778/3705829.3705838},
url = {https://doi.org/10.14778/3705829.3705838},
year = {2025}
}Incoming Citations (Sorted by Pagerank)
Showing 1 of 1 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 10,385 | N2E: A General Framework to Reduce Node-Differential Privacy to Edge-Differential Privacy for Graph Analytics | 2026 | SIGMOD | 5.093636e-05 |
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 2 of 2 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 2,012 | Understanding the Sparse Vector Technique for Differential Privacy | 2017 | VLDB | 9.3073552e-05 |
| 2,542 | Privacy at Scale: Local Differential Privacy in Practice | 2018 | SIGMOD | 8.4460386e-05 |
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 10,335 | A General Framework for Per-record Differential Privacy | 2026 | SIGMOD |
| 2 | 5,555 | Practical Differential Privacy via Grouping and Smoothing | 2013 | VLDB |
| 3 | 611 | Optimizing Linear Counting Queries Under Differential Privacy | 2010 | PODS |
| 4 | 11,280 | Confidence Intervals for Private Query Processing | 2024 | VLDB |
| 5 | 11,203 | Personalized Truncation for Personalized Privacy | 2024 | SIGMOD |
| 6 | 7,968 | Optimizing Fitness-For-Use of Differentially Private Linear Queries | 2021 | VLDB |
| 7 | 3,935 | Output Perturbation with Query Relaxation | 2008 | VLDB |
| 8 | 1,504 | A Data- and Workload-Aware Algorithm for Range Queries Under Differential Privacy | 2014 | VLDB |
| 9 | 9,569 | Free Gap Information from the Differentially Private Sparse Vector and Noisy Max Mechanisms | 2020 | VLDB |
| 10 | 2,012 | Understanding the Sparse Vector Technique for Differential Privacy | 2017 | VLDB |