Reproduce “DProvDB: Differentially Private Query Processing with Multi-Analyst Provenance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shufan Zhang, Xi He
DProvDB introduces a differentially private query processing system designed for multi-analyst environments, featuring a fine-grained privacy provenance framework to track budget usage per analyst and per data view. The system demonstrates enhanced query utility and proportional fairness among analysts with varying privilege levels, while maintaining strong privacy guarantees even under collusion.
Key insights
- Tracking fine-grained privacy provenance enables equitable and efficient privacy budget allocation in multi-analyst environments, adapting to varying privilege levels and query demands.
- Employing an additive Gaussian mechanism with correlated noise significantly improves query utility (more accurate answers) compared to independent noise addition, especially with increased analyst interaction on shared data.
- Dynamic budget allocation to views, coupled with provenance tracking, optimizes overall system utility by preventing premature budget exhaustion and adapting to evolving query workloads.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.