Discovery, Replication and Reuse of Sensitive Scientific Data with PSI
Talkat Data Privacy: From Foundations to Applications (Simons Institute workshop, 2019) · 2019-03-04..03-08·
summary
A deployable system called PSI sits on Dataverse and generates differentially private summary releases for sensitive repositories, with a budgeter GUI that lets data owners allocate epsilon (example 0.05, delta 1e-8) and see natural‑language error promises (e.g. mean error ≈1.198), an R library of algorithms for means, histograms, regressions and matching, automated data‑use license generation from metadata, JSON metadata that records mechanism and noise, plans for a certified runnable service, and a cryptographically secure RNG and usability testing showing practical accessibility.
AI-written summary · about
chapters (18)
- 0:00Intro
- 3:00Replication is the foundation
- 6:03Overview
- 9:50Tools
- 14:34Contributors
- 17:16Library of Functions
- 21:10The Problem
- 21:41The Model
- 24:19Scientific Workflow
- 26:07Curator Workflow
- 27:00Exploratory Analysis
- 27:56Workflow
- 29:15Users
- 31:04Differential Privacy
- 33:42Continuous Variables
- 37:57Usability
- 39:16Conclusions
- 39:56Future directions
related talks
39mData Privacy Engineering to power Responsible Enterprise AI
13mDPSynth: From Research to Production—Engineering Differentially Private Synthetic...
01h 00mPractical Difficulties in the Application of Differential Privacy for the Release of Large-Scale...
24mData Versioning Explained the Open Source Way
11mBuilding a Complete Export Ecosystem—From DSAR Automation to Privacy Center
36mData Versioning - does it mean what we think it means?
21mBuilding Trust in Your Data - Improving Your Data Quality with Automated Framework on AWS
41mUsing Data Riches: A Tale of two projects