Negotiating Privacy/Utility Trade-Offs under Differential Privacy
summary
The College Scorecard project links Department of Education cohorts to IRS earnings data and moves from ad hoc suppression to differential privacy, producing millions of statistics across institutions and programs in 2020–21. A negotiation workflow and a prototype interface let analysts adjust epsilon budget and granularity to meet concrete fitness‑for‑use error targets; outcome: more published statistics with comparable accuracy, while IRS gains principled incremental risk accounting and fewer arbitrary suppressions.
AI-written summary · about
chapters (7)
- 0:00Start
- 5:37Informal Privacy Protection
- 7:47Differential Privacy
- 7:59Background on Differential Privacy
- 10:14Release Description
- 14:01Outcomes
- 14:41Conclusions and Challenges
related talks
01h 05mMarkets and Incentives for Transacting on (Personal) Information, Part III
20mShare data without revealing personal information
30mAccuracy First: Selecting a Differential Privacy Level for Accuracy-Constrained...
27mPrivacy-preserving statistical analyses
19mPractical Considerations for Differential Privacy
10mCommunicating Differential Privacy Guarantees to Data Subjects
06mLessons Learned in Communicating Differential Privacy to Data Subjects
11mMy $5MM Differential Privacy Visualizations