1 / 27

Applying Modern Crypto to Advance Social Science

Explore the relationships among individuals within a society and improve the public value of academic research by applying modern cryptography techniques. Protect data privacy, enable secure data processes, and conduct real social science research.

nklein
Télécharger la présentation

Applying Modern Crypto to Advance Social Science

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Applying Modern Crypto to Advance Social Science Ran Cohen (MIT & Northeastern) YarkinDoroz (NJIT) ShafiGoldwasser (MIT) Jason Owen-Smith (IRIS/University of Michigan) Kurt Rohloff (NJIT)

  2. Social Science Relationships among individuals within a society

  3. Social Science + PRIVACY Relationships among individuals within a society

  4. IRIS • Institute for Research on Innovation and Science • Consortium of major research universities • Founded in 2015 • Located at University of Michigan Understand, explain, and improve the public value of academic research

  5. The System Today data process processed DB raw DB Members (universities) Users (researchers)

  6. Accessing IRIS Data output verification final output data process confirm processed DB raw DB output query Virtual secure data enclave (VDE) Lawyer-based security

  7. IRIS Data Use Agreement Lawyer-based security

  8. Accessing Census Data Census data is protected by Title 13 29 FSRDC (Federal Statistical Research Data Centers) data process processed DB raw DB

  9. The System Today • Pros: • It works! • Cons: • Data is visible to users • A lot of bureaucracy • Manual verification of the output • Data visible to IRIS

  10. Evolution of Modern Crypto Proof of concept Feasibility Asymptotic efficiency Concrete efficiency

  11. MPC/FHE

  12. Phase 1 – Encrypted DB output verification final output 2PC confirm output query data process processed DB raw DB Encrypt

  13. Phase 1 – Encrypted DB • Pros: • Data is hidden from users • Less bureaucracy • Control over the leakage • Cons: • Increased runtime • Manual verification of the output • Data visible to IRIS

  14. Proof of Concept Research question: explain the size, composition, and level of student participation of grant funded research teams • Regression analysis • OLS regression • Logistic regression • Poisson regression • Probit regression • Negative binomial regression • Gradient descent • Iterative optimization algorithm to find local miminum • Implemented using the PALISADE FHE library

  15. Concrete Queries • Percentage of students funded on grant Computed using OLS regression • Student participation (at least one student paid on the grant)Computed using Logistic regression • Project size (number of total people employed by a grant)Computed using Poisson regression

  16. Runtime Two test databases • 10,000 records, 11 properties • ~23 seconds per iteration

  17. Runtime Two test databases • 10,000 records, 11 properties • ~23 seconds per iteration • 50,000 records, 14 properties • ~80 seconds per iteration • 7 digits of precision

  18. Phase 2 – Distributed Key output verification final output 2PC confirm output query data process processed DB raw DB Encrypt

  19. Phase 2 – Distributed Key output verification final output MPC confirm output query data process processed DB raw DB Encrypt

  20. Phase 3 – Differential Privacy output verification final output MPC confirm output query data process processed DB raw DB Encrypt

  21. Phase 3 – Differential Privacy final output confirm MPC + DP output query data process processed DB raw DB Encrypt

  22. Phase 4 – Secure Data Process final output confirm MPC + DP output query data process processed DB raw DB Encrypt

  23. Phase 4 – Secure Data Process final output confirm MPC + DP output query secure data process

  24. Summary • Protect IRIS data using FHE & MPC • Better privacy – more data contributors • Less bureaucracy – more researchers • Conduct real social science research

More Related