Computing Statistics When Data Cannot Be Shared
The Inter-university Consortium for Political and Social Research (ICPSR) at the University of Michigan is collaborating with Stealth Software on a system to securely compute statistics on sensitive databases without revealing the underlying data that will be constructed with the latest developments in secure Multiparty Computation (MPC). Data commonly collected about education will be synthesized by members of our research team, so that the MPC system can be tested on large scale databases with no risks to subjects.
The ICPSR will be responsible for the following:
1. ICPSR will assist Stealth Software in defining statistical algorithms to be implemented in the prototype MPC system.
2. ICPSR will create synthetic data for testing the MPC system.
3. In consultation with developers at Stealth Software, ICPSR will design and implement a system for testing MPC prototype on a cloud computing platform. Successive versions of the prototype will be installed for testing as they are delivered.
4. ICPSR will create realistic analytical scenarios involving statistical computations performed by the MPC system. These scenarios will be used to test the successive versions of the MPC prototype as increasingly complex statistical procedures are implemented.
Funding: Laura and John Arnold Foundation (subcontract: Services Subcontract Agreement)
Funding Period: 7/1/2017 to 6/30/2020