Riivo Talviste "Applying Secure Multi-Party Computation in Practice" (supervisors sen. researcher Sven Laur (Institute of Computer Science, UT) and project manager Dan Bogdanov (AS Cybernetica))
Data is useful only when used. This is especially true if one is able to combine several data sets. For example, combining income and educational data, it is possible for a government to get a return of investment overview of educational investments. The same is true in private sector. Combining data sets of financial obligations of their customers, banks could issue loans with lower credit risks.
However, this kind of data sharing is often forbidden as citizens and customers have their privacy expectations. Moreover, such a combined database becomes an interesting target for both hackers as well as nosy officials and administrators taking advantage of their position.
Secure multi-party computation is a technology that allows several parties to collaboratively analyse data without seeing any individual values. This technology is suitable for the above mentioned scenarios protecting user privacy from both insider and outsider attacks. With first practical applications using secure multi-party computation developed in 2000s, the technology is now mature enough to be used in distributed deployments and even offered as part of a service.
In this work, we present solutions for technical difficulties in deploying secure multi-party computation in real-world applications. We will first give a brief overview of the current state of the art, bring out several shortcomings and address them.
The main contribution of this work is an end-to-end process description of deploying secure multi-party computation for the first large-scale registry-based statistical study on linked databases. Involving large stakeholders like government institutions introduces also some non-technical requirements like signing contracts and negotiating with the Data Protection Agency.
Looking into the future, we propose to deploy secure multi-party computation technology as a service on a federated data exchange infrastructure. This allows privacy-preserving analysis to be carried out faster and more conveniently, thus promoting a more informed government.