CAREER: Distributed Differential Privacy via Secure Multiparty Computation

University Of Vermont & State Agricultural College, Burlington VT

Investigators

Abstract

The increasing rate of collection and analysis of personal data has led to new privacy and security concerns. Differential privacy is a promising framework for protecting individual privacy, but deploying it in practice remains a challenge. Bugs in differential privacy systems are difficult to find, and can result in unexpected privacy failures. In addition, such systems often require collecting sensitive data on central servers; a compromise of these servers could result in a catastrophic loss of privacy. This project aims to develop tools for addressing both challenges. The project's novelties are: (a) new techniques for verifying that programs correctly implement differential privacy, and (b) new applications of cryptography to protect the security of data during processing. The project's broader significance and importance lies in its potential to enable the broader deployment of correct, secure implementations of formal privacy guarantees for individuals in data processing systems. The central goal of this project is to enable the construction of correct, scalable systems that satisfy differential privacy without the need for a trusted data curator. To this end, the project aims to design both new secure protocols and new techniques for ensuring the correctness of systems built with those protocols. Specific research goals of the project include (1) the design of new secure protocols that leverage properties of differential privacy to increase performance and scale to millions of participants; (2) new tools for evaluating these protocols at scale; (3) new automated program analyses for verifying the correctness of secure protocols; and (4) new program analyses and automated testing approaches for checking the correctness of differentially private systems built on secure protocols. This project includes the development of educational materials, including a programming-oriented textbook suitable for an undergraduate course on secure computation. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.