Statistical inference in privacy-preserving data processing systems

Recent advances in mathematical and computational sciences have provided innovative ways to explore large and complex datasets. New solutions extract meaning from big data, offering ways to make informed decisions that drive advances in the standard of living. However, individuals, governments and organizations often rely on solutions that need shared databases of personal information. Consequentially, end-users often express security and privacy concerns and wish to have the legal right to gain control of data that is being shared and to understand its processing. Unfortunately, data aiming to bring insights on habits, culture or personal opinions is rarely complete.

Existing inferential approaches for big data problems have not adapted; providers have currently no means to attend user worries without decays in the quality of their services. Hence, there exists an urgent need to better understand and improve on privacy-preserving distributed approaches to inference and uncertainty quantification on big data problems with incomplete information.

This project explores distributed statistical and computational approaches in order to create flexible privacy-preserving inferential frameworks addressing data-driven discovery challenges concerned with human behaviour, consumer or health data. The work brings together an interdisciplinary team committed to strengthen the understanding of the mathematical and computational methods that drive improvements within modern digital societies and economies.