SID: Secure data sharing for Industrial Digitalization

Differential privacy and federated learning are promising methods for protecting sensitive data, but there are currently obstacles to the widespread use of these technologies within Swedish industry. The technology can be easily misunderstood and misused. Such misunderstandings sometimes lead to privacy guarantees without practical value that can nevertheless give the impression that data is protected. Computational and communication performance can also be a challenge when privacy-preserving mechanisms are applied together with distributed machine learning.

This project will carry out case studies with real datasets and perform privacy and performance analysis of differential privacy and federated learning. The case studies, with associated analysis and identification of pitfalls, will serve as a practical guide for the application of privacy-preserving data analysis. The goal is to make privacy-preserving techniques ready for widespread use in Swedish industry, and promote data-driven development by enabling increased use of sensitive data for analysis and machine learning.