An event series for Turing-Roche partnership updates, knowledge sharing and new perspectives. Find out more about the series.
This event will be exploring the topic of synthetic data- which is artificially created data that simulates accurate, real-world settings. These datasets allow researchers to generate insights without having to access sensitive data.
About the event
During the event we'll be hearing from Florimond Houssiau, Research Associate at the Alan Turing Institute and Marta Battle and Finn Janson who are both Data Scientists at Roche.
Florimond will be discussing how to evaluate the privacy of synthetic data: synthetic data has been proposed as a solution to using and sharing sensitive data without privacy issues, as records in a synthetic dataset are not linked to real persons. Recent research has however shown that this is not enough in itself to protect the privacy of users. In this talk, he will present a toolbox being developed at Turing to comprehensively evaluate privacy risks of synthetic data generators for a wide range of datasets.
Marta and Finn will discuss how to unlock access to patient data, by generating synthetic data that leaks no information about individual patients. The talk will focus on the state of the technology, its challenges and how they could be overcome. This includes approaches to quality and privacy evaluation, and in particular, challenges in generating high-quality longitudinal datasets.
We will be hosting a Q&A at the event where we hope to get insights from the audience on how to innovate and solve problems in synthetic data generation. We also hope to have discussions about this area pre and post the event via our Turing-Roche Slack Workspace which you can join here.
You can watch a recording of this event here.