Analyse sensitive data with your own teams and outside partners
For internal analysts as well as contracted consultants and external analysts, in work ranging from disease and medication research to energy forecasting and data exchange between government bodies.
Request demoTrusted for research and data sharing by leading organisations
Why the usual ways of sharing data hold research back
Personal data cannot simply be passed on
Research data often describes real people in privacy-sensitive detail. Sharing it with a colleague or an outside analyst is never free of risk.
Anonymisation breaks the data or leaves a risk
Traditional anonymisation forces a choice. Remove enough detail to protect people and the analysis loses what it needs. Keep more of it, and individuals can still be re-identified.
The real data should stay where it is
Sometimes real data cannot or should not leave its source, whether across a border, to an external party or within your own organisation. Data minimisation means real records only move when strictly necessary.
What shared research data needs, and how BlueGen provides it
No way back to individuals
Whoever works with the data, a colleague or a contracted consultant, must not be able to trace any record to a real person, even when the research asks for fine detail.
Generation steered by your research question
Your research question sets the configuration for the generation process. The parts of the data it depends on keep the highest quality, and less related parts are loosened slightly so privacy stays fully protected. The result can in principle be shared onwards too.
Quality close to the real data
Research stands or falls with quality. Statistical properties, patterns and trends should match the real data as closely as possible, including how several variables relate to each other.
Contextually aware synthetic data
BlueGen generates synthetic data that is contextually aware and representative, so relations between columns carry over along with each column's own distribution. Set several columns against each other for the real and the synthetic data, and you can see how closely the patterns match.
Evidence you can build on
Before a finding from synthetic data goes into your research, you want evidence that the synthetic data is fit for that specific question.
Evaluations built around research questions
The BlueGen evaluation report draws on a library of evaluations for research questions that come up often. Switch on survival analysis, for example, and the report produces the Kaplan-Meier plot and the hazard ratio table directly. If needed, a hypothesis or model developed on the synthetic data can always be applied to the real data afterwards.
500 synthetic ICU patients, near-identical results to the real data
Strict privacy rules kept TU Eindhoven's ICU records inside myDRE, its secure research environment, out of reach for students working on acute kidney injury (AKI) prediction. BlueGen trained inside myDRE, so the real records never left it. Models trained on 500 synthetic or on 500 real patients were tested on the same 800 held-out patients and performed almost identically. For the research, this means students can work on the prediction problem with data that matches the real cohort, contains no identifiable individual and can be published alongside the study.
See what synthetic data can do for your research
- Bring a research question you are working on
- See how BlueGen would shape the data around it