An open-source infrastructure for privacy preserving analysis
Answering many of the questions in health care often requires incorporating data that are located at different sources. Typically, data from different parties are analyzed by centralizing them. That is, data are brought to where the algorithms are. Unfortunately, this raises heavy concerns regarding the privacy and security of sensitive patient data.
Federated learning has emerged as a technological solution to address these concerns. In this novel paradigm, algorithms are brought to where the data are. This way, we can maximize the potential of multiple datasets while minimizing data leaks and privacy risks.
This led us to create vantage6, our open-source platform for supporting the development of federated learning projects.
vantage6 is built on top of three pillars
...using three basic principles
Each party is in control of their own data.
Parties are allowed to have differences in hardware and operating systems.
Support for horizontally- and vertically-partitioned data.