Privacy Preserving Data Analysis is becoming ever more important as society would like to benefit from the data available, while also not infringing in people’s privacy. DataSHIELD comprises a set of tools, which allows researchers to analyse data, without giving the researcher access to individual level data. In this manner it would allow a researcher to explore aggregate statistics of data, such as the mean age of a population, but not disclose any information about the ages of the individual patients.
As part of the MIRACUM effort to create a privacy preserving hospital analysis network, MIRACUM researchers took part in the DataSHIELD Workshop 2018 and presented their contribution to the DataSHIELD community. Julian Gruendner (Erlangen) presented “A queue-Poll Extension for standardised, monitored, indirect and secure DataSHIELD access to your data” and Stefan Lenz (Freiburg) presented “Deep Learning with Boltzmann Machines” adjusted to work with DataSHIELD to ensure privacy preserving data analysis. Daniela Zöller (Freiburg) presented a “Distributed regression modelling for selecting markers in DataSHIELD”, which allows for privacy preserving selection of the most important variables in a given dataset. The workshop was a large success and the participants gained a good insight into the issues of privacy preserving analysis, the newest statistical methods in this area, as well as the infrastructure necessary for a data analysis network. The relationships fostered during this workshop will benefit the MIRACUM project in the future and first collaborations have already been established between the MIRACUM consortium and other member of the DataSHIELD community, which will help to streamline future development of the necessary DataSHIELD infrastructure.