On October 1st the Institute started an FFG funded research project which aims to systematically validate the feasibility of using deep recurrent neural network architectures to generate synthetic sequential raw data that preserve individual privacy and, at the same time, retain enough information to be used for market research. In this project we collaborate with Mostly AI Solutions MP GmbH, George Labs GmbH and Statistik Austria.

Generative deep neural networks have recently become a highly active research field of artificial intelligence, with impressive demonstrations for synthetic image generation. The combination and application of these developed methods to sequential personal data could provide a viable solution to the utilization problem of growing amount of available personal data, while safeguarding the privacy of individuals. Together with our consortium partners, we are going to design and set up a virtual data lab that will allow us to systematically investigate the conditions under which a variety of generative models are able to derive synthetic replicas that capture structure and correlations, while protecting individual-level privacy.

More information on the project will be shared via the project website: www.anonymousbigdata.net