AVATAR – Anonymization of personal health data by generating digital avatars

The aim of the cluster project AVATAR is to explore new concepts for the privacy-compliant exchange and analysis of personal health data by generating "avatars", i.e. a digital data set representing a patient. Thus, personal data (e.g. from patients) can be in a pseudonymized or anonymized form for analysis and use in biomedical research.

In the AVATAR competence cluster, the conflict between the practical use of data and data protection requirements is to be resolved by means of a new anonymization approach. All collected data, e.g. from studies, medical care, fitness and health applications, are fed into a data pool. On demand, data with the required characteristics are extracted from this pool to generate digital avatars. These avatars serve as artificial persons based on real data. The advantage: it is no longer possible to draw conclusions about the real data donors, and for the first time there is a practical use perspective for the large amount of existing data.

Fraunhofer IDMT designs and develops solutions for technical data protection in the project on the basis of so-called privacy enhancement technologies (PET) such as data anonymization and pseudonymization, differential privacy or secure federated learning for application-specific processing of medical data. This includes, for example, personal MRI recordings, EEG recordings, but also simple personal data such as name, date of birth, etc.

The key challenge here is to address the privacy and utility requirements of the data equally by defining privacy and utility metrics and selecting, adapting, and integrating appropriate PETs. These components should be selectable and customizable during avatar creation, depending on the use case, in order to ensure effective data usage while providing sufficient data protection.

Our motivation

AVATAR offers the opportunity to test the practical relevance of various privacy enhancement technologies in cooperation with application partners and to make them available for a potentially large number of use cases by means of the novel avatar-based approach. We hope that this will provide new impetus for research and further development, as well as expanded application and usage options for the technologies used.

The consortium consists of 18 partners, including three networks, nine research institutions and five companies:

• medways e.V. (consortium coordinator)
• ITnet e.V.
• InfectoGnostics Forschungscampus Jena e.V.
• Jena University hospital (UKJ)
• Fraunhofer IDMT
• DLR Institute for Data Science (DLR DW)
• Fraunhofer IOSB-AST
• Ilmenau University of Technology (TU Ilmenau)
• ISMA AG
• Ernst Abbe University of Applied Sciences Jena (EAH Jena)
• Friedrich Schiller University Jena (FSU Jena)
• SRH Waldklinikum Gera
• rooom AG
• Data in Motion GmbH DIM)
• Navimatix GmbH
• JEN-OPHTHALMO
• Königsee Implants GmbH
• LIEBENSTEIN LAW - Law firm for commercial law, Frankfurt am Main

Federal Ministry of Education and Research (BMBF) within the framework of the call for proposals "Research Network Anonymization for Secure Data Use".

November 2022 – November 2025