Privacy-preserving analysis and learning in secure and distributed enclaves and exascale systems funded by DOE ASCR

Overall Objectives


  • Develop Argonne PPFL (APPFL) framework that implements differentially private (DP) algorithms for training federated learning (FL) models with the biomedical datasets from multiple organizations and

  • Integrate, deploy, and demonstrate the proposed framework with our existing secure computing and data infrastructure.

To accomplish our objectives, we will create a privacy-preserving AI/ML architecture, which will allow us to validate APPFL framework with real-world, multi-modal biomedical data repositories that align with the NIH Bridge2AI pilot flagship data generation projects.