The Microsoft Research Privacy in AI group explores questions related to Privacy-Preserving Machine Learning and Federated Learning. Some of the topics of interest covering research areas in PPML and FL are:
- Privacy attacks, including membership inference, poisoning attacks, and reconstruction attacks
- Privacy metrics and properties of machine learning pipelines
- Privacy mitigations in machine learning, such as differentially private model training
- Causal Federated Learning
- Federated optimization
- Multi-party data collaboration
- Privacy in Federated Learning
- Adversarial robustness in Federated Learning
- Hierarchical training
- Genetic algorithms for Federated Learning