Background

Federated learning is a subfield of machine learning that enables multiple, independent entities to collectively train a single model without sharing their data. This can be useful for efficiency reasons, but also is often desirable for privacy. To improve the privacy of these systems, end-to-end encryption can be used to share the raw data being trained on or intermediate results without revealing additional information.

Research Goals

The scholar team will develop and evaluate federated learning frameworks incorporating techniques such as differential privacy, secure aggregation, and encrypted communication to protect sensitive information during training.

Skills Needed

Strong C++ programming skills; comfortability with formal mathematics; basic understanding of cryptographic protocols and machine learning

Skills Gained

Applied cyrptography; building secure applications; understanding of federated machine learning