Md Rafi Ur Rashid

Federated language models vulnerable to privacy attacks, research finds

A new research paper details methods for extracting private user data from federated language models. The study highlights that intermediate model snapshots can be more vulnerable to privacy leakage than the final train…