Saeed Mahloujifar

DP-ZO balances privacy and performance in language model training.

New mechanisms enhance privacy while preserving data utility in machine learning.

Examining feature inversion in deep learning and its implications for privacy.