Abstract

I will discuss the recent methods we developed in method to audit differential privacy of ML models without changing the pipeline for training and without additional costs.

Video Recording