The author investigates the risks associated with machine learning model leaks and explores how synthetic data can impact the inference of sensitive attributes by attackers.
