Empirical Privacy Metrics in Synthetic Data Generation - The Bad, the Ugly, and the Good

Explore the complexities of synthetic data generation and empirical privacy metrics in this 14-minute conference talk from PEPR '24. Delve into the world of fake data creation that mimics real statistical distributions, examining the privacy claims made by synthetic data vendors. Critically analyze the effectiveness and reliability of empirical privacy metrics used to determine data safety. Uncover marketing myths surrounding synthetic data and privacy claims, and gain insights into the necessary steps for integrating these tools into a robust privacy strategy. Learn about the potential benefits and pitfalls of synthetic data in data analysis, sharing, and commercialization while considering privacy compliance and anonymity concerns.