We have an invited guest for tomorrow’s reading group, Georgi Ganev, presenting his paper: Robin Hood and Matthew Effects: Differential Privacy Has Disparate Impact on Synthetic Data
Short description: Synthetic data is a promising privacy enhancing technology (PET) which has the potential to serve in a variety of use cases such as private data release, data de-biasing, data augmentation. In this talk, we will focus on synthetic data produced by generative machine learning algorithms, why privacy is not automatic, briefly introduce DP, and what effects or distortions it brings to the (synthetic) table.
Short speaker bio: Georgi is a PhD student at UCL studying the intersection between (generative) machine learning and privacy. He also leads the research team at Hazy, a synthetic data company, and volunteers as a research scientist at OpenMinded, an open source community building privacy preserving tools.