Access to data is key for scientific progress. Yet, data is often sensitive in nature and must, therefore, fulfil certain requirements to ensure privacy. For instance, publicly available data should be arranged such that its users cannot trace the data back to any individual in the dataset.

In this course, you will learn about a solid mathematical framework called Differential Privacy (DP), which enables you to quantify the loss in privacy by releasing data or results. By the end of this course, you will be familiar with algorithms that give you control over the privacy loss due to such releases. More concretely, you can report summary statistics, apply Machine Learning techniques, and generate synthetic data (i.e., artificial data that mimic properties of sensitive data) in ways that ensure DP.

All these approaches, in essence, add noise to the data. This noise affects parameter estimates and their uncertainty, and may, therefore, undermine the validity of classical test statistics and resulting inferences. We link this issue to Robust Statistics (RS), and investigate how to correctly perform inference in the context of DP.

Vrije Universiteit Amsterdam (VU Amsterdam) awards credits based on the ECTS system. Contact hours listed under a course description may vary due to the combination of lecture-based and independent work required for each course therefore, CEA's recommended credits are based on the ECTS credits assigned by VU Amsterdam. 1 ECTS equals 28 contact hours assigned by VU Amsterdam.