Developers are under constant pressure to focus efforts on those features that would bring the greatest benefits to their users. It would help to collect metrics that measure how different parts of the software are used, but unfortunately many users are reluctant to supply such data.

Differential Privacy encompasses a variety of techniques where a controlled amount of noise is added to data to allow the reporting of aggregated information without the ability to identify individuals. It remains an active area of research.

This talk provides a brief overview of the development of the subject over the last 20 years. It explains the key terminology and describes the main techniques available while providing a taster of some of the mathematics involved.

Finally it proposes ways these techniques might be used to address telemetry in Fedora.