2 Differential privacy for machine learning (part-1)
This chapter covers:
- Introducing the concept and definition of differential privacy
- Formulating widely adopted differential privacy mechanisms in use today, that have served as the most important building blocks in various privacy-preserving algorithms and applications
- Implementing various properties of differential privacy under different scenarios (e.g., post processing property, group property and composition properties)
We have investigated various privacy-related threats and vulnerabilities in machine learning, and numerous concepts of privacy-enhancing technologies in the previous chapter. In this book, we will learn the details about several important and popular privacy enhancing technologies. The very first one we are going to introduce here is the differential privacy. Differential privacy is one of the most popular privacy protection schemes in use today. It introduces a very interesting concept on how to make a dataset robust enough to changes of any single sample in the dataset, while computing the data statistics (e.g., machine learning models could be considered as certain very complex statistics to describe the distribution of its training data). In this chapter, we will walk through our 1st part of differential privacy for machine learning.
