Hands-On Differential Privacy

Chapter 1. Welcome to Differential Privacy

If you are new to the concept of differential privacy, you’ve found the right place. This chapter will present the historical background and conceptual intuition of differential privacy. This chapter is designed to achieve two main objectives:

1. Offer a brief history of data privatization, culminating in contemporary examples of differential privacy deployments

2. Provide a simple practical example of data privatization risks in a classroom setting that demonstrates basic key terms

This chapter is meant to be a soft introduction to differential privacy. You will learn about core DP concepts with less of a focus on math and code. The book will gradually mix in more formal mathematical language and implementations of algorithms as you progress through the chapters.

History

The idea that computing statistics on a data set can leak information about individual data points is not new. In fact, many of the fundamental differential privacy papers^1,2 from the 21st century cite research from the 1970s and ’80s.

In 1977, Tore Dalenius sought to formalize the notion of statistical disclosure control.³ As part of this work, Dalenius contended that the goal should be disclosure control rather than elimination, adding that “elimination of disclosure is possible only by elimination of statistics.” This idea of controlling disclosure will be important throughout the book; the goal is to quantify and calibrate the risk of disclosure, rather than eliminate ...