What is differential privacy?

Differential privacy is a mathematical model used to protect individual privacy when analysing large datasets. It adds a small amount of random noise to the data, ensuring that insights can be drawn about the overall population without revealing sensitive information about specific individuals in the dataset. A key aspect of differential privacy is the concept of “epsilon” (ɛ), which defines the amount of noise added to the data. Epsilon is often referred to as the “privacy budget” or “privacy parameter.”

Differential privacy aims to protect individual privacy by ensuring that sensitive information about any one person cannot be easily inferred. A notable practical application is its use by the US Census Bureau in 2020 to protect the sensitive demographic data of its citizens.