Differential privacy refers to a set of techniques aiming to modify the data points in a dataset in a way that makes it difficult to detect the presence of any particular data point, while preserving certain global statistical properties of the dataset. The goal is to ensure individual privacy while keeping the data useful in the aggregate.