Differential privacy is a property of a randomized algorithm that computes aggregate information about a large data set without revealing detailed information about the individual data. Specifically, an algorithm is $$\varepsilon$$-differentially private if the probability of any particular outcome of the algorithm is changed by a factor of at most $$e^{\varepsilon}$$ by including any additional data point. Thus there is a limit to how much information can leak through the algorithm about a user when that user adds their data.