This 2014 monograph by Cynthia Dwork and Aaron Roth, published in the Foundations and Trends in Theoretical Computer Science series, is the standard reference text for differential privacy. By the time it appeared, the field begun in 2006 had accumulated a body of definitions, mechanisms, and theorems scattered across conference papers; this book pulled them into one coherent, rigorous account that practitioners and researchers still cite as the canonical source.
It lays out the core machinery in order. It formalizes the definition of differential privacy and its relaxations. It presents the Laplace mechanism, which adds noise drawn from the Laplace distribution and is the workhorse for queries measured by their worst-case sensitivity, and the Gaussian mechanism, which adds normally distributed noise and underpins the slightly weaker but very useful approximate form of the guarantee. Crucially, it develops the composition theorems that govern the privacy budget, the parameter usually written as epsilon, that quantifies how much privacy is spent: every query against the data consumes some budget, and the book makes precise how those costs add up across many queries, which is exactly the accounting any real deployment must do.
For a business reader, the significance is less any single result than what the book represents: differential privacy graduating from a research curiosity into engineering. When Apple, Google, or the Census Bureau speak of an epsilon, a privacy budget, or the Gaussian mechanism, this is the shared vocabulary and the shared proofs they are drawing on. Anyone seriously evaluating a privacy claim with a stated epsilon is, in effect, checking it against the framework set down here.