As the unofficial American summer has chimed in, and we start heading to the beach, it is time for some light summer reading. Perhaps this light fare is what you need in contrast to what you may be reading about large-scale database analysis in the press or elsewhere. Sifting through large databases, whether they are government, commercial, or private, may make us think in terms of privacy: what pieces of data can be omitted while still retaining some usefulness as a whole? You may have heard of k-anonymity, but that is not the target here. We are talking about differential privacy, originally conceived by Dwork et al. in the 2000s.

Simson Garfinkel brings us the perfect 222+-page "beach book," or perhaps "general audience book" depending on your vacationing preferences, to delve a bit deeper into the concepts of "Differential Privacy." The author takes an experiential approach of describing differential privacy via his time at places of employment where those ideas could be or have been applied. In three chapters (plus the introductory chapter), with many (sometimes humorous) black-and-white illustrations, cartoons, tables, and diagrams, extensive references, he brings the topic to the reader while still referring to the solid mathematical foundations. While the book is available for purchase, you can simply download a properly sanctioned (think: open access) PDF from the MIT Press book website. Ironically, that inserts a download URL into the PDF with the date it was downloaded, plus the MIT Press username. Try it for yourself!

In the Preface, the author sets up the motivation for the book, how he met some of the key figures from the field, and how he got more interested in the topic. He also sets expectations, and immediately points the reader to appropriate resources beyond this light reading, depending on their background depth (policymakers, mathematicians, computer scientists, statisticians, programmers, etc).

In the Introduction, the reader learns about the conceptualization of the differential privacy idea, going through various namings, such as incremental privacy, and the evolution of the various mathematical ideas connected to it. In connection to learning about Dwork's journey on this concept, we also hear about the author's interaction with the privacy field on that subject matter in a semi-autobiographical approach. Moreover, the importance of applying this fundamental idea to the 2020 Census is highlighted by a series of real-world examples that explain how microdata can be used to reconstruct otherwise anonymized statistical database data and more. It also contains Daniel Solove's taxonomy on privacy terminology.

In Chapter 1, the focus is on "Concepts and Theories," where the author sprinkles in a bit more math. The basic concept of Differential Privacy is introduced as a "mathematical framework based on a definition of privacy loss that has formal guarantees." Adding noise is a key idea as a trade-off between privacy and accuracy. Here the reader learns about composability as something differential privacy can do, but k-anonymity cannot. The rest of the chapter brings in many real-world examples, including the author's interaction with the Census Bureau to satisfy its privacy needs and its application of differential privacy principles for the 2020 Census.

Chapter 2 covers "Differential Privacy Issues," seen as the teething problems of a technology that is only about 20 years young. Issues such as privacy vs. accuracy trade-offs, semantic interpretations, and issues with practical applications such as an ambulance sent to the wrong house due to inaccuracies. Other concerns are expressions of proper privacy policies, and other application domain problems.

Lastly, in Chapter 3 the author discusses "Future Directions," extending possible directions from the state of the art of differential privacy. The topics explored here include combining differential privacy with other security and privacy concepts, such as Trusted Execution Environments, Secure Multiparty Computation, and Homomorphic Encryption, among others.

This book is aimed at a broader audience, but it does provide technical details for those wishing to explore further. Think of it as Cliff Notes of Differential Privacy to get you bootstrapped. The 13-page long references at the end of the book, plus the 'Further Reading' section, will lead the reader to more technical knowledge, should their interest be piqued.

I really enjoyed reading this book. It is book is light and entertaining reading and provides anecdotal background information for how we can deal with (differential) privacy in a more critical and pragmatic way. Read it for yourself. On the beach, on your commute, or elsewhere.

Sven Dietrich reviews technology and security books for IEEE Cipher. He welcomes your thoughts at spock at ieee dot org