Ektelo is a programming framework and system that aids programmers in developing differentially private programs with high utility. Ektelo can be used to author programs for a variety of statistical tasks that involve answering counting queries over a table of arbitrary dimension.

In Ektelo, a differentially private program is described as a plan over a high level library of operators. Operators are organized into classes based on their functionality, which includes data transformations, data reductions, query selection and execution, and inference methods to combine noisy answers into a consistent estimate.

The operator-based approach has following benefits:

• Privacy for free. Any plan written in Ektelo will satisfy differential privacy. The privacy of each operator has been vetted and the system ensures that any composition of operators satisfies differential privacy.
• Modularity. Ektelo enables code reuse as the same operator can be used in multiple programs. This helps safety as the number of privacy-sensitive components is kept small.
• Transparency. By expressing algorithms as plans with operators from operator classes, it is easier to compare/contrast competing algorithms.
• Flexibility. Users can take existing operators and recombine them in arbitrary ways – allowing them to invent new algorithms that borrow ideas from the state-of-art – without the need for privacy analysis.

Supporting publications

Dan Zhang, Ryan McKenna, Ios Kotsogiannis, Michael Hay, Ashwin Machanavajjhala, and Gerome Miklau. 2018. EKTELO: A Framework for Defining Differentially-Private Computations. In Proceedings of the 2018 International Conference on Management of Data (SIGMOD ‘18). ACM, New York, NY, USA, 115-130. DOI: https://doi.org/10.1145/3183713.3196921

Contributors

• Dan Zhang, UMass Amherst
• Ryan McKenna, UMass Amherst
• Ios Kotsogiannis, Duke University
• George Bissias, UMass Amherst
• Michael Hay, Colgate University
• Ashwin Machanavajjhala, Duke University
• Gerome Miklau, UMass Amherst

Acknowledgements

This project is supported by the NSF, DARPA and Center for Data Science at the University of Massachusetts College of Information and Computer Sciences.