The next Comète seminar will take place on Friday, November 22nd at 2pm in Salle Emmy Noether (LIX - Alan Turing building, École Polytechnique). Ashish Dandekar, ENS, will talk about Calibration of noise for a privacy-preserving mechanism.

Abstract: The calibration of noise for a privacy-preserving mechanism depends on the sensitivity of the query and the prescribed privacy level. A data steward must make the non-trivial choice of a privacy level that balances the requirements of users and the monetary constraints of the business entity. We study various sources of randomness that are involved in the design of a privacy-preserving mechanism, namely the explicit randomness induced by the noise distribution and the implicit randomness induced by the data-generation distribution. The study leads us to a probabilistic calibration of privacy-preserving mechanisms with quantifiable privacy guarantees. We instantiate it for the Laplace mechanism by providing analytical results. We propose a cost model that bridges the gap between the privacy level and the compensation budget estimated by a GDPR compliant business entity. We illustrate a realistic scenario wherein the use of fine-tuning of the Laplace mechanism avoids the overestimation of the compensation budget.