Google Research

Locally Private k-Means in One Round

International Conference on Machine Learning (ICML) (2021), pp. 1441-1451

Abstract

We study k-means clustering in the non-interactive (aka one-round) local model of differential privacy. We give an approximation algorithm that requires a single round of communication and achieves an approximation ratio arbitrarily close to the best non private approximation algorithm. To show the flexibility of our framework, we also demonstrate that it yields a similar near-optimal approximation algorithm in the (one-round) shuffle model.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work