FriendlyCore: Practical Differentially Private Aggregation

Eliad Tsfadia, Edith Cohen, Haim Kaplan, Yishay Mansour, Uri Stemmer
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:21828-21863, 2022.

Abstract

Differentially private algorithms for common metric aggregation tasks, such as clustering or averaging, often have limited practicality due to their complexity or to the large number of data points that is required for accurate results. We propose a simple and practical tool FriendlyCore that takes a set of points D from an unrestricted (pseudo) metric space as input. When D has effective diameter r, FriendlyCore returns a “stable” subset C⊆D that includes all points, except possibly few outliers, and is guaranteed to have diameter r. FriendlyCore can be used to preprocess the input before privately aggregating it, potentially simplifying the aggregation or boosting its accuracy. Surprisingly, FriendlyCore is light-weight with no dependence on the dimension. We empirically demonstrate its advantages in boosting the accuracy of mean estimation and clustering tasks such as k-means and k-GMM, outperforming tailored methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-tsfadia22a, title = {{F}riendly{C}ore: Practical Differentially Private Aggregation}, author = {Tsfadia, Eliad and Cohen, Edith and Kaplan, Haim and Mansour, Yishay and Stemmer, Uri}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {21828--21863}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/tsfadia22a/tsfadia22a.pdf}, url = {https://proceedings.mlr.press/v162/tsfadia22a.html}, abstract = {Differentially private algorithms for common metric aggregation tasks, such as clustering or averaging, often have limited practicality due to their complexity or to the large number of data points that is required for accurate results. We propose a simple and practical tool $\mathsf{FriendlyCore}$ that takes a set of points ${\cal D}$ from an unrestricted (pseudo) metric space as input. When ${\cal D}$ has effective diameter $r$, $\mathsf{FriendlyCore}$ returns a “stable” subset ${\cal C} \subseteq {\cal D}$ that includes all points, except possibly few outliers, and is guaranteed to have diameter $r$. $\mathsf{FriendlyCore}$ can be used to preprocess the input before privately aggregating it, potentially simplifying the aggregation or boosting its accuracy. Surprisingly, $\mathsf{FriendlyCore}$ is light-weight with no dependence on the dimension. We empirically demonstrate its advantages in boosting the accuracy of mean estimation and clustering tasks such as $k$-means and $k$-GMM, outperforming tailored methods.} }
Endnote
%0 Conference Paper %T FriendlyCore: Practical Differentially Private Aggregation %A Eliad Tsfadia %A Edith Cohen %A Haim Kaplan %A Yishay Mansour %A Uri Stemmer %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-tsfadia22a %I PMLR %P 21828--21863 %U https://proceedings.mlr.press/v162/tsfadia22a.html %V 162 %X Differentially private algorithms for common metric aggregation tasks, such as clustering or averaging, often have limited practicality due to their complexity or to the large number of data points that is required for accurate results. We propose a simple and practical tool $\mathsf{FriendlyCore}$ that takes a set of points ${\cal D}$ from an unrestricted (pseudo) metric space as input. When ${\cal D}$ has effective diameter $r$, $\mathsf{FriendlyCore}$ returns a “stable” subset ${\cal C} \subseteq {\cal D}$ that includes all points, except possibly few outliers, and is guaranteed to have diameter $r$. $\mathsf{FriendlyCore}$ can be used to preprocess the input before privately aggregating it, potentially simplifying the aggregation or boosting its accuracy. Surprisingly, $\mathsf{FriendlyCore}$ is light-weight with no dependence on the dimension. We empirically demonstrate its advantages in boosting the accuracy of mean estimation and clustering tasks such as $k$-means and $k$-GMM, outperforming tailored methods.
APA
Tsfadia, E., Cohen, E., Kaplan, H., Mansour, Y. & Stemmer, U.. (2022). FriendlyCore: Practical Differentially Private Aggregation. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:21828-21863 Available from https://proceedings.mlr.press/v162/tsfadia22a.html.

Related Material