Movatterモバイル変換


[0]ホーム

URL:


Oblivious Sampling Algorithms for Private Data Analysis

Part ofAdvances in Neural Information Processing Systems 32 (NeurIPS 2019)

AuthorFeedbackBibtexMetaReviewMetadataPaperReviewsSupplemental

Authors

Sajin Sasy, Olga Ohrimenko

Abstract

We study secure and privacy-preserving data analysisbased on queries executed on samples from a dataset.Trusted execution environments (TEEs) can be used toprotect the content of the data during query computation,while supporting differential-private (DP) queries in TEEsprovides record privacy when query output is revealed.Support for sample-based queries is attractivedue to \emph{privacy amplification}since not all dataset is used to answer a query but only a small subset.However, extracting data samples with TEEswhile proving strong DP guarantees is nottrivial as secrecy of sample indices has to be preserved.To this end, we design efficient secure variants of common sampling algorithms.Experimentally we show that accuracy of modelstrained with shuffling and sampling is the same fordifferentially private models for MNIST and CIFAR-10,while sampling provides stronger privacy guarantees than shuffling.


Name Change Policy

Requests for name changes in the electronic proceedings will be accepted with no questions asked. However name changes may cause bibliographic tracking issues. Authors are asked to consider this carefully and discuss it with their co-authors prior to requesting a name change in the electronic proceedings.

Use the "Report an Issue" link to request a name change.


[8]ページ先頭

©2009-2025 Movatter.jp