We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Cryptography and Security

Title: Plume: Differential Privacy at Scale

Abstract: Differential privacy has become the standard for private data analysis, and an extensive literature now offers differentially private solutions to a wide variety of problems. However, translating these solutions into practical systems often requires confronting details that the literature ignores or abstracts away: users may contribute multiple records, the domain of possible records may be unknown, and the eventual system must scale to large volumes of data. Failure to carefully account for all three issues can severely impair a system's quality and usability.
We present Plume, a system built to address these problems. We describe a number of sometimes subtle implementation issues and offer practical solutions that, together, make an industrial-scale system for differentially private data analysis possible. Plume is currently deployed at Google and is routinely used to process datasets with trillions of records.
Subjects: Cryptography and Security (cs.CR); Distributed, Parallel, and Cluster Computing (cs.DC)
Cite as: arXiv:2201.11603 [cs.CR]
  (or arXiv:2201.11603v1 [cs.CR] for this version)

Submission history

From: Sergei Vassilvitskii [view email]
[v1] Thu, 27 Jan 2022 16:02:26 GMT (892kb,D)

Link back to: arXiv, form interface, contact.