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

Download:

Current browse context:

cs.LG

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 > Machine Learning

Title: Differentially Private Normalizing Flows for Privacy-Preserving Density Estimation

Abstract: Normalizing flow models have risen as a popular solution to the problem of density estimation, enabling high-quality synthetic data generation as well as exact probability density evaluation. However, in contexts where individuals are directly associated with the training data, releasing such a model raises privacy concerns. In this work, we propose the use of normalizing flow models that provide explicit differential privacy guarantees as a novel approach to the problem of privacy-preserving density estimation. We evaluate the efficacy of our approach empirically using benchmark datasets, and we demonstrate that our method substantially outperforms previous state-of-the-art approaches. We additionally show how our algorithm can be applied to the task of differentially private anomaly detection.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Data Structures and Algorithms (cs.DS); Machine Learning (stat.ML)
Cite as: arXiv:2103.14068 [cs.LG]
  (or arXiv:2103.14068v1 [cs.LG] for this version)

Submission history

From: Rachel Cummings [view email]
[v1] Thu, 25 Mar 2021 18:39:51 GMT (6977kb,D)

Link back to: arXiv, form interface, contact.