Continual release moment estimation with differential privacy
Kalinin N, Upadhyay J, Lampert C. 2025. Continual release moment estimation with differential privacy. Advances in Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 38, 64002–64047.
Download (ext.)
Conference Paper
| Published
| English
Scopus indexed
Author
Corresponding author has ISTA affiliation
Department
Series Title
Advances in Neural Information Processing Systems
Abstract
We propose Joint Moment Estimation (JME), a method for continually and privately estimating both the first and second moments of a data stream with reduced noise compared to naive approaches. JME supports the matrix mechanism and exploits a joint sensitivity analysis to identify a privacy regime in which the second-moment estimation incurs no additional privacy cost, thereby improving accuracy while maintaining privacy. We demonstrate JME’s effectiveness in two applications: estimating the running mean and covariance matrix for Gaussian density estimation and model training with DP-Adam.
Publishing Year
Date Published
2025-12-02
Proceedings Title
Advances in Neural Information Processing Systems
Publisher
Neural Information Processing Systems Foundation
Acknowledgement
We thank Monika Henzinger for her valuable feedback and insightful discussions on earlier versions
of this draft. We are also grateful to Mher Safaryan for his contributions to discussions on DP-Adam.
Additionally, we thank Ryan McKenna for suggesting Joint Clipping as a baseline.
Jalaj Upadhyay’s research was funded by the Rutgers Decanal Grant no. 302918, NSF CNS 2433628,
Google Research Scholar Award, and Google Seed Fund Grant. A part of this work was done while
visiting the Institute of Science and Technology Austria (ISTA).
Nikita Kalinin’s research was funded in part by the Austrian Science Fund (FWF) [10.55776/COE12]
Volume
38
Page
64002-64047
Conference
NeurIPS: Neural Information Processing Systems
Conference Location
San Diego, CA, United States
ISBN
ISSN
IST-REx-ID
Cite this
Kalinin N, Upadhyay J, Lampert C. Continual release moment estimation with differential privacy. In: Advances in Neural Information Processing Systems. Vol 38. Neural Information Processing Systems Foundation; 2025:64002-64047. doi:10.52202/085713-1924
Kalinin, N., Upadhyay, J., & Lampert, C. (2025). Continual release moment estimation with differential privacy. In Advances in Neural Information Processing Systems (Vol. 38, pp. 64002–64047). San Diego, CA, United States: Neural Information Processing Systems Foundation. https://doi.org/10.52202/085713-1924
Kalinin, Nikita, Jalaj Upadhyay, and Christoph Lampert. “Continual Release Moment Estimation with Differential Privacy.” In Advances in Neural Information Processing Systems, 38:64002–47. Neural Information Processing Systems Foundation, 2025. https://doi.org/10.52202/085713-1924.
N. Kalinin, J. Upadhyay, and C. Lampert, “Continual release moment estimation with differential privacy,” in Advances in Neural Information Processing Systems, San Diego, CA, United States, 2025, vol. 38, pp. 64002–64047.
Kalinin N, Upadhyay J, Lampert C. 2025. Continual release moment estimation with differential privacy. Advances in Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 38, 64002–64047.
Kalinin, Nikita, et al. “Continual Release Moment Estimation with Differential Privacy.” Advances in Neural Information Processing Systems, vol. 38, Neural Information Processing Systems Foundation, 2025, pp. 64002–47, doi:10.52202/085713-1924.
All files available under the following license(s):
Copyright Statement:
This Item is protected by copyright and/or related rights. [...]
Link(s) to Main File(s)
Access Level
Open Access
Google Scholar
ISBN Search