Auditing $f$-differential privacy in one run

Saeed Mahloujifar, Luca Melis, Kamalika Chaudhuri
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:42615-42641, 2025.

Abstract

Empirical auditing has emerged as a means of catching some of the flaws in the implementation of privacy-preserving algorithms. Existing auditing mechanisms, however, are either computationally inefficient – requiring multiple runs of the machine learning algorithms or suboptimal in calculating an empirical privacy. In this work, we present a tight and efficient auditing procedure and analysis that can effectively assess the privacy of mechanisms. Our approach is efficient; Similar to the recent work of Steinke, Nasr and Jagielski (2023), our auditing procedure leverages the randomness of examples in the input dataset and requires only a single run of the target mechanism. And it is more accurate; we provide a novel analysis that enables us to achieve tight empirical privacy estimates by using the hypothesized $f$-DP curve of the mechanism, which provides a more accurate measure of privacy than the traditional $\epsilon,\delta$ differential privacy parameters. We use our auditing procure and analysis to obtain empirical privacy, demonstrating that our auditing procedure delivers tighter privacy estimates.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-mahloujifar25a, title = {Auditing $f$-differential privacy in one run}, author = {Mahloujifar, Saeed and Melis, Luca and Chaudhuri, Kamalika}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {42615--42641}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/mahloujifar25a/mahloujifar25a.pdf}, url = {https://proceedings.mlr.press/v267/mahloujifar25a.html}, abstract = {Empirical auditing has emerged as a means of catching some of the flaws in the implementation of privacy-preserving algorithms. Existing auditing mechanisms, however, are either computationally inefficient – requiring multiple runs of the machine learning algorithms or suboptimal in calculating an empirical privacy. In this work, we present a tight and efficient auditing procedure and analysis that can effectively assess the privacy of mechanisms. Our approach is efficient; Similar to the recent work of Steinke, Nasr and Jagielski (2023), our auditing procedure leverages the randomness of examples in the input dataset and requires only a single run of the target mechanism. And it is more accurate; we provide a novel analysis that enables us to achieve tight empirical privacy estimates by using the hypothesized $f$-DP curve of the mechanism, which provides a more accurate measure of privacy than the traditional $\epsilon,\delta$ differential privacy parameters. We use our auditing procure and analysis to obtain empirical privacy, demonstrating that our auditing procedure delivers tighter privacy estimates.} }
Endnote
%0 Conference Paper %T Auditing $f$-differential privacy in one run %A Saeed Mahloujifar %A Luca Melis %A Kamalika Chaudhuri %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-mahloujifar25a %I PMLR %P 42615--42641 %U https://proceedings.mlr.press/v267/mahloujifar25a.html %V 267 %X Empirical auditing has emerged as a means of catching some of the flaws in the implementation of privacy-preserving algorithms. Existing auditing mechanisms, however, are either computationally inefficient – requiring multiple runs of the machine learning algorithms or suboptimal in calculating an empirical privacy. In this work, we present a tight and efficient auditing procedure and analysis that can effectively assess the privacy of mechanisms. Our approach is efficient; Similar to the recent work of Steinke, Nasr and Jagielski (2023), our auditing procedure leverages the randomness of examples in the input dataset and requires only a single run of the target mechanism. And it is more accurate; we provide a novel analysis that enables us to achieve tight empirical privacy estimates by using the hypothesized $f$-DP curve of the mechanism, which provides a more accurate measure of privacy than the traditional $\epsilon,\delta$ differential privacy parameters. We use our auditing procure and analysis to obtain empirical privacy, demonstrating that our auditing procedure delivers tighter privacy estimates.
APA
Mahloujifar, S., Melis, L. & Chaudhuri, K.. (2025). Auditing $f$-differential privacy in one run. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:42615-42641 Available from https://proceedings.mlr.press/v267/mahloujifar25a.html.

Related Material