Noise-Aware Differentially Private Variational Inference

Talal Alrawajfeh, Joonas Jälkö, Antti Honkela
Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, PMLR 258:4987-4995, 2025.

Abstract

Differential privacy (DP) provides robust privacy guarantees for statistical inference, but this can lead to unreliable results and biases in downstream applications. While several noise-aware approaches have been proposed which integrate DP perturbation into the inference, they are limited to specific types of simple probabilistic models. In this work, we propose a novel method for noise-aware approximate Bayesian inference based on stochastic gradient variational inference which can also be applied to high-dimensional and non-conjugate models. We also propose a more accurate evaluation method for noise-aware posteriors. Empirically, our inference method has similar performance to existing methods in the domain where they are applicable. Outside this domain, we obtain accurate coverages on high-dimensional Bayesian linear regression and well-calibrated predictive probabilities on Bayesian logistic regression with the UCI Adult dataset.

Cite this Paper

BibTeX
@InProceedings{pmlr-v258-alrawajfeh25a, title = {Noise-Aware Differentially Private Variational Inference}, author = {Alrawajfeh, Talal and J{\"a}lk{\"o}, Joonas and Honkela, Antti}, booktitle = {Proceedings of The 28th International Conference on Artificial Intelligence and Statistics}, pages = {4987--4995}, year = {2025}, editor = {Li, Yingzhen and Mandt, Stephan and Agrawal, Shipra and Khan, Emtiyaz}, volume = {258}, series = {Proceedings of Machine Learning Research}, month = {03--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v258/main/assets/alrawajfeh25a/alrawajfeh25a.pdf}, url = {https://proceedings.mlr.press/v258/alrawajfeh25a.html}, abstract = {Differential privacy (DP) provides robust privacy guarantees for statistical inference, but this can lead to unreliable results and biases in downstream applications. While several noise-aware approaches have been proposed which integrate DP perturbation into the inference, they are limited to specific types of simple probabilistic models. In this work, we propose a novel method for noise-aware approximate Bayesian inference based on stochastic gradient variational inference which can also be applied to high-dimensional and non-conjugate models. We also propose a more accurate evaluation method for noise-aware posteriors. Empirically, our inference method has similar performance to existing methods in the domain where they are applicable. Outside this domain, we obtain accurate coverages on high-dimensional Bayesian linear regression and well-calibrated predictive probabilities on Bayesian logistic regression with the UCI Adult dataset.} }
Endnote
%0 Conference Paper %T Noise-Aware Differentially Private Variational Inference %A Talal Alrawajfeh %A Joonas Jälkö %A Antti Honkela %B Proceedings of The 28th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2025 %E Yingzhen Li %E Stephan Mandt %E Shipra Agrawal %E Emtiyaz Khan %F pmlr-v258-alrawajfeh25a %I PMLR %P 4987--4995 %U https://proceedings.mlr.press/v258/alrawajfeh25a.html %V 258 %X Differential privacy (DP) provides robust privacy guarantees for statistical inference, but this can lead to unreliable results and biases in downstream applications. While several noise-aware approaches have been proposed which integrate DP perturbation into the inference, they are limited to specific types of simple probabilistic models. In this work, we propose a novel method for noise-aware approximate Bayesian inference based on stochastic gradient variational inference which can also be applied to high-dimensional and non-conjugate models. We also propose a more accurate evaluation method for noise-aware posteriors. Empirically, our inference method has similar performance to existing methods in the domain where they are applicable. Outside this domain, we obtain accurate coverages on high-dimensional Bayesian linear regression and well-calibrated predictive probabilities on Bayesian logistic regression with the UCI Adult dataset.
APA
Alrawajfeh, T., Jälkö, J. & Honkela, A.. (2025). Noise-Aware Differentially Private Variational Inference. Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 258:4987-4995 Available from https://proceedings.mlr.press/v258/alrawajfeh25a.html.

Related Material