Blind Justice: Fairness with Encrypted Sensitive Attributes

Niki Kilbertus, Adria Gascon, Matt Kusner, Michael Veale, Krishna Gummadi, Adrian Weller
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:2630-2639, 2018.

Abstract

Recent work has explored how to train machine learning models which do not discriminate against any subgroup of the population as determined by sensitive attributes such as gender or race. To avoid disparate treatment, sensitive attributes should not be considered. On the other hand, in order to avoid disparate impact, sensitive attributes must be examined, e.g., in order to learn a fair model, or to check if a given model is fair. We introduce methods from secure multi-party computation which allow us to avoid both. By encrypting sensitive attributes, we show how an outcome-based fair model may be learned, checked, or have its outputs verified and held to account, without users revealing their sensitive attributes.

Cite this Paper


BibTeX
@InProceedings{pmlr-v80-kilbertus18a, title = {Blind Justice: Fairness with Encrypted Sensitive Attributes}, author = {Kilbertus, Niki and Gascon, Adria and Kusner, Matt and Veale, Michael and Gummadi, Krishna and Weller, Adrian}, booktitle = {Proceedings of the 35th International Conference on Machine Learning}, pages = {2630--2639}, year = {2018}, editor = {Dy, Jennifer and Krause, Andreas}, volume = {80}, series = {Proceedings of Machine Learning Research}, month = {10--15 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v80/kilbertus18a/kilbertus18a.pdf}, url = {https://proceedings.mlr.press/v80/kilbertus18a.html}, abstract = {Recent work has explored how to train machine learning models which do not discriminate against any subgroup of the population as determined by sensitive attributes such as gender or race. To avoid disparate treatment, sensitive attributes should not be considered. On the other hand, in order to avoid disparate impact, sensitive attributes must be examined, e.g., in order to learn a fair model, or to check if a given model is fair. We introduce methods from secure multi-party computation which allow us to avoid both. By encrypting sensitive attributes, we show how an outcome-based fair model may be learned, checked, or have its outputs verified and held to account, without users revealing their sensitive attributes.} }
Endnote
%0 Conference Paper %T Blind Justice: Fairness with Encrypted Sensitive Attributes %A Niki Kilbertus %A Adria Gascon %A Matt Kusner %A Michael Veale %A Krishna Gummadi %A Adrian Weller %B Proceedings of the 35th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jennifer Dy %E Andreas Krause %F pmlr-v80-kilbertus18a %I PMLR %P 2630--2639 %U https://proceedings.mlr.press/v80/kilbertus18a.html %V 80 %X Recent work has explored how to train machine learning models which do not discriminate against any subgroup of the population as determined by sensitive attributes such as gender or race. To avoid disparate treatment, sensitive attributes should not be considered. On the other hand, in order to avoid disparate impact, sensitive attributes must be examined, e.g., in order to learn a fair model, or to check if a given model is fair. We introduce methods from secure multi-party computation which allow us to avoid both. By encrypting sensitive attributes, we show how an outcome-based fair model may be learned, checked, or have its outputs verified and held to account, without users revealing their sensitive attributes.
APA
Kilbertus, N., Gascon, A., Kusner, M., Veale, M., Gummadi, K. & Weller, A.. (2018). Blind Justice: Fairness with Encrypted Sensitive Attributes. Proceedings of the 35th International Conference on Machine Learning, in Proceedings of Machine Learning Research 80:2630-2639 Available from https://proceedings.mlr.press/v80/kilbertus18a.html.

Related Material