Blind Justice: Fairness with Encrypted Sensitive Attributes

By Niki Kilbertus, Adrià Gascón, Matt J. Kusner, Michael Veale, Krishna P. Gummadi, Adrian Weller. 2018

167 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Safety Research, Privacy & Security, AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety Research

In Privacy & Security

In AI Ethics, Fairness & Society

Citations in the corpus

QUOTIENT: Two-Party Secure Neural Network Training and Prediction (2019, neutral): “This framework has been used in several previous works [14, 29, 41–44].”
Federated Machine Learning (2019, methodology): “Ref [33] uses MPC protocols for model training and verification without users revealing sensitive data.”
Improving Fairness in Machine Learning Systems: What Do Industry Practitioners Need? (2018, background): “ic information in the first place” (R15). Recent work in the fair ML literature has proposed encryption mechanisms that ensure any collected individuallevel demographics can only be used for auditing [64, 105]. But interviewees emphasized that, while such technical solutions are an important prerequ”
What We Can't Measure, We Can't Understand: Challenges to Demographic Data Procurement in the Pursuit of Fairness (2020, methodology): “Other proposed methods build on this third-party arrangement, ensuring cryptographic or differential privacy from the point of collection [43, 48].”
Fairness without Demographics through Adversarially Reweighted Learning (2020, background): “They generally assume that the ML model has access to the protected-features, albeit in encrypted form via secure multi-party computation [50, 34, 25], or in a privacy preserving form by employing differentially private learning [50, 27].”
Algorithms that remember: model inversion attacks and data protection law (2018, background): “isaged as better managed using edge computing, such as the delivery of advertisements or the personalization of news media [17], or discrimination auditing and ‘de-biasing’ of machine learning models [18], limiting the sensitive data leaving devices users directly control. Yet such systems, no longe”
Transparency: Motivations and Challenges (2017, background): “nswer as to whether better outcomes will be achieved by hiding or revealing sensitive attributes. Some have attempted to use technical privacy methods to try simultaneously to get the benets of both [33]. Recent work explores multi-agent reinforcement learning in repeated social interactions to begi”
A Review on Fairness in Machine Learning (2022, methodology): “A few other studies have investigated a distributed setting that includes a “main” party that wishes to train a fair ML model over the data it holds, and a third party to which the sensitive attributes are outsourced [73, 98, 121].”
Awareness in practice: tensions in access to sensitive attribute data for antidiscrimination (2019, background): “Veale and Binns [76] and Kilbertus et al [41] propose approaches to dealing with such information deficits without collecting or revealing sensitive data.”
Transparency: Motivations and Challenges (2019, background): “Some have attempted to use technical privacy methods to try simultaneously to get the benefits of both [33].”
Bias in data‐driven artificial intelligence systems—An introductory survey (2020, background): “9 GDPR special category data such as ethnicity often requires explicit consent (Kilbertus et al., 2018).”