Blind Justice: Fairness with Encrypted Sensitive Attributes

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.

arXiv commentpublished at ICML 2018

References in corpus (3)

• Equality of Opportunity in Supervised Learning
• On the (im)possibility of fairness
• From Parity to Preference-based Notions of Fairness in Classification

Cited by in corpus (8)

• Detecting Shortcut Learning for Fair Medical AI using Shortcut Testing
• Fair Learning with Private Demographic Data
• "What We Can't Measure, We Can't Understand": Challenges to Demographic Data Procurement in the Pursuit of Fairness
• Privacy-preserving Federated Bayesian Learning of a Generative Model for Imbalanced Classification of Clinical Data
• A Sequentially Fair Mechanism for Multiple Sensitive Attributes
• Measuring Fairness Under Unawareness of Sensitive Attributes: A Quantification-Based Approach
• Accelerating Federated Learning in Heterogeneous Data and Computational Environments
• Near-Optimal Procedures for Model Discrimination with Non-Disclosure Properties