Achieving Differential Privacy and Fairness in Logistic Regression

Abstract

Machine learning algorithms are used to make decisions in various applications. These algorithms rely on large amounts of sensitive individual information to work properly. Hence, there are sociological concerns about machine learning algorithms on matters like privacy and fairness. Currently, many studies focus on only protecting individual privacy or ensuring fairness of algorithms. However, how to meet both privacy and fairness requirements simultaneously in machine learning algorithms is under exploited. In this paper, we focus on one classic machine learning model, logistic regression, and develop differentially private and fair logistic regression models by combining functional mechanism and decision boundary fairness in a joint form. Theoretical analysis and empirical evaluations demonstrate our approaches effectively achieve both differential privacy and fairness while preserving good utility.