Abstract:As one of the most popular statistical and machine learning models, logistic regression with regularization has found wide adoption in biomedicine, social sciences, information technology, and so on. These domains often involve data of human subjects that are contingent upon strict privacy regulations. Concerns over data privacy make it increasingly difficult to coordinate and conduct large-scale collaborative studies, which typically rely on cross-institution data sharing and joint analysis. Our work here focuses on safeguarding regularized logistic regression, a widely-used statistical model while at the same time has not been investigated from a data security and privacy perspective. We consider a common use scenario of multi-institution collaborative studies, such as in the form of research consortia or networks as widely seen in genetics, epidemiology, social sciences, etc. To make our privacy-enhancing solution practical, we demonstrate a non-conventional and computationally efficient method leveraging distributing computing and strong cryptography to provide comprehensive protection over individual-level and summary data. Extensive empirical evaluations on several studies validate the privacy guarantee, efficiency and scalability of our proposal. We also discuss the practical implications of our solution for large-scale studies and applications from various disciplines, including genetic and biomedical studies, smart grid, network analysis, etc.

Minimax Bounds for Distributed Logistic Regression

Lower Bounds for Learning Distributions under Communication Constraints via Fisher Information

Distributed Ordinal Regression Over Networks

A Minimax Lower Bound for Low-Rank Matrix-Variate Logistic Regression

Distributed Coordinate Descent for L1-regularized Logistic Regression

Distributed Logistic Regression for Massive Data with Rare Events

Minimax rates for latent position estimation in the generalized random dot product graph

A Few Interactions Improve Distributed Nonparametric Estimation, Optimally

Supporting Regularized Logistic Regression Privately and Efficiently

Distributed Nonparametric Regression under Communication Constraints

Beyond Minimax Rates in Group Distributionally Robust Optimization via a Novel Notion of Sparsity

High-Dimensional Graphical Model Selection Using $\ell_1$-Regularized Logistic Regression

Oblivious sketching for logistic regression

Distributionally Robust Logistic Regression

Minimax Optimal Algorithms with Fixed-$k$-Nearest Neighbors

Minimax rates of l p-losses for high-dimensional linear regression models with additive measurement errors over l q-balls

Dimension-free uniform concentration bound for logistic regression

On the Fundamental Limit of Distributed Learning with Interchangable Constrained Statistics

Logistic Regression: Tight Bounds for Stochastic and Online Optimization

Distributed Bandit Learning: Near-Optimal Regret with Efficient Communication.

PrivLogit: Efficient Privacy-preserving Logistic Regression by Tailoring Numerical Optimizers