Stability Enhanced Privacy and Applications in Private Stochastic Gradient Descent

Abstract

Private machine learning involves addition of noise while training, resulting in lower accuracy. Intuitively, greater stability can imply greater privacy and improve this privacy-utility tradeoff. We study this role of stability in private empirical risk minimization, where differential privacy is achieved by output perturbation, and establish a corresponding theoretical result showing that for strongly-convex loss functions, an algorithm with uniform stability of $\beta$ implies a bound of $O(\sqrt{\beta })$ on the scale of noise required for differential privacy. The result applies to both explicit regularization and to implicitly stabilized ERM, such as adaptations of Stochastic Gradient Descent that are known to be stable. Thus, it generalizes recent results that improve privacy through modifications to SGD, and establishes stability as the unifying perspective. It implies new privacy guarantees for optimizations with uniform stability guarantees, where a corresponding differential privacy guarantee was previously not known. Experimental results validate the utility of stability enhanced privacy in several problems, including application of elastic nets and feature selection.

References in corpus (9)

• Scikit-learn: Machine Learning in Python
• Deep Learning with Differential Privacy
• Differential Privacy and Machine Learning: a Survey and Review
• Swapout: Learning an ensemble of deep architectures
• Privacy for Free: Posterior Sampling and Stochastic Gradient Monte Carlo
• Learning with Differential Privacy: Stability, Learnability and the Sufficiency and Necessity of ERM Principle
• Stability and Convergence Trade-off of Iterative Optimization Algorithms
• Generalization Bounds for Uniformly Stable Algorithms
• To Drop or Not to Drop: Robustness, Consistency and Differential Privacy Properties of Dropout