Timezone: »

Learning with User-Level Privacy
Daniel Levy · Ziteng Sun · Kareem Amin · Satyen Kale · Alex Kulesza · Mehryar Mohri · Ananda Theertha Suresh

Thu Dec 09 08:30 AM -- 10:00 AM (PST) @
We propose and analyze algorithms to solve a range of learning tasks under user-level differential privacy constraints. Rather than guaranteeing only the privacy of individual samples, user-level DP protects a user's entire contribution ($m \ge 1$ samples), providing more stringent but more realistic protection against information leaks. We show that for high-dimensional meanestimation, empirical risk minimization with smooth losses, stochastic convex optimization, and learning hypothesis classes with finite metric entropy, the privacy cost decreases as $O(1/\sqrt{m})$ as users provide more samples. In contrast, when increasing the number of users $n$, the privacy cost decreases at a faster $O(1/n)$ rate. We complement these results with lower bounds showing the minimax optimality of our algorithms for mean estimation and stochastic convex optimization. Our algorithms rely on novel techniques for private mean estimation in arbitrary dimension with error scaling as the concentration radius $\tau$ of the distribution rather than the entire range.

Author Information

Daniel Levy (Stanford University)
Ziteng Sun (Cornell University)
Kareem Amin (Google Research)
Satyen Kale (Google)
Alex Kulesza (Google)
Mehryar Mohri (Google Research & Courant Institute of Mathematical Sciences)
Ananda Theertha Suresh (Google)

More from the Same Authors