Privacy Amplification of Iterative Algorithms via Contraction Coefficients
| Autor: | Flavio P. Calmon, Mario Diaz, Shahab Asoodeh |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
FOS: Computer and information sciences
Computer Science - Machine Learning Data processing Computer Science - Cryptography and Security Computer Science - Information Theory Information Theory (cs.IT) 020206 networking & telecommunications Machine Learning (stat.ML) 02 engineering and technology 010501 environmental sciences 01 natural sciences Machine Learning (cs.LG) Total variation Borda–Carnot equation Stochastic gradient descent Statistics - Machine Learning ComputingMethodologies_SYMBOLICANDALGEBRAICMANIPULATION 0202 electrical engineering electronic engineering information engineering Differential privacy Contraction (operator theory) Algorithm Cryptography and Security (cs.CR) 0105 earth and related environmental sciences Mathematics |
| Zdroj: | ISIT |
| DOI: | 10.48550/arxiv.2001.06546 |
| Popis: | We investigate the framework of privacy amplification by iteration, recently proposed by Feldman et al., from an information-theoretic lens. We demonstrate that differential privacy guarantees of iterative mappings can be determined by a direct application of contraction coefficients derived from strong data processing inequalities for $f$-divergences. In particular, by generalizing the Dobrushin's contraction coefficient for total variation distance to an $f$-divergence known as $E_{\gamma}$-divergence, we derive tighter bounds on the differential privacy parameters of the projected noisy stochastic gradient descent algorithm with hidden intermediate updates. Comment: Submitted for publication |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|