Cryptography, Security and Privacy
Information Privacy I

Privacy Amplification of Iterative Algorithms via Contraction Coefficients

Shahab Asoodeh, Flavio Calmon, Mario Diaz

Date & Time

01:00 am – 01:00 am


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.


Shahab Asoodeh

Harvard University

Flavio Calmon

Harvard University

Mario Diaz

Universidad Nacional Autonoma de Mexico

Session Chair

Prakash Narayan

University of Maryland, College Park