Dipartimento di Scienze Statistiche

DSS Statistics Seminar  January 19, 2026, 10:30  In person.Room V (CU002)  Webinar^.https://uniroma1.zoom.us/j/83625004899?pwd=bXCtz0 mp759PUh2lkqT0BUoVa0Uegg.1  Passcode: 123456      From insufficient statistics to Bayesian privacy     Christian Robert Universitè Paris Dauphine, PSL & University of Warwick   From insufficient statistics to Bayesian privacy Christian Robert Université Paris Dauphine, PSL & University of Warwick While several results in the literature demonstrate that Bayesian inference approximated by MCMC output can achieve differential privacy with zero or limited impact on the ensuing posterior, we reassess this perspective via an alternate “exact" MCMC perturbation within a federated learning setting. Since the ensuing privacy criterion is mostly related to a slowing-down of MCMC convergence rather than a generic gain in protecting data privacy, we propose an alternative decision-theoretic framework that accommodates more realistic privacy constraints.