Minimax rate for multivariate data under componentwise local differential privacy constraints

par Chiara Amorino

mardi 24 mars 2026, 14:00 → 15:00 Europe/Paris

Our research analyses the trade-off between maintaining privacy and preserving statistical accuracy when dealing with multivariate data subject to componentwise local differential privacy (CLDP). Under CLDP, each component of the private data is released through a separate privacy channel. This allows for varying levels of privacy protection for different components or for the privatization of each component by different entities, each with their own distinct privacy policies. We develop general techniques for establishing minimax bounds that quantify the statistical cost of privacy as a function of the privacy levels a₁,…,a_d of the d components. The versatility and efficiency of these techniques are demonstrated through various statistical applications. 

The talk is based on a joint work with A. Gloter.