Universidade Federal do Ceará
Mestrado e Doutorado em Ciências da Computação

Título: Differentially Private Selection using Smooth Sensitivity

Data: 28/08/2024
Horário: 09h30
Local:  Sala de Seminários – Bloco 952

Resumo:

Differentially private selection mechanisms offer strong privacy guarantees for queries whose canonical outcome is the top-scoring element r within a finite set R according to a dataset-dependent utility function. While selection queries are pervasive throughout data science, there are few mechanisms to ensure their privacy. Additionally, the vast majority focus on achieving differential privacy (DP) through global sensitivity, possibly corrupting the query result with excessive noise and maiming downstream inferences. We propose the Smooth Noisy Max (SNM) algorithm to alleviate this issue. In particular, the SNM algorithm leverages the notion of smooth sensitivity to provably provide smaller (upper bounds on) expected errors compared to methods based on global sensitivity under mild conditions. Empirical results show that our algorithm is more accurate than state-of-the-art differentially private selection methods in three applications: percentile selection, greedy decision trees, and random forest.

Banca examinadora:

• Prof. Dr. Javam de Castro Machado (MDCC/UFC – Orientador)
• Prof. Dr. César Lincoln Cavalcante Mattos (MDCC/UFC)
• Prof. Dr. Victor Aguiar Evangelista de Farias (UFC)
• Prof. Dr. Daniel Cardoso Moraes de Oliveira (UFF)
• Prof. Dr. Diego Parente Paiva Mesquita (FGV)