Status : Verified
Personal Name Abad, Earl Andrea G.
Resource Title On the release of differentially private graph statistics and synthetic graphs
Date Issued December 2019
Abstract The analysis of networks is a significant endeavor that enables researchers to understand certain patterns and behavior, which benefits individuals and society as a whole. However, this is accompanied by privacy risks among participating individuals. Mechanisms to preserve privacy in databases have been proposed throughout the years. One of them is differential privacy, a perturbation technique wherein noise is added to the aggregated query answers to protect individuals and hide their participation in a database. In the context of networks, differential privacy is still young. It roughly falls into two directions: the release of differentially private graph statistics (i.e., the use of direct approach) and the release of synthetic graphs with differentially private parameters (i.e., the use of model-based approach). In this study, we see how each technique provides utility and privacy, and which network generation models are appropriate for different datasets and queries.
Degree Course MS Computer Science
Language English
Keyword Differential Privacy, privacy, complex networks, small world networks, graph statistics
Material Type Thesis/Dissertation
Preliminary Pages
129.39 Kb