Many data analysis tasks rely on the abstraction of a graph to represent relations between entities, with attributes on the nodes and edges. Since the relationships encoded are often sensitive, we seek effective ways to release representative graphs which nevertheless protect the privacy of the data subjects. Prior work on this topic has focused primarily on the graph structure in isolation, and has not provided ways to handle richer graphs with correlated attributes.
We introduce an approach to release such graphs under the strong guarantee of differential privacy. We adapt existing graph models, and introduce a new one, and show how to augment them with meaningful privacy. This provides a complete workflow, where the input is a sensitive graph, and the output is a realistic synthetic graph. Our experimental study demonstrates that our process produces useful, accurate attributed graphs.
[ bib | Alternate Version | poster | slides | .pdf ] Back
This file was generated by bibtex2html 1.92.