Paper 12 (Research track)

A Declarative Approach to Anonymize Linked Data Graphs

Author(s): Rémy Delanaux, Angela Bonifati, Marie-Christine Rousset, Romuald Thion

Abstract: The growth of the LOD (Linked Open Data) graph due to the injection of new information from individuals, enterprises and governments leads to rethink the available data protection methods mainly conceived for relational data.
In this paper, we introduce a declarative framework for privacy-preserving Linked Data publishing in which privacy and utility policies are seamlessly encoded as SPARQL queries.

We focus on the problem of query-driven anonymization of a graph dataset prior to its publication in the LOD graph.
We formalize the problem of finding sequences of modifications to a graph database guaranteeing that both privacy and utility criteria as encoded in their respective policies are satisfied.

We propose a data-independent method that inspects only the privacy and utility policies and does not need to navigate the graph instance in order to determine the sequence of operations needed for policy application.
We prove the soundness of our method and gauge its effectiveness and runtime by means of an experimental assessment.

Finally, we conclude the paper by discussing the generality of our
declarative framework and its possible extensions.

Keywords: Linked Data Anonymization; Data Privacy; Linked Open Data; Privacy Policy; Utility Policy; Privacy/Utility Tradeoff; Declarative Privacy; RDF Graph Anonymization

Share on

Leave a Reply

Your email address will not be published. Required fields are marked *