Knowledge Graph Publishing with Anatomy, Toward a New Privacy and Utility Trade-Off

The impact of the Big and Open data phenomena, as well as the need for a certain confidentiality, obliges the emergence of new knowledge management tasks. Anonymization is one of them and directly impacts knowledge discovery. In fact, the utility of an anonymized dataset, i.e. the capacity to discover insights, largely depends on the level of privacy one has selected. In this paper, we tackle Knowledge Graphs, i.e. data represented with the Resource Description Framework data model, and present a new anonymization operation that proposes a trade-off between data privacy and utility. This operation corresponds to an extension of anatomy, originally designed for the relational model, where the knowledge graph’s ontology is used to improve the utility. We evaluate our prototype on several synthetic datasets and demonstrate the potential of this novel approach.