Semantic-based graph data anonymization for big data analysis

Abstract

Publishing individual-related data for big data analysis such as scientific research and merchant analysis has become frequent in this decade. Most of these data can be represented as graphs, with real world entities as graph nodes and interrelationships among entities as graph edges. Mining these released data, or corresponding graphs, may facilitate the forming of judicious strategies for marketing or promoting public health. However, individual data inevitably contain private information. How to prevent potential adversaries from recognizing the mapping between a particular graph node and a real world individual is critical for data providers. Existing tabular or graph-based techniques usually solve data anonymization problem only from one aspect, i.e., either table or graph, not from both. In this paper, we propose a semantic-based data anonymization method which employs entity ontology to anonymize the graph data for publication. Our study demonstrates that the proposed method can answer complex queries with assured privacy.