Clustering-based Safety Grouping Strategy for Bipartite Graph Data Publishing

The current data publishing process for bipartite graph data only focusing on privacy security and ignoring the needs of the application of the original data, we consider the data application requirements and data features by node clustering properties. Then we propose a clustering-based safety grouping strategy that has provable guarantees to resist a variety of attacks, and introduce two algorithms with the different strategies, first-clustering and clustering-while-grouping. Theoretical analysis and experimental results show that the method can well improve the availability of released data and avoid privacy disclosure in the graph data publishing. The clustering-based safety grouping strategy can offer strong tradeoffs between privacy and utility for data publishing with a variety of complex relationships of individual interaction not just in digital libraries.