Privacy preserving subgraph isomorphism query for dynamic graph database

Abstract

Graph is an important structure for presenting the data with internal connections. Subgraph isomorphism query is one of the most important operations in graph database. Unfortunately, the calculation of subgraph isomorphism query is very complex. The data owners choose to store the graph data in the cloud server and let the cloud server undertake the complex query calculation. In order to solve the security risks from the untrusted cloud server, several privacy preserving subgraph isomorphism query schemes have been proposed. Nonetheless, all existing schemes are designed for static graph database. Once the graph database is dynamically updated, these schemes will not work well anymore. In order deal with this problem, we propose a privacy preserving subgraph isomorphism query scheme for dynamic graph database. We use enumeration and hash mapping methods to quickly generate feature vectors for graphs. We use clustering algorithm to group feature vectors and use feature vector list to store a group of similar feature vectors, which not only reduces the size of the tree index, but also makes the index easy to be updated. In order to realize privacy protection and subgraph isomorphic query, the improved secure Euclidean distance algorithm is adopted in our scheme. We consider three types of operations: addition, deletion and modification, which are used to update the index and database synchronously. When the index needs to be updated, our scheme can quickly locate the index nodes that need to be updated and make the update operation. Security analysis and experiments show that our scheme is secure and efficient.