Enabling efficient privacy-preserving subgraph isomorphic query over graphs

Benefiting from the cloud computing, graph owners tend to store their data graphs on the cloud server to release the local computation and storage burden. The complex subgraph isomorphic query, as an important graph query operation, is completed by the cloud server in this way. In order to realize privacy-preserving subgraph isomorphic query over graphs in cloud servers, some schemes have been proposed. However, these schemes need to mine specific subgraphs as index features, which causes high computation burden to graph owner. In addition, they need to traverse the entire index to find the data graphs that meet the requirements, which makes the query inefficient. In this paper, we focus on implementing an efficient and privacy-preserving subgraph isomorphic query over graphs, and give a practical scheme. We extract trees and cycles from the data graphs as features instead of the specific subgraphs, which reduces the precomputation overhead for the graph owner. We generate fingerprint vectors for data graphs based on the features, and construct a tree-based index for vectors. The tree based index enables our scheme to provide faster query service by cutting off the branch structure that does not meet the requirements without traversing the whole index. Our scheme follows the "filtering-and-verification " principle and ensures the efficiency of the query process. To support graph semantic security without privacy leakage, we adopt the secure inner product computation technology to protect the privacy in the index. This technology ensures that the index and the query process will not disclose the privacy of users. The security proof and the performance evaluation show that our scheme is secure and efficient. (c) 2022 Elsevier B.V. All rights reserved.