A novel subgraph K+-isomorphism method in social network based on graph similarity detection

• Published in: Soft computing (Berlin, Germany) Vol. 22; no. 8; pp. 2583 - 2601
• Main Authors: Rong, Huan, Ma, Tinghuai, Tang, Meili, Cao, Jie
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2018; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Clusters; Computational Intelligence; Control; Engineering; Graph theory; Isomorphism; Mathematical Logic and Foundations; Mechatronics; Methodologies and Application; Privacy; Publishing; Robotics; Similarity; Social networks; Subgraph; anonymization; Social network; Privacy preservation; Subgraph similarity detection; Maximum common subgraph
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-017-2513-y

In this paper, we propose a novel K + -isomorphism method to achieve K -anonymization state among subgraphs or detected communities in a given social network. Our proposed K + -isomorphism method firstly partitions the subgraphs we have detected into some similar-subgraph clusters followed by graph modification conducted in every cluster. In this way, it is feasible to publish preserved structures of communities or subgraphs and every preserved structure actually represents a cluster of at least K subgraphs or communities which are isomorphic to each other. The contributions of this paper are listed as follows: On the one hand, we improve a maximum common subgraph detection algorithm, MPD - V, which is a core technique for graph similarity detection involved in partition phase of our proposed K + -isomorphism method; on the other hand, with minor adjustment, we utilize some current techniques as an innovative combination to finish the partition and modification of similar-community cluster in K + -isomorphism method. The experiments have shown that the improved MPD - V method has much better efficiency to search larger common subgraphs with acceptable performance compared with its prototype and other techniques. Moreover, our proposed K + -isomorphism method can achieve the K -isomorphism state with less modification of original network structure, or lower anonymization cost compared to the current K -isomorphism method.