We have proposed a subgraph generalization approach for sharing information between terrorist or criminal social networks. The subgraph generalization tries to retain the degree of privacy in information sharing and yet share information with high utility so that an effective social network analysis can still be conducted. In our experiment, it shows that the proposed technique is promising to reduce the error substantially when we compare the computation of closeness
centrality without and with information sharing by subgraph generalizations. In the future, we shall further investigate other generalization techniques and determine what information to be shared under different circumstances or analysis.