Chapter 3 - Privacy-Preserving Social Network Integration, Analysis, and Mining

Abstract

Social network analysis and mining has drawn a significant attention in the recent years due to the proliferation of online communities and the advance of information-sharing technologies. Social network analysis and mining techniques discover the knowledge embedded in the structure of social networks. Such knowledge is useful in many domains such as marketing, management, epidemiology, homeland security, and psychology. Sharing of social networks between organizations enables knowledge discovery from an integrated and larger social network obtained from multiple sources. However, concerns over respect for individuals’ privacy usually prohibit the possibility of information sharing. Early research work on respect for privacy focused on relational data and some recent works have extended this to social network data. The work on privacy of social network data relies on anonymity and perturbation. These techniques are developed for the purpose of data publishing but the utility of the published data on social network analysis and mining has not been explored. The global network structure is mostly retained after perturbation and anonymization, so that analysis can be conducted on the structural properties. However, without integrating the social networks with those of other organizations, the results of social network and analysis mining tasks can only be obtained from an incomplete social network since each organization owns a partial network of the complete social network. It is desirable to share the insensitive and generalized information to support social network analysis and mining, whilst at the same time respecting individuals’ privacy. In this chapter, we explore the subgraph generalization approach to construct generalized social networks in which only insensitive and generalized information is shared. We also investigate how to integrate the generalized information with social networks of other organizations in order to achieve better performance in the social network analysis and mining tasks. At the same time, we need to ensure that a prescribed level of privacy leakage tolerance is satisfied. The measurement of privacy leakage should be independent of the privacy-preserving techniques of integrating social network data. Privacy-preserving social network integration, analysis, and mining are beneficial to many practical domains, such as public health, national security, marketing, psychology, and many others. Social network sharing will be encouraged and knowledge can be discovered to provide better support for practical applications such as investigating the spreading of contagious diseases and identifying the communication patterns across and within terrorist or criminal groups.