Privacy preservation based on clustering perturbation algorithm for social network

Abstract

With a vigorous development of information release for social network, it is now an urgent question to protect sensitive information. The sensitive information concerned by the attackers may be usually located in a local group of large scale social networks and the operations of privacy protection are needed to minimize changes for overall structure of the network to maintain data availability. In this paper, a clustering perturbation algorithm to preserve privacy for social network was proposed considering preservation privacy of vertices properties and community structures simultaneously. The proposed algorithm introduced a strategy of exchanging attributes between vertices with same degree randomly to induce attackers to search for false targets and maintain whole structure of network. Furthermore, a perturbation strategy with tiny influences based on local clustering and modifying edges complementarily was adopted to decrease the risk of privacy disclosure considering minimum loss of network structure and data information. The experimental results showed that the proposed algorithm has more advantages over other existing state-of-the-art approaches in privacy preservation and effectiveness of social network.