Friendship links-based privacy-preserving algorithm against inference attacks

Abstract

Directly publishing the original data of social networks may compromise personal privacy because social relationship data contain sensitive information about users. To protect the social relationships against inference attacks and achieve the trade-off between privacy and utility, we propose a privacy protection algorithm that combines the friendship links of central nodes (PPCN) in a dynamic social network. In the preparation work, we design two indices for user influence based on the characteristics of social networks that can identify central nodes (Definition 1) in a network. Operating central nodes can effectively protect user privacy and improve algorithm efficiency. Then we propose the PPCN algorithm to classify the friendship links of central nodes into three levels, which achieves the trade-off between privacy and utility. Considering that the added links may increase the risk of privacy disclosure, a substitution coefficient θ (Definition 4) is designed to measure the probability of two strangers becoming friends. Experimental results show that the privacy-utility trade-off (PUTO) value of the PPCN algorithm is an average 29.43% lower than that of other methods, achieving a better trade-off between privacy and structural utility. In addition, the PPCN algorithm only runs for 3.59 s, which performs better than most algorithms.