Preserving weighted social networks privacy using vectors similarity

Aiming at weighted social networks, a random perturbation method based on vectors similarity is proposed. It can protect structures and edge weights of weighted social networks in multiple release scenarios. First, it partitions weighted social networks into t sub-graphs by the segmentation method based on vertex cluster using edge space of graph theory, describes these sub-graphs by vectors, and constructs vector set models of weighted social networks. Then, it adopts weighted Euclidean distance as the metrics of vectors similarity to construct the released candidate sets of t sub-graphs according to the threshold designated by publishers. Finally, it randomly selects vectors from the candidate sets to construct the released vector set, and builds the published weighted social networks based on the released vector set. The proposed method can resist multiple vertex recognition attacks, force the attackers to re-identify in a large result set that the existential probabilities of the vectors are same, and increase the uncertainty of recognition. The experimental results on the actual datasets demonstrate that the proposed method can preserve the security of individuals´ privacy, meanwhile it can protect some structure characteristics for social networks analysis and improve the released data utility.