Node Protection in Weighted Social Networks

Weighted social network has a broad usage in the data mining fields, such as collaborative filtering, influence analysis, phone log analysis, etc. However, current privacy models which prevent node re-identification for the social network only dealt with unweighted graphs. In this paper, we make use of the special characteristic of edge weights to define a novel k-weighted-degree anonymous model. While keeping the weight utilities, this model helps prevent node re-identification in the weighted graph based on three distance functions which measure the nodes' difference. We also design corresponding algorithms for each distance to achieve anonymity. Some experiments on real datasets show the effectiveness of our methods.