Publishing graph node strength histogram with edge differential privacy

Abstract

Protecting the private graph data while releasing accurate estimate of the data is one of the most challenging problems in data privacy. Node strength combines the topological information with the weight distribution of the weighted graph in a natural way. Since an edge in graph data oftentimes represents relationship between two nodes, edge-differential privacy (edge-DP) can protect relationship between two entities from being disclosed. In this paper, we investigate the problem of publishing the node strength histogram of a private graph under edge-DP. We propose two clustering approaches based on sequence-aware and local density to aggregate histogram. Our experimental study demonstrates that our approaches can greatly reduce the error of approximating the true node strength histogram.