Graph Publishing with Local Differential Privacy for Hierarchical Social Networks

Abstract

With the spread of social network services, social media applications have obtained a significant amount of personal data and relational information from their users. In the era of big data, data should to be shared to make maximum use of its potential value. Therefore, the problem of personal privacy protection has become increasingly important. At present, differential privacy is a strictly proven privacy protection model, which has garnered people’s attention and research in many fields. However, the existing social network differential privacy publishing technology mainly focuses on the centralized model, that is, the data collector is assumed to be credible. The hierarchical random graph model, which satisfies differential privacy to the local model, was applied in our study. We improved efficiency and accuracy using the Monte Carlo Markov chain. The experiment showed that the local differential privacy had better utility than the centralized differential privacy under the same differential privacy conditions.