Mutual friend attack prevention in social network data publishing

Abstract

Due to increasing demand of publishing social network data, privacy has raised more concern for data publisher. There are different risks and attacks still exist that can breach user privacy. Online social network such as Facebook, Google Plus and LinkedIn provide a feature that allows finding out number of mutual friends (NMF) between two users. Adversary can use such information to identify individual user and his/her connections. As published dataset itself reveals mutual friends information for each connection, it becomes very easy for an adversary to re-identify the individual user. Existing anonymization techniques for mutual friends attack are based on edge anonymization. It performs edge anonymization operation without considering the NMF-requirement of other edges that results into more edge insertion operations. Due to that, the data utility of anonymized dataset is very low. In this paper, we propose the anonymization approach that works on the mutual friend sequence. It ensures that, there exist at least k elements in mutual friend sequence that holds same value. The vertex selection process to increase the number of mutual friend (NMF) for one edge reduces the mutual friend anonymization requirement for other edges too. The experimental results demonstrate that the proposed anonymization approach preserve the privacy and the utility of the published dataset against mutual friend attack.