Privacy Preserving Data Sharing in Online Social Networks

Social networks pervaded human lives in mostly each aspect. The vast amount of sensitive data that users produce and exchanged on these platforms call for intensive concern about information and privacy protection. Moreover, the users’ statistical usage data collected for analysis is also subject to leakage and therefor require protection. Although there is an availability of privacy preserving methods, they are not scalable, or tend to underperform when it comes to data utility and efficiency. Thus, in this paper, we develop a novel approach for anonymizing users’ statistical data. The data is collected from the user’s behavior patterns in social networks. In particular, we collect specific points from the user’s behavior patterns rather than the entire data stream to be fed into local differential privacy (LDP). After the statistical data has been anonymized, we reconstruct the original points using nonlinear techniques. The results from this approach provide significant accuracy when compared with the straightforward anonymization approach.