Determining privacy utility trade-off for Online Social Network data publishing

Abstract

Online Social Network (OSN) data holders like Facebook, Twitter, Linked-In release their data to third parties such as researchers, data mining practitioners etc. Third parties mine the released data and help data holders gain deeper insights about the network. Releasing the social network graph in its actual form results in loss of privacy. As a result OSN users could end up losing trust that they have on the data holders which would degrade the growth of social capital immensely. To prevent unwanted privacy breaches the social network graph is anonymized before it is released. Various graph anonymization algorithms could be used for anonymizing the social network graph. These algorithms perturb actual graph to produce the final graph which could be released for mining. Perturbation reduces utility of the graph and gives better privacy protection. Graph released with fewer modifications would have greater utility but would also increase the risk of privacy breaches. Balancing the right combination of privacy-utility is a challenging task. Hence, in this work we implement and validate a solution which helps the data holders choose the best edge anonymizing scheme that could guarantee an optimal privacy utility trade-off for publishing OSN data.