PRESERVING PRIVACY IN: PUBLISHING SOCIAL NETWORK DATA

Abstract

Recently, following the stupendous growth of various social networking sites, a vast amount of social network data has been collected. To put this data in use for commercial and research gains, more and more social network data have been published in one way or another. So an outright concern which comes up is preserving privacy in publishing social network data. With some local knowledge about individuals in a social network, an adversary may attack the privacy of some victims easily. Unfortunately, most of the previous studies on privacy preservation data publishing can deal with relational data only, and cannot be applied to social network data. The basic objective of this dissertation was working towards preserving privacy in social network data. Specifically, two types of privacy attacks were identified: neighborhood attacks and friendship attacks. Bin Zhou and Jian Pei [10] proposed a scheme for anonymization of social networks against neighborhood attacks. Later, Tripathy and Panda[13] improved their algorithm for graph isomorphism by using adjacency matrix instead of min DFS code. Tai and Yang[14] provided a different solution for anonymization against friendship attacks. In this dissertation we propose a modified approach to their algorithms which anonymizes the social network against both neighborhood and friendship attacks simultaneously. Thus, the published social network will be privacy preserved against adversary attacks based on the vertex degree and neighborhood graph knowledge about individuals