Resisting structural re-identification in anonymized social networks

Abstract

We identify privacy risks associated with releasing network data sets and provide an algorithm that mitigates those risks. A network consists of entities connected by links representing relations such as friendship, communication, or shared activity. Maintaining privacy when publishing networked data is uniquely challenging because an individual's network context can be used to identify them even if other identifying information is removed. In this paper, we quantify the privacy risks associated with three classes of attacks on the privacy of individuals in networks, based on the knowledge used by the adversary. We show that the risks of these attacks vary greatly based on network structure and size. We propose a novel approach to anonymizing network data that models aggregate network structure and then allows samples to be drawn from that model. The approach guarantees anonymity for network entities while preserving the ability to estimate a wide variety of network measures with relatively little bias.

Reidentification and k-anonymity: a model for disclosure risk in graphs

“…We require that the clustering algorithm returns a partition of the vertex set V of the graph and that each cluster or class of vertices contains at least k vertices. In [10], the authors use simulated annealing in order to find a partition of the vertices that satisfy k-anonymity and minimizes information loss, via a maximum likelihood approach. Heuristic methods are nice, because they work.…”

A Privacy Protection Method for Social Network Data against Content/Degree Attacks

“…Hay et al [7] considered such an adversary and proposed a method that produces at least k candidate answers to any structural queries, namely k-candidate anonymity. Zou et al [18] further observed that an adversary can perform different structural queries to re-identify a specific person.…”

“…In order to achieve the anonymization of social network data, nodes and edges in the social network can be modified, added, deleted, or clustered. Several methods have been proposed to preserve the privacy of social networks in the past [6], [7], [9], [13], [15], [17], [18]. However, most of them considered only structural information [7], [9], [17], [18].…”

“…Several methods have been proposed to preserve the privacy of social networks in the past [6], [7], [9], [13], [15], [17], [18]. However, most of them considered only structural information [7], [9], [17], [18]. Some recent work [6], [13], [15] considered both structural information and content information, but have some limitations as described below.…”

Towards a Privacy-Preserving Solution for OSNs

“…All attributes are available in plaintext to the OSN service providers and, depending on the configurations, some of them are available to third parties. It is not surprising that a subset of the profile attributes can already identify a user, even after anonymization [4,6]. Therefore, it is an interesting task to design a solution for users to: (1) protect their private profile attributes; (2) establish friendship with strangers based on their profile similarities (this is the main reason why users want to publish their profiles).…”