Zheleva ACM 2009
This a Paper discussed in Social Media Analysis 10-802 in Spring 2011.
Citation
Zheleva, E. and Getoor L. To Join or Not to Join: The Illusion of Privacy in Social Networks with Mixed Public and Private User Profiles. In ACM 2009, April 20-24, Madrid, Spain
Online version
Summary
This paper investigates what sensitive information can be inferred from friendship and group membership information in social networks such as Facebook, Orkut and Flickr. In such social networks the persons profile information can be marked as private, but friendship links and group affiliations are often visible to the public. The paper proposes eight privacy attacks using different classifiers and features.
Sensitive information is in this paper defined as attributes such as age, political affiliation or location. In the social network there are users for which this information is hidden and others for which it is observed, depending on their privacy settings. The goal of the paper is to predict those values for the users that hide this information. The approach consist of Naive Bayes classifier learning for the a specialized graphical model.
Their experiments show that groups can leak a significant amount of information, although not joining homogeneous groups preserves privacy better. On the other hand it turned out that link based methods did not reveal that much information. Although related work Liben-Nowell PNAS 2005 [1] shows that on other datasets the links actually do help.
References 1. D. Liben-Nowell, J. Novak, R. Kumar, P. Raghavan and A. Tomkins. Geographic routing in social networks. PNAS, 102(33):11623–11628, August 2005
