Privacy Aspects of Social Graphs
A defining feature of Web 2.0 is the explicit mapping of social connections. Beyond online social networks, everything from blogs to email services to Wikipedia is mapping social connections to add social context. This is causing some unexpected privacy problems and requiring us to combine sociology, graph theory, and inference control. Unfortunately, most results thus far have been discouraging for privacy: anonymising social graphs for research seems impossible, de-anonymising pseudonymous users by correlating multiple networks seems practical, and social graphs are difficult to keep private if a small number of nodes are compromised. This talk will survey these results and address the privacy implications of displaying a small sample of edges for each node for promotional and disambiguation purposes.
Background Reading: http://www.cl.cam.ac.uk/~jcb82/doc/8_friends_paper.pdf