Correlation of seemingly innocuous information can create inference chains that tell much more about individuals than they are aware of revealing. However, while media coverage occasionally draws attention to privacy leaks on individual web sites, there is still no comprehensive analysis of the fundamental risks that users face in their online worlds. This project pursues such a study, focusing in particular on the threat of personal, yet publicly available information that can be correlated with modern multimedia retrieval and content analysis technologies. One thrust of the work is informing users about potential risks by exposing the broader possibilities that arise with more sophisticated privacy attacks. A second thrust concerns understanding the control that users can exercise over their privacy in the light of such potential. When analyzing the impact of "global inference" on privacy, the primary conceptual challenge concerns understanding the trade-off between the benefits that providing personal information to web services offers, versus the risks that doing so entails. By combining expertise from two traditionally separate communities---network security and multimedia---this work advances the start of privacy protection in an area that is poised to raise in importance as more information moves into public spaces. The project enables users to better understand the threats they are facing by developing scenarios that intuitively demonstrate the relevant effects; and it develops novel tools supporting them in better protecting their privacy. The research results will benefit the user community by promoting risk awareness and empowering control. The project also includes outreach activities including working with Berkeley Foundation for Opportunities in Information Technology (BFOIT) to recruit minority students in this research, organizing summer schools, and raising public risk awareness by working with Identity Theft 911.
TC: Medium: Understanding and Managing the Impact of Global Inference on Online Privacy