User Authentication from Web Browsing Behavior

As anticipated in True Names by Vernor Vinge, identity has been recognized as our most valued possession in cyberspace. Attribution is a key concept in enabling trusted identities and deterring malicious activities. As more people use the Web to communicate, work, and otherwise have fun, is it possible to uniquely identify someone based on their Web browsing behavior or to differentiate between two persons based solely on their Web browsing histories? Based on a user study, this paper provides some insights into these questions. We describe characteristic features of Web browsing behavior and present our algorithm and analysis of an ensemble learning approach leveraging from those features for user authentication.