Learning Temporal User Pro les of Web Browsing Behavior

As more people use the Web to gather information, communicate, work, and otherwise have fun, learning user pro les of Web browsing behavior can help personalize search results and identify persons of interest. Learning individual user pro les from Web browsing behavior also has applications in cybersecurity as a continuous soft biometric technique to verify a user's claim of identity, in online opinion mining to eliminate duplicate users, in fraud detection to discriminate between normal and anomalous behavior, in adaptive user interfaces to contextualize the presentation of information, and in recommender systems based on collaborative ltering techniques. The contribution of this paper is a novel method for modeling Web browsing behavior and learning user pro les using conditional random elds with experimental results in authenticating users from a user study.