Mining Web Usage Profiles from Proxy Logs: User Identification

Personal web-surfing habits of individual users vary drastically. Thus, the selection and frequency of requested URLs can be viewed as an individual finger-print allowing to identify a user, which has many security and target marketing applications. In this paper, to model web usage, we derive web usage profiles for each user and employ instance-based classification methods to predict the most likely user in the monitored network. Thus, the proposed methods are scalable for large numbers of users. In our experimental evaluation, the accuracy of user identification is verified.