Towards Web Usage Attribution via Graph Community Detection in Grouped Internet Connection Records

Abstract

Internet connection records can be very useful to digital forensic analysts in producing Internet history timelines and making deductions about the cause and effect of activity. However, the available data may include only a subset of the data that would be available from physical extraction. For example, the new UK legislation allows the collection of host website details, time of access and subscriber details, but not the specific uniform resource locator visited. Here, we investigate how to process data from Internet connections records to extract the websites, and construct the sessions of activity that are likely to be idiosyncratic the individual users, from the set of multiple possible users. We demonstrate how to display Internet history sessions as a network and perform graph community detection, showing a scheme for breaking up the component parts of the Internet history sessions into groups. We also introduce the use of websites' relative popularity for identifying websites that are likely to be meaningful to particular users of particular devices, further improving the accuracy of attributing a particular activity session to a particular user at a particular point in time.