Tor is a low-latency network built for anonymity online. Due to ethical concerns, attacks and protocol changes cannot be attempted on the actual network and as a result, considerable work has been done in an attempt to precisely model the network itself. While many simulation platforms attempt to model the Tor network for research purposes, most platforms contain simple unrealistic models of the behavior of actual Tor users. Without an appropriate model of network traffic, the simulated network cannot fully mimic the real Tor network, potentially weakening any previous conclusions. This thesis presents an examination of the effects of clustering algorithm and distance measure choices on the quality of a user model. In particular, the goal is to improve a previously established generative model of outbound data with the end goal of realistically simulating user traffic of the Tor network.
"},{"label":"Degree","value":["Bachelor of Arts with Departmental Honors"]},{"label":"Administrative Note","value":"2015