Poster: shaping network topology for privacy and performance

While it is important to design anonymity systems to be robust against attacks, it is also important to provide good performance to users. We explore ways to improve the security and performance of anonymity systems by building both security and performance properties into the network topology. In particular, we study an expander graph based network topology and apply link-based performance metrics in order to build the topology graph. Such a network can be constructed to have enhanced performance and similar security properties to restricted route topologies with random links. Results show that a sparse, D-regular expander graph topology provides nearly the same security, as measured by the likelihood of an incoming stream exiting through any node in the network, as with a fully-connected graph. Further, when the expander graph is constructed with a bias towards faster links, there is a considerable gain in performance without much loss of security.