Measuring latency for live video calls routed via vanilla and modified Tor

Abstract: Tor is an anonymity network which consists of relays (or nodes) dispersed over the world. By interlaying a connection with the relays it becomes more difficult to conclude who the sender and receiver of the connection is, however the latency is also increased. A user can choose to only connect via relays in a certain country, which would affect latency and anonymity. In this thesis we routed a live video calling application through Tor, and measured how much the usage of Tor increased the latency compared to a live video calling application not routed through Tor. We also explored different approaches to lower the latency by using modified versions of Tor, where the relays were strictly limited to either Sweden or Germany. Lastly we calculated the effects on anonymity the modified versions of Tor had, using several anonymity metrics. We concluded that the usage of vanilla Tor increased the median latency by 517 ms, compared to when Tor was not used. Limiting the relays to Sweden increased the median latency by 300 ms for Tor users in Sweden, i.e. the increase was reduced by 42 %. Limiting relays to Germany had similar latency results. However, the modifications decreased the anonymity of the user when German relays were used, and even more so when Swedish relays were used, due to Sweden’s smaller number of Tor users and relays. We also concluded that at least 39%of the additional latency for vanilla Tor could be attributed to the fact that Tor only uses the transport protocol TCP instead of both UDP and TCP. Finally, we recognized that there exist more complex methods to modify Tor which would yield better results.