Detecting Internet Filtering from Geographic Time Series
Joss Wright, Alexander Darer, Oliver Farnan · arXiv (Cornell University) · 2015
We propose an approach based on principle component analysis to identify per-country anomalous periods in traffic usage as a means to detect internet filtering, and demonstrate the applicability of this approach with global usage statistics from the Tor Project. In contrast to previous country-specific investigations, our techniques use deviation from global patterns of usage to identify countries straying from predicted behaviour, allowing the identification of periods of filtering and related events in any country for which usage statistics exist. To our knowledge the work presented here is the first automated approach to detecting internet filtering at a global scale. We demonstrate the applicability of our approach by identifying known historical filtering events as well as events injected synthetically into a dataset, and evaluate the sensitivity of this technique against different classes of censorship events. Importantly, our results show that usage of circumvention tools, such as those provided by the Tor Project, act not only as direct indicators of network censorship but also as a meaningful proxy variable for related events such as protests in which internet use is restricted.