DATA METHODOLOGY

The Great Fire Wall (GFW) blocks "unwanted" websites mainly via three techniques: DNS poisoning, TCP RST injection, and IP blocking. Based on our understanding of GFW's mechanisms, we develop packet-level detectors to measure whether, and how the GFW blocks a website.

The "total list" of blocked websites is generated by testing alexa top websites. The list presents a picture of GFW's censorship policy. It can also be used for censorship circumvention. For example, if one website is only blocked by TCP RST, the website can avoid GFW's blocking by deploy HTTPS. We run the test periodically to keep the list up-to-date.

The "top 50" is a list of selected websites that are frequently visited by Chinese circumventors. The list serves as a navigation page for new circumventors to start exploring uncensored Internet.

Although GFW is "good enough" to block most of unwanted web visits, its implementation is not stable, occasionally letting blacked listed web visits going through. Its instability could have negative effects on the accuracy of our measurement results. While we design our measurement with this factor in mind, our results may still have a few errors/inaccuracies. You are welcome to correct us if you find some inaccurate results.

Further Notes:

The general access testing is more reliable. Sometime, the specific blocking mechanism (DNS, RST, IP) is not reported because of false negative of GFW. repeat same test could have more reliable results. In particular, the testing for DNS poisoning and TCP RST should be free of false positive, but have some false negative because of GFW’s problem. Repeated testing should reduce the effect of false negative.