Cracked: The Great Firewall of China

Computer experts from the University of Cambridge claim not only to have breached the Great Firewall of China but to have found a way to use the firewall to launch denial of service attacks against specific IP addresses in the country.

The firewall, which uses routers supplied by Cisco, works in part by inspecting web traffic for certain keywords the Chinese government wish to censor, including political ideologies and groups it finds unacceptable.

The Cambridge research group tested the firewall by firing data packets containing the word "Falun" at it, a reference to the banned Falun Gong religious group. The researchers found it was possible to circumvent the Chinese intrusion detection systems (IDS) by ignoring the forged TCP resets injected by the Chinese routers, which would normally force the endpoints to abandon the connection.

Richard Clayton of the University of Cambridge computer laboratory explained: "The machines in China allow data packets in and out but send a burst of resets to shut connections if they spot particular keywords. If you drop all the reset packets at both ends of the connection, which is relatively trivial to do, the web page is transferred just fine."

Clayton added this means the Chinese firewall can be used to launch denial of service attacks against specific IP addresses within China, including those of the Chinese government itself.

The IDS uses a stateless server, which examines each data packet both going in and out of the firewall individually, unrelated to any previous request. By forging the source address of a packet containing a "sensitive" keyword, people could trigger the firewall to block access between source and destination addresses for up to an hour at a time.

If an attacker had identified the machines used by regional government offices, they could block access to Windows Update, or prevent Chinese embassies abroad from accessing specific Chinese web content.

Clayton said: "Due to the design of the firewall, a single packet addressed from a high party official could block their web access."

Even though this technique would only block communication between two particular points on the internet, the researchers calculated that a lone attacker using a single dial-up connection could still generate a "reasonably effective" denial of service attack. If an attacker generated 100 triggering packets per second, and each packet caused 20 minutes of disruption, 120,000 pairs of endpoints could be prevented from communicating at any one time.

Clayton, speaking at the Sixth Workshop on Privacy Enhancing Technologies in Cambridge last week, said the researchers had reported their findings to the Chinese Computer Emergency Response Team.

Tom Espiner writes for ZDNet UK