Internet Filtering in China

T he government of the People's Republic of China has a longstanding set of policies restricting their citizens' exposure to information. 1 The Internet poses a new challenge to such censorship because of the breadth of online content, the rapidity with which sources of content can be moved or mirrored, and because content sources are often remote from Chinese jurisdiction. A desire to capture the economic benefits of networked computing while maintaining control over the public's Internet exposure has led to a variety of strategies to split the difference between allowing unfettered access to the global Net and refusing to countenance any deployment beyond trusted elites. We collected data on the methods, scope, and depth of selective barriers to Internet usage through networks in China. Tests conducted from May through November 2002 indicated at least four distinct and independently operable Internet filtering methods — web server IP address, DNS server IP address, keyword, and DNS redirection — with a quantifiable leap in filtering sophistication beginning in September 2002. As with most technical filtering regimes, whether implemented at the client, Internet service provider, or backbone level, no list of the sites blocked or the methodologies used to block them has been made available by those doing the filtering. Further, while the government-connected Internet Society of China (not a chapter of the international Internet Society) has asked ISPs and content creators to sign a pledge that includes self-filtering, few official statements document that government-maintained Web filtering exists, much less the criteria employed and thresholds necessary to elicit a block. We therefore investigated the growing methods of Internet filtering, and collected and distributed a list of blocked sites and pages on our web site at http://cyber.law.harvard.edu/ filtering/china — a diverse list that is large in absolute terms, even if small relative to the size of the Internet and to the total amount of still-undocumented blocked content. Such a list lets us assess the nature and scope of filtering in China, paying particular attention to inaccessible non-sexually-explicit Web sites. Our testing relied on two separate data collection methods — through modems and open proxy servers. From 20 March to 6 May 2002, we connected with an international telephone call by modem using dialup accounts with several Chinese ISPs. After 6 May, our modems were unable to negotiate a handshake with modems answering at any Chinese ISPs, a failure consistent across multiple phone lines, locations, …