CIRC09 – Censorship and surveillance on the Chinese Internet

Liao Hang Teng of the Oxford Internet Institute is interested in ways that Chinese internet authorities are mediating between chaos and control. He introduces us to two terms – zhi and luan 亂. Zhi means “order, governance, control or cure”, while luan means “disorder, instability and chaos”. People in China talk about history in terms of periods of zhi or luan, Liao tells us, whether periods of time were chaotic or ordered.

Beijing has been replacing the luan of the internet with the zhi of control, he tells us. A 2007 study suggests that 80% of Chinese users prefer control of the internet, and prefer that the government do it. There’s a sense that “too much freedom leads to luan”, like the 1989 Tiananmen protests, or like Taiwan’s government.

Liao suggests we stop thinking about the Great Firewall and think instead about Great Dams and Great Canals. What we’re seeing is “zoning technology versus dynamic order” – a decision to open free speech zones, much as “free-speech zones” are becoming unfortunately common at US protests. “Freedom is being introduced as a special exception permitted by Beijing.”

An example of a controlled speech zone is Baidu Baike, a participatory encyclopedia introduced as an alternative to the Chinese-language Wikipedia. William Chang, technology officer of Baidu, suggested that it makes sense that Chinese users wouldn’t want to use “a service based out there” – “It’s very natural for China to make its own projects.”

While Baidu Baike has grown sharply, it hasn’t caught up with Chinese Wikipedia. Liao sees it as a failed experiment, because it hasn’t managed to embrace the diversity of the Chinese wikipedia. He uses outlinks from the sites and the language sets represented as a proxy for linguistic diversity, and sees much less diversity in the Baidu project. He also notes that Baidu got the “grass mud horse” incident wrong, reporting on it as an animal, while the Chinese Wikipedia understood that it was a complex parody and statement about free speech.

Referring back to the 2007 survey that showed enthusiasm for government control, Liao wonders if there’s a missing option – perhaps control of the internet should be in the hands of the internet community or of civil society.

My friend and colleague Rebecca MacKinnon reports on her recent paper published in First Monday on blog censorship. She notes that in talking about censorship in China, we need to consider both censorship of sites outside the firewall (i.e., the blocking of Human Rights Watch inside China) as well as censorship inside the firewall, the censorship of content on domestic commercial websites, the takedown of domestically hosted websites and the shutdown of data centers.

Rebecca has studied the mechanisms used to censor on sites within China, and notes some early work she did with Human Rights Watch, which suggests that corporate censorship wasn’t uniform, wich meant that companies were making choices, acting in reaction to government demands, but offering their own interpretations. Nart Villeneuve at Citizen Lab followed up on this work, and was able to discover different levels of filtering between Baidu, which censored a great deal, and Yahoo, MSN and Google which censored less (they’re listed in order of decreasing censorship.)

Because the Great Firewall is so effective at blocking access to blogging platforms, most Chinese users publish on domestic platforms. Her recent research demonstrates that there’s incredible variation in what can and cannot be published on these platforms. Using excerpts from Chinese blogs and Xinhua news reports, she tested 108 potentially sensitive texts on 15 different platforms. Filtering differed wildly – one platform filtered 60 of 108 texts, while one filtered only one, and there’s a wide spread of results in between those extremes.

She outlines five ways content is blocked from publication:
– The platform simply refuses certain sensitive posts
– Posts are held for moderation, and never appear
– Posts are visible to the authors, but never to the public
– Posts are published, but removed within 24 hours
– Posted, but sensitive words are replaced with asterisks – Blogbus, for instance, blocks out all mentions of Hu Jintao, even in Xinhua articles.

One filtering method was only used by Microsoft’s Windows Live service – sensitive articles could be published and would be seen in Hong Kong or the US, but MSN geofiltered to ensure posts couldn’t be read from within China.

Rebecca tells us that very sensitive texts, like a provocative Bao Tong essay, were either not censored at all, or censored by very few platforms. She believes that the censorship method tends to overfocus on certain words, while very controversial texts might get through if written in “official language”.

Why is there such variation in platforms? We’ve only got theories, she tells us – it might have to do with relationships between certain editors and government officials. It’s going to require more study… and it’s important to do this study because this form of filtering is deeply important to Chinese users and may represent a pattern of censorship we’re likely to see in the rest of the world, not just in China.

Dave Lyons has been studying “the Golden Shield project“, which he argues may be one of the world’s great misunderstood efforts. Essays by American commentators on the project – an effort to bring computers and databases to the Chinese police force – portray it as “Police State 2.0” (Naomi Klein) or as “too creepy to bear repeating”, as James Fallows writes. Lyons thinks these authors are badly misunderstanding what’s being implemented and what it means for China.

The focus of the Golden Shield project is on bringing computers to the police, from the lowest provincial levels, up to the Ministry of Public Security, and developing eight databases: population management, criminal records, fugitives, driver’s licenses, stolen vehicles, stolen property, national security, and border control. Some of those sound pretty sinister, while others are the sorts of databases we expect police to have. Lyons has focused on the “population management” database – described as “the dragon’s head of Golden Shield” – which tracks hukou registration (a registration of your city of residence) and the “second generation” national ID card, which features an RFID chip. Most of the country has moved to this new card, which stores a great deal of personal data.

Lyons points out that forgery is a serious problem in China. It’s possible to buy fake cards quite easily – indeed, in Mission Impossible 3, a single frame of the film features a piece of graffiti advertising the services of a ban jun, a forger, with a phone number. This was sufficiently provocative that the film was never screened in China… but it circulated widely via pirate DVDs. As a result, the ban jun advertised on the wall needed to change his phone number – he got so many calls that it got in the way of his business, weeding out the merely curious from potential customers. More ambitious forgers have been known to advertise by scrawling their numbers on police cars. So while the idea of a database that tracks all Chinese people is very scary, the reality may be a lot less so. “The biggest project of the golden shield has to do with accurately identifying citizens”, which can lead towards service delivery and accurate statistics, much like the census in the US.

Another part of the Golden Shield is a video surveillance system, using CCTV. He shows us a picture of a surveillance center in Shenyang, which looks pretty intimidating. He points out that the name for this project is sometimes translated as “Skynet”, which hearkens back to the evil computer in the Terminator movies. But a second surveillance room is a photo of Chicago’s “Operation Virtual Shield”. Lyons wonders whether what we’re seeing is China catching up with how the rest of the world does law enforcement, not breaking new ground. He notes that the most frightening tech in facial recognition isn’t used in China – it’s used in Las Vegas casinos to weed out hustlers and card sharks.

Finally, Lyons suggests that we misunderstand if we connect the Golden Shield to internet censorship – while there are at least a dozen organizations involved with internet censorship, Skynet is focused much more closely to the ground.

Lyons’s talk receives a great deal of commentary from the audience, including Chinese scholars sharing their experience of how casually regulations are enforced. One talks about visiting a Chinese cybercafe without a registration card – the cybercafe administrator simply swiped her in with his own card.

Sarah Cook of Freedom House wonders if Lyons is being fair in comparing surveillance in Chicago to that in Beijing, given that residents in Chicago have legal protections they can rely on which are denied to Chinese citizens – Lyons concedes her point. But there’s another message here – simply understanding security practices based on what systems are in place is insufficient – you’ve got to look at how those systems are embedded in societies and how they’re actually carried out.