My colleague Hal Roberts, I and friends at Berkman released a paper today that attempts to estimate usage of circumvention tools, tools used to evade internet filtering. We were specifically interested in trying to compare usage of different types of tools – sophisticated blocking-resistant tools like Tor and Ultrasurf, ad-supported web proxies like Proxeasy or HideMyAss, and VPN-based systems like Hotspot Shield and Relakks. Unlike in our previous study of some of these tools, we weren’t trying to compare the functionality of these very different tools, or evaluate their performance – we just wanted to answer the question, “How many people use this tool?”
That’s not an easy question to answer. For blocking resistant tools, we used estimates provided by the tool operators. (Tor is very good about publicly posting usage metrics, and Jacob Appelbaum points out that we’re able to access anonymized logs to conduct our own analysis. We could also, Appelbaum points out, run our own Tor node and extrapolate traffic from that node. I hope we’re able to do so in a later version of the research.) We did much the same for VPN operators, though here we sent a survey out to as many companies as we could find and tried to extrapolate from the responses we got to offer an estimate of tool usage for the whole space. Because web proxies are, at their heart, script-driven websites, we’re able to estimate their usage by building a catalog and using traffic statistics from Google Ad Planner to estimate usage.
There’s a lot of extrapolation in our findings, and we tried to make it clear that we were trying to calculate orders of magnitude of usage, not more granular numbers. With that caveat, the interesting finding were that total usage of the tools we studied was under 19 million users per month. That’s a pretty big number, in absolute terms, but surprisingly small given the large number of people accessing the Internet from countries where internet censorship is widespread. If that 19 million estimate (which represented the high end of our range of estimates) is correct, that number would represent 3% of the 562 million internet users in nations that filter the web aggressively.
Of course, not all proxy users are in nations like China or Iran – some are in an unfiltered country like Mexico, and using proxies to access content that’s been geographically restricted – television shows made available on Hulu.com, for instance. It’s very hard to estimate how many users are in highly censored countries, but it’s worth noting that the economics of simple web proxies (which represent 15 million users in that high-end estimate) mean that their operators prefer to serve American users trying to access Facebook within a filtered high school network rather than Iranian users (as they are more likely to click on English-language banner ads)… and many use geolocation software to prevent users from less-profitable nations from using the tools.
There are a couple of possible conclusions we could draw from these usage statistics. One is that people in censored countries either don’t know enough about these tools, why they might want them or how to find them. That’s the logic behind efforts like the Sesawe project, which is promoting tools, localizing them into appropriate languages and providing guides to circumventing censorship.
Another possibility is that there’s reasonably widespread knowledge of these tools, but less appetite for them than we might hope. While an unfiltered internet is critical for some users – academics, foreign policy experts, activists – for many, a censored internet is a stimulating and diverting space. David Talbot wrote an excellent piece for Technology Review on this phenomenon earlier this year, helping English-speaking readers get a sense for the dynamism and complexity of conversations on China’s heavily censored internet. When countries like China block access to social media sites like Twitter or YouTube, but provide alternatives – censored, but in local languages – those tools tend to gain traction quickly. I’ve seen estimates of Chinese Twitter users as between 50,000 and 100,000 users – Sina.com’s microblogging service claims 20 million users. Given that disparity – and the fact that microblogging is still an early adopter phenomenon – it’s worth asking whether those Sina microbloggers are ignorant of Twitter, or whether they made a decision to user the platform their friends are using rather than a “freer” platform. As Evgeny Morozov commented in a Technology Review piece on our new study, “nothing is irreplaceable online.”
I made the case some months back that circumvention can’t be the only pillar of a US government internet freedom strategy – not everyone wants or needs these tools. That observation – and this paper – aren’t meant to be a case against developing better circumvention tools or promoting these tools. Instead, it’s our way of trying to get more people thinking about the tough challenge we’ve been wrestling with – how do we think about internet censorship if it’s possible – maybe even likely – that many people aren’t interested in making an effort to access an uncensored internet?