The OpenNet Initiative has documented Internet filtering in more than 40 countries worldwide. Filtering allows governments to prevent people within their borders from accessing websites that governments find offensive for any of a variety of political, social, or security reasons. Circumvention tools, on the other hand, enable people inside those countries to access filtered sites, so a person in China using a circumvention tool can access "falundafa.org" even though the site is otherwise blocked in the country.
Examples of these circumvention tools include publicly/charitably funded tools like Ultrasurf, Freegate, and Tor that have been well covered in the media. But they also include mostly fee-based virtual private network (VPN) services; free, but anonymous protocol-based proxies; and mostly advertising-supported, simple web proxies. All these tools share the same basic mechanism -- they allow access to a blocked site by ferrying requests for the blocked site through an (ideally) unblocked intermediary machine.
The Berkman Center for Internet & Society at Harvard University recently issued a report examining the use of these tools. We found that, at most, only about 3 percent of people in filtering countries -- and likely many fewer -- use circumvention tools. Of the three types of tools whose use we could measure -- Ultrasurf, Freegate, and Tor; VPN services; and simple web proxies -- we found that simple web proxies were by far the most widely used. And we found that when users in filtered countries search for these tools, they overwhelmingly search for generic terms -- mostly the English word "proxy" -- that return lists of simple web proxies.
We don't know for sure what any circumvention tool or class of circumvention tools is used for, but we have lots of circumstantial evidence that simple web proxies are generally used for nonpolitical purposes like browsing social media, pornography, and gambling sites. So even the 3 percent number may overstate the use of such tools to access the politically relevant material that most people think of when they think of promoting Internet freedom in authoritarian countries.
Many press reports have portrayed circumvention tools as a clear path toward Internet (and possibly even broader political) freedom. But our findings challenge that simple story with simple numbers. To be clear, even at this proportionally small level of usage, these tools are doing a huge amount of absolute good by offering the opportunity to fulfill the basic human right to freedom of expression for as many as tens of millions of people. And the impact of the small proportion of people using these tools may be magnified if we assume those circumvention users distribute otherwise censored information out to their communities.
But even so, it is hard to argue that these tools are having revolutionary impact with such little proportional usage or even that there is a simple way to scale their usage dramatically enough to achieve such impact.
Access Versus Inconvenience
The core question that remains unanswered is why so few people use these tools. We published an evaluation of circumvention tools in 2007 that found that most of these tools work, in the broad sense that they enable users to access sites that would otherwise be blocked. But we also found that the performance of the tools ranged from slow to painfully slow, that the usability of most of the tools could be improved, and that there were significant security problems with several of the tools. These issues, particularly the slow performance of the tools, may be the main factors limiting their usage.
But a bigger limiting factor may be that users simply do not want to access filtered content badly enough to outweigh the inconvenience of using the tools.
This does not necessarily mean that users do not want to circumvent Internet control. It may mean that circumvention tools are not sufficient to provide access to the kinds of content that people want. Internet control involves much more than simply filtering some foreign sites.
Distributed denial of service attacks commonly take down controversial sites by flooding them with traffic, often from thousands of compromised computers. Many online journalists-bloggers have been arrested by their home countries, including prominent Iranian blogger Hossein Derakhshan, who was recently sentenced to 19 1/2 years in prison. China has experimented with (but for now backed away from) mandating the installation of client-side filtering software that would break all existing circumvention tools. We even know of many examples of governments shutting down the Internet entirely, either in specific regions or in the entire country, during times of political crisis. Circumvention tools do not help provide access to any of the content that is blocked by these other sorts of control.
Probably most importantly, governments use traditional knock-on-door enforcement by government agents to control locally published data. In many filtered countries, the large majority of traffic is to sites hosted locally. For example, about 95 percent of web page views in China are for websites hosted locally in China. This high level of interest in locally hosted content makes sense -- Chinese people like to read content written in Chinese by other Chinese people about local Chinese topics. But the organizations and individuals publishing these sites are subject to government agents knocking on doors to enforce content regulations -- to fine or shut down businesses or throw individuals in prison.
In a world in which users are overwhelmingly interested in viewing and creating content locally and in which all of that local content is subject to this traditional knock-on-door regulation, the benefit of circumvention tools for these users (and therefore their role in promoting Internet freedom) is limited. And so, in this sense it is not entirely surprising that only a small percentage of filtered users go to the trouble of using them.
Mitigating authoritarian Internet control by governments will likely require a much broader approach that focuses not only on circumvention tool development, but also on the much harder solutions to the various other kinds of Internet control.
Hal Roberts is a fellow at the Berkman Center for Internet & Society at Harvard University. The views expressed in this commentary are the author's own and do not necessarily reflect those of RFE/RL.
Examples of these circumvention tools include publicly/charitably funded tools like Ultrasurf, Freegate, and Tor that have been well covered in the media. But they also include mostly fee-based virtual private network (VPN) services; free, but anonymous protocol-based proxies; and mostly advertising-supported, simple web proxies. All these tools share the same basic mechanism -- they allow access to a blocked site by ferrying requests for the blocked site through an (ideally) unblocked intermediary machine.
The Berkman Center for Internet & Society at Harvard University recently issued a report examining the use of these tools. We found that, at most, only about 3 percent of people in filtering countries -- and likely many fewer -- use circumvention tools. Of the three types of tools whose use we could measure -- Ultrasurf, Freegate, and Tor; VPN services; and simple web proxies -- we found that simple web proxies were by far the most widely used. And we found that when users in filtered countries search for these tools, they overwhelmingly search for generic terms -- mostly the English word "proxy" -- that return lists of simple web proxies.
We don't know for sure what any circumvention tool or class of circumvention tools is used for, but we have lots of circumstantial evidence that simple web proxies are generally used for nonpolitical purposes like browsing social media, pornography, and gambling sites. So even the 3 percent number may overstate the use of such tools to access the politically relevant material that most people think of when they think of promoting Internet freedom in authoritarian countries.
Many press reports have portrayed circumvention tools as a clear path toward Internet (and possibly even broader political) freedom. But our findings challenge that simple story with simple numbers. To be clear, even at this proportionally small level of usage, these tools are doing a huge amount of absolute good by offering the opportunity to fulfill the basic human right to freedom of expression for as many as tens of millions of people. And the impact of the small proportion of people using these tools may be magnified if we assume those circumvention users distribute otherwise censored information out to their communities.
But even so, it is hard to argue that these tools are having revolutionary impact with such little proportional usage or even that there is a simple way to scale their usage dramatically enough to achieve such impact.
Access Versus Inconvenience
The core question that remains unanswered is why so few people use these tools. We published an evaluation of circumvention tools in 2007 that found that most of these tools work, in the broad sense that they enable users to access sites that would otherwise be blocked. But we also found that the performance of the tools ranged from slow to painfully slow, that the usability of most of the tools could be improved, and that there were significant security problems with several of the tools. These issues, particularly the slow performance of the tools, may be the main factors limiting their usage.
But a bigger limiting factor may be that users simply do not want to access filtered content badly enough to outweigh the inconvenience of using the tools.
This does not necessarily mean that users do not want to circumvent Internet control. It may mean that circumvention tools are not sufficient to provide access to the kinds of content that people want. Internet control involves much more than simply filtering some foreign sites.
Distributed denial of service attacks commonly take down controversial sites by flooding them with traffic, often from thousands of compromised computers. Many online journalists-bloggers have been arrested by their home countries, including prominent Iranian blogger Hossein Derakhshan, who was recently sentenced to 19 1/2 years in prison. China has experimented with (but for now backed away from) mandating the installation of client-side filtering software that would break all existing circumvention tools. We even know of many examples of governments shutting down the Internet entirely, either in specific regions or in the entire country, during times of political crisis. Circumvention tools do not help provide access to any of the content that is blocked by these other sorts of control.
Probably most importantly, governments use traditional knock-on-door enforcement by government agents to control locally published data. In many filtered countries, the large majority of traffic is to sites hosted locally. For example, about 95 percent of web page views in China are for websites hosted locally in China. This high level of interest in locally hosted content makes sense -- Chinese people like to read content written in Chinese by other Chinese people about local Chinese topics. But the organizations and individuals publishing these sites are subject to government agents knocking on doors to enforce content regulations -- to fine or shut down businesses or throw individuals in prison.
In a world in which users are overwhelmingly interested in viewing and creating content locally and in which all of that local content is subject to this traditional knock-on-door regulation, the benefit of circumvention tools for these users (and therefore their role in promoting Internet freedom) is limited. And so, in this sense it is not entirely surprising that only a small percentage of filtered users go to the trouble of using them.
Mitigating authoritarian Internet control by governments will likely require a much broader approach that focuses not only on circumvention tool development, but also on the much harder solutions to the various other kinds of Internet control.
Hal Roberts is a fellow at the Berkman Center for Internet & Society at Harvard University. The views expressed in this commentary are the author's own and do not necessarily reflect those of RFE/RL.
