Technology Review: Blogs: Guest Blog: Why Twitter Doesn't Mean the End of Iranian Censorship Use of proxies and Twitter is rising in Iran, but that doesn't automatically mean victory against state censorship. By Hal Roberts As a Fellow at the Berkman Center for Internet & Society at Harvard University, Hal Roberts performs primary research into global Internet filtering. Here he offers his perspective on the post-election Internet crackdown in Iran. Amid post-election protests in Iran, the government has apparently increased its filtering of sites, such as Twitter and Facebook, that host potentially offensive (to the government) content--and even reportedly turned off for a short period the Internet connection to the rest of the world. A question simple to ask--but difficult to answer--is whether Iranians are successfully bypassing the filtering through proxies or other filtering circumvention tools. Academic research has established for years that the government of Iran closely filters its Internet connections, blocking sites that it does not like (mostly pornographic ones, but political and religious sites as well). The government of Iran can do this easily because virtually all traffic flows through a single government-controlled ISP. (In fact, Iran for years used McAfee SmartFilter, a product of a U.S. company, to perform this filtering, but it uses its own homegrown filtering tools now.) Some users combat this filtering by employing proxies, routing their traffic through a machine outside of Iran so that the Iran filter sees only traffic to that proxy, effectively exchanging Iran's control of the network for the proxy's control of its network. Iran responds by blocking these proxies as it finds them, and these proxy users respond by continually looking for new, unblocked proxies or by using tools like UltraSurf that do the work of filtering out government interference themselves. Data about proxy use is naturally hard to find (the point is to hide the users' usage), but our best data indicate that interest in using proxies has increased substantially over the past year and has doubled in the past week. But such use is confined to a small portion of Iranian Internet users; it's in the low single percentage points. Google searches for "proxy," for instance, remain orders of magnitude less popular than searches for "election." Likewise, a steady flow of information about the protests has come out of Twitter, but the number of Iranian users actually Twittering seems to be a tiny portion of Iranians. As far as we can tell, the Iranian government has done a pretty good job of blocking its citizens' Web requests to sites that it does not want them to see, including during the current crisis. But new technologies make the battle over filtering harder to judge. Even though the government has reportedly blocked Twitter.com, a defining attribute of Twitter is that it is an open system in that it allows a wide diversity of external tools and sites to read from and write to its service through its programming interface. Jonathan Zittrain and John Palfrey point out that as content is divorced from delivery through such open systems, blocking, for example, Twitter-as-a-network-system much harder than simply blocking Twitter the site, since there are dozens of tools and sites that directly read and write the Twitter data stream. And as with other recent global crises, the widespread use of distributed denial of service (DDoS) attacks has made it possible to filter a site by flooding it with so much data that it can no longer respond to legitimate users, rendering proxies useless for those sites. The tools to launch DDoS attacks, including simple Twitter campaigns to overload a list of sites, have become easily available, so both pro-government and protest actors have directed these attacks at each other's sites. But the technical issue of whether a given site returns a response for a given set of people captures only one small part of the larger problem of determining who controls the flows of information on the Internet and through media and social networks in general. A fuller approach to the problem is to think about those flows of information and how they are being filtered, by social and political as well as technical means. We should ask, for example, whether the information from the core group of proxy/Twitter users is filtering out to the wider Iranian and global communities, how it is flowing to and through those communities, and what effect the information is having as it filters out. The answers to those questions are impossible to determine in real time from the outside, given the chaos and confusion of the situation. As with the protests, time and perspective will tell.