Shifting Borders Steven J. Murdoch &…

                   Shifting Borders
              Steven J. Murdoch & Ross Anderson


         The Internet once promised a brave new borderless world,
         but the reality is more complex


    In A Declaration of the Independence of Cyberspace, John Perry Barlow called
for communities built around the Internet to be independent of national gov-
ernments and borders: a Utopian ideal that has failed to materialise. The In-
ternet does have borders, for similar reasons that national boundaries exist:
they ease administration, permit collective defence and can be founded in
culture.
    While it is true that Internet borders do not have to be the same as polit-
ical boundaries, the two have naturally mirrored each other. This is hardly
a surprise since the Internet was built on the infrastructure of telecommuni-
cations companies, often controlled or regulated by nation states.
    Even during the early 1980s, during the infancy of public networks and
before what we now call the Internet had left the laboratory, this pattern was
established. Although bulletin board services were run by enthusiasts rather
than commercial interests, and driven by ideals similar to Barlow's, locality
mattered. Since the communication hubs were accessed by telephone, each
was popular only within the local call area.
    Communication across borders was possible, but initially limited. Later,
as the Internet became commonplace, global communication was seen as
the norm, but international traffic was still charged at a higher rate and so
discouraged. Nonetheless, the Internet has changed the nature of borders,
and particularly their power to control the flow of ideas. The traditional
censorship methodologies that nations built over centuries do not work par-
ticularly well on the Internet.
    National borders are becoming more porous to information. Organisa-
tions previously had to publish through posters, leaflets or radio within their
home country, risking material being banned. Now they can host their ma-
terial overseas. Even where there is no initial concern about censorship, a lot
of hosting is done for cost and efficiency reasons in the US. When a dispute
arises, it is judged by the more liberal values there.
    Since this has become apparent, many countries have implemented mea-
sures to restrict access to content that they declare to be illegal or immoral,

Originally published in: Index on Censorship, Volume 36, Issue 4, pages 156­159,
November 2007. DOI: 10.1080/03064220701740525
                                                                              2


despite being acceptable in the country where it is hosted. The common
mechanism is to block communications with the computer on which the
content is stored. Information coming from or going to a banned Internet
address is misdirected or dropped entirely. The technology for performing
the blocking varies by country, as does the nature of the material filtered, but
the overall concept remains the same.
    This type of blocking is fairly easy to circumvent for those with enough
persistence or technical knowledge, but censors are not aiming for the unat-
tainable goal of completely censoring forbidden information. Instead, it is
sufficient to block most of the banned content from most of the people. Com-
bined with surveillance and threats of punishment, this is felt sufficient to
encourage self-censorship, and thus to halt or slow the social changes the
censor fears.
    The strengths and vulnerabilities of this style of blocking largely follow
from the `end-to-end principle', one of the central tenets in the design of the
Internet. This calls for the intelligence of the network to be moved to the
edges, rather than being distributed within it. The idea was that by con-
structing a dumb network, with clever endpoints, new Internet applications
could be developed without altering the core of the network.
    When information is pushed to the edges, blocking based on the address
of the endpoints is fairly effective. However, while the end-to-end principle
has been maintained in the technical details of how computers communi-
cate over the Internet, when it comes to content, the end-to-end principle is
increasingly breaking down, requiring the methods of blocking to become
more complex. Community-based sites, known as Web 2.0, are one exam-
ple of applications that shift the intelligence away from the end-user's com-
puter. The code and data are kept on central servers, and the user requires
only a web browser. This idea is not new: Geocities, founded in 1994, al-
lowed people to publish content without requiring them to manage their
own server. What has changed is the popularity and diversity of applica-
tions that are available.
    Social networking sites like Facebook, blogging platforms like Blogger,
and other community publishing venues like Wikipedia can all have their
addresses blocked, rendering them inaccessible within a country. But cen-
sors are faced with the choice of blocking all of the site or none of it. All
the disparate communities sharing the same site are behind the same ad-
dress, and probably only a tiny fraction are the target of censorship. This is
a very different situation from German ISPs blocking a handful of websites
dedicated to promoting neo-Nazism.
    The `collateral damage' of blocking a huge website is significant. For ex-
ample in 2006, when Google's Blogger and Yahoo Geocities were blocked
in India, large online protests were organised. The Chinese blocking of
Wikipedia was unpopular; the site was often unblocked only to be later
blocked again. Even users of the sites who had no interest in the material
                                                                              3


the censors disliked were frustrated by the censorship. Indeed, blocking a
major service like Google or Skype can do actual economic damage to a de-
veloping country.
     The tensions surrounding censorship of community sites are unlikely to
change, since `network effects' lead to popular sites becoming even more
popular. This phenomenon was described in Metcalfe's law, which states
that the value of a network grows with the square of the number of mem-
bers. Network effects lead to industry consolidation, as with Google buying
their video product's competitor YouTube, and Yahoo! buying Flickr. It also
leads to `winner takes all' behaviour-people will want to join the most pop-
ular community while the second-best languishes in obscurity.
     But if address-based blocking is no longer feasible, what are the alter-
natives? One is to block by keyword, as is already done by China. Here,
regardless of the address of a site, if a banned word is used the material is
blocked. However, this approach requires more expensive equipment and is
not reliable. Websites can also prevent this blocking, whether intentionally
or unintentionally, by obscuring the information as it is transmitted. Then
the censor is back in the previous situation: faced with a large site containing
a small amount of offensive material, he has to block it all, or not at all.
     That leaves countries with the option of asking the site operator to re-
move the offending content. Even if the content is legal in the country where
it is hosted, the country in which the content is delivered still has leverage.
Thailand has on occasion blocked all of YouTube.
     Countries can also censor content if the provider has a physical pres-
ence there, regardless of where the content is actually hosted. In 2000, the
US-based company Yahoo! lost a court case in France over whether they
were required to prevent the sale of Nazi memorabilia, and earlier this year
a Brazilian court ordered YouTube to remove an unauthorised video of a
celebrity.
     Even if the provider does not have a base in the offended country, a cen-
sor can move against the provider's suppliers, and in particular those who
purchase advertising. This was the UK government's strategy in the 1960s
to close down pirate radio stations that operated from ships outside Britain's
territorial waters. This is an interesting case as the censor's goal was not an
issue of faith or morals, but simply maintaining a level of state control of
broadcast media that is nowadays considered excessive almost everywhere.
     As community-based sites grow, network effects will cause their value to
increase and so countries will be more hesitant in blocking them. However,
this trend might not continue indefinitely. Industry consolidation means
that a country could `punish' an operator not by blocking the offending site-
if that would cause too much collateral damage-but by blocking a different
site owned by the same company. YouTube might be the dominant video-
sharing site, but it is owned by Google, which plays in the more competitive
search-engine advertising market. Even in countries where YouTube does
                                                                            4


not have a legal presence, Google might be forced to remove its videos by
threats to block its advertising.
    All of a sudden, computer industry structures matter when it comes to
censorship, where previously we had to worry about governments, and (to
a much lesser extent) about the media industry. Important factors include
whether controversial matter is hosted on a dominant site or niche interest.
Corporate boundaries, rather than national ones, define how information
will be restricted.
    While John Perry Barlow's ideals of a borderless Internet have not come
to pass, the barriers that have been created are different. Hopefully they are
lower; it's a lot easier to set up companies than nation states. And change
is particularly rapid in the computer business ­ so we shall see where these
borders will move next.

Steven J. Murdoch is a researcher at the University of Cambridge. He is
currently developing censorship resistant technologies for the Tor Project
Ross Anderson is Professor of Security Engineering at Cambridge. He was
a pioneer of peer-to-peer systems and of security economics, and chairs the
Foundation for Information Policy Research