to produce the blockpage. ISPs in Iran, Singapore, Thailand, and Yemen all use redirection to
a blockpage. Uzbekistan uses redirection but instead of redirecting to a blockpage the filters
send users to Microsoft’s search engine at www.live.com, suggesting that the government
wishes to conceal that fact that blocking has taken place.
There are thus various degrees of transparency in Internet filtering. Where blockpages are
used, it is clearly apparent to users when a requested Web site has been intentionally
blocked. Other countries give no indication that a Web site is blocked. In some cases, this is
a function of the blocking technique being used. Some countries, such as Tunisia and Uzbekistan, appear to deliberately disguise the fact that they are filtering Internet content, going a
step farther to conceal filtering activity beyond the failure to inform users that they are being
filtered.
Another subset of countries, including Bahrain and United Arab Emirates, employ a hybrid
strategy, indicating clearly to users that certain sites are blocked while obscuring the blocking
of other sites behind the uncertainty of connection errors that could have numerous other
explanations. In Bahrain, users normally receive a blockpage. However, for the specific site
www.vob.org, Bahrain uses DNS tampering that results in an error. In United Arab Emirates
all blocked sites with the exception of www.skype.com returned a blockpage. There is an apparent two-tiered system in place. They are willing to go on the record as blocking some sites,
and not for others.
Providing a blockpage informing a user that their choice of Web site is not available by
action of the government is still short of providing a rationale for the blocking of that particular
site, or providing a means for appealing this decision. Very few countries go this far. A small
group of countries, including Saudi Arabia, Oman, and United Arab Emirates, and some ISPs
in Iran, allow Internet users to write to authorities to register a complaint that a given Web site
has been blocked erroneously.
Centralized filtering regimes require all Internet traffic to pass through the same filters. This
results in a consistent view of the Internet for users within the country; all users experience the
same degree of filtering. This is most commonly implemented at the international gateway.
When filtering is delegated to the ISP level, and hence decentralized, there may be significant
differences among ISPs regarding the filtering techniques used and the content that is filtered.
In this case, access to Web sites may vary substantially depending on the blocking choices of
individual ISPs. (Table 1.4 presents the use of centralized and/or decentralized filtering strategies across the countries in the study, and the resulting consistency in filtering within each
country.) In Iran there is considerable variation in the blocking among ISPs. For example, one
ISP blocks considerably less political content than the other six ISPs tested. Only one ISP out
of the five tested in Azerbaijan, AzNet, blocks access to a considerable amount of social content, most of which is pornographic, while the others block access to only a single IP address.
In Myanmar, there is substantial variation in the filtering between the two ISPs tested. One filters much more pornography, while the other blocks a significantly greater portion of politically
oriented Web sites. In the United Arab Emirates, an ISP that serves primarily the free-trade
16
Robert Faris and Nart Villeneuve
a blockpage. Uzbekistan uses redirection but instead of redirecting to a blockpage the filters
send users to Microsoft’s search engine at www.live.com, suggesting that the government
wishes to conceal that fact that blocking has taken place.
There are thus various degrees of transparency in Internet filtering. Where blockpages are
used, it is clearly apparent to users when a requested Web site has been intentionally
blocked. Other countries give no indication that a Web site is blocked. In some cases, this is
a function of the blocking technique being used. Some countries, such as Tunisia and Uzbekistan, appear to deliberately disguise the fact that they are filtering Internet content, going a
step farther to conceal filtering activity beyond the failure to inform users that they are being
filtered.
Another subset of countries, including Bahrain and United Arab Emirates, employ a hybrid
strategy, indicating clearly to users that certain sites are blocked while obscuring the blocking
of other sites behind the uncertainty of connection errors that could have numerous other
explanations. In Bahrain, users normally receive a blockpage. However, for the specific site
www.vob.org, Bahrain uses DNS tampering that results in an error. In United Arab Emirates
all blocked sites with the exception of www.skype.com returned a blockpage. There is an apparent two-tiered system in place. They are willing to go on the record as blocking some sites,
and not for others.
Providing a blockpage informing a user that their choice of Web site is not available by
action of the government is still short of providing a rationale for the blocking of that particular
site, or providing a means for appealing this decision. Very few countries go this far. A small
group of countries, including Saudi Arabia, Oman, and United Arab Emirates, and some ISPs
in Iran, allow Internet users to write to authorities to register a complaint that a given Web site
has been blocked erroneously.
Centralized filtering regimes require all Internet traffic to pass through the same filters. This
results in a consistent view of the Internet for users within the country; all users experience the
same degree of filtering. This is most commonly implemented at the international gateway.
When filtering is delegated to the ISP level, and hence decentralized, there may be significant
differences among ISPs regarding the filtering techniques used and the content that is filtered.
In this case, access to Web sites may vary substantially depending on the blocking choices of
individual ISPs. (Table 1.4 presents the use of centralized and/or decentralized filtering strategies across the countries in the study, and the resulting consistency in filtering within each
country.) In Iran there is considerable variation in the blocking among ISPs. For example, one
ISP blocks considerably less political content than the other six ISPs tested. Only one ISP out
of the five tested in Azerbaijan, AzNet, blocks access to a considerable amount of social content, most of which is pornographic, while the others block access to only a single IP address.
In Myanmar, there is substantial variation in the filtering between the two ISPs tested. One filters much more pornography, while the other blocks a significantly greater portion of politically
oriented Web sites. In the United Arab Emirates, an ISP that serves primarily the free-trade
16
Robert Faris and Nart Villeneuve