Smartfilter: Miscategorization and Filtering in Saudi Arabia and UAE

November 28, 2013

Tagged: , , ,

Categories: Reports and Briefings, Research News

By Bennett Haselton

What do the Trinity Davison Lutheran Church, the Filipino American Women’s Network, the Tuscon Jazz Institute, the Sacramento Police Activities League, the Pan Iranist Party of Iran, and the Salvation Army of Houston, Texas have in common?

Their websites are all blocked for Internet users in Saudi Arabia, and for most users in the United Arab Emirates. This is not because the governments of these countries disapprove of these websites. Rather, these sites are blocked by ISPs because of their use of the U.S.-made Internet blocking program “Smartfilter”, a program that often mistakenly categorizes these sites and many others as “pornography.”

The Smartfilter web blocking product has existed since 1996. In 2008, McAfee acquired Secure Computing, the company that manufactured Smartfilter. McAfee runs a support site at http://www.trustedsource.org/ that allows users to look up how sites are classified in Smartfilter’s rating system. For example, the screen captures here, here and here show http://www.trinitydavison.org/, http://www.paniranist.org/, and http://www.sahouston.org/ categorized as “pornography.” The lookup form does not give any reason why the sites would have been categorized as such, and none of these or other sites in our report contained any content that could be viewed as pornographic. The lookup form does, however, provide the option to email feedback to McAfee asking for a site’s classification to be changed. We did not submit any requests to change the classifications of any of the sites we found to be blocked.

trinity-miscat
Figure 1 – An example of a site miscategorized as pornography by Smartfilter

A previous OpenNet Initiative report, “West Censoring East”, focused on the issue of American and Canadian software companies selling Internet filtering programs to government censors in the Middle East and North Africa. Countries like Saudi Arabia, UAE, Kuwait, Bahrain, Oman, Tunisia, and Iran have used Smartfilter at various times for nationwide Internet censorship. Currently, Smartfilter is only known to be used in Saudi Arabia and in UAE for customers of Etisalat, one of the UAE’s two main Internet providers. (The other Internet provider in the UAE, a telecom company known as “du”, uses the Canadian-made Netsweeper filter.)

McAfee maintains a centralized database of site categorizations. All Smartfilter customers download periodic updates to the database. If a site is newly categorized or re-categorized in Smartfilter’s database, these updates will be pushed out to customers usually within 24 hours. The Citizen Lab confirmed in summer 2013 that Smartfilter was still in use in Saudi Arabia and on UAE Etisalat by creating new websites and submitting them to Smartfilter to be blocked. The Citizen Lab verified that the sites became blocked in Saudi Arabia and on UAE Etisalat a few days later, while a control group of similar websites created at the same time did not become blocked.

Network administrators can choose the categories of sites to be blocked by Smartfilter. Our current tests indicate that Smartfilter is being used in Saudi Arabia and on UAE Etisalat with the “pornography” category blocked. We tentatively believe that the “nudity” category is also blocked in both Saudi Arabia and UAE since two sites in our sample—http://www.antgallery.org and http://www.midwivesofmaine.org—were classified under “nudity” (but not “pornography”) on trustedsource.org and were also found to be blocked while testing in both countries. With a sample size of just two sites, however, we decline to draw any definitive conclusions.

Our results also suggest that other related SmartFilter categories—such as “incidental nudity”, “sexual materials,” and “provocative attire”—are probably not blocked, with the exception of “incidental nudity,” which was is blocked in UAE but not Saudi Arabia (see below). For each of those categories, we found sites which trustedsource.org classified in that category, but  were still available in both countries:

 

 

 

 

 

 

Using an ad hoc combination of domain name sampling and automated web spidering, we found over 100 sites that were mistakenly classified as “pornography” by Smartfilter, and hence were blocked in Saudi Arabia and on UAE Etisalat as of writing. As our testing could only sample a small portion of the Web, our list should not be taken to represent more than a  fraction of sites misclassified by Smartfilter.

For most domain names, it was impossible to discern why the sites may have been classified as “pornography.” In some cases, how,  probable reasons for blocking could be reasonably inferred. Some blocked sites were using domain names that sounded as if they might have previously been registered as pornographic websites. The Internet Archive’s “Wayback Machine” can be used to check whether any such domains might have hosted pornographic content in the past. Through the site, we checked the history of six domains whose names sounded as if they could plausibly have been used to host pornography: bearyouth.org, bigmatt.com, caseydonovan.org, easyaction.com, gerl.org, and theteacherspet.com. Of these, we found that bearyouth.org, easyaction.com and gerl.org all hosted adult content at some point in the past when the domains had different owners. It is reasonable to assume that most of the other domain names in our sample—such as 14thbrooklyn.org (The Society for the Historical Preservation of the Fourteenth Brooklyn) and FreeInTheAbingtons.org (a guide to low-cost recreational activities in Northeastern Pennsylvania)—are unique enough that it is unlikely they were previously owned by other parties, much less by pornographic websites.

Some sites were likely blocked due to the presence of a “trigger word” appearing in their domain name. The sites http://www.AgainstPornography.org and http://www.SearchingForMySpermDonorFather.org were presumably blocked because of the occurrence of the words “porn” and “sperm.” ScunthorpeGilbertAndSullivan.org (a Gilbert and Sullivan Society based in Scunthorpe, England) was presumably blocked due to the presence of the word “cunt” in the domain name. (This particular word filtering error is so common that it is known as the “Scunthorpe problem”.) BigHornBasinSFW.org, the website for the Big Horn Basin, Wyoming chapter of Sportsmen for Fish & Wildlife, was presumably blocked because of the inclusion of the string “nsfw” (“not suitable for work”), an acronym commonly used to indicate pornographic or otherwise workplace-inappropriate content.

The following is a sample of websites that we found to be blocked in Saudi Arabia and UAE Etisalat as a result of the site being misclassified by Smartfilter as pornography. Testing in Saudi Arabia was performed over October 16-17, 2013 on a VPN connection using HideMyAss Pro software, connected through an IP address registered to “National Computer Systems Co.” in Riyadh. We re-tested on November 27th using a connection through the King Abdulaziz City For Science And Technology, and found the sites listed below were all still blocked. Testing from within the UAE was performed on November 3rd, 2013 by an anonymous volunteer. When a blocked site was accessed from within Saudi Arabia, it would display this page indicating that the site had been blocked. When a blocked site was accessed on UAE Etisalat, it would bring up this error page, displaying the message:

uae-computergarage

It is not clear whether this page is intended to mislead users into thinking the website is temporarily down (as opposed to being blocked by the ISP). While the error message seems to indicate a temporary failure, the title of the error page (clearly visible in the screen capture linked above) is listed as “Blocked URL”. It is also not clear why the error message is displayed in English.

In our list of sites below, each entry is accompanied by a screen capture of the site being categorized on McAfee’s trustedsource.org site as pornography, along with a screen capture of the site being blocked from within Saudi Arabia. We did not collect screen captures of blocking within UAE, other than the single screen capture linked above, due to the time constraints on our volunteer tester preventing the collection of every site. Smartfilter might change its classifications of these sites after this post is published; the most current classifications can be looked up using McAfee’s form at http://trustedsource.org/en/feedback/url?action=checksingle.

For each list entry, the letters “ts” are linked to a screen capture of the site’s categorization on McAfee’s trustedsource.org site, and the letters “sa” are linked to a screen capture of the site being blocked in Saudi Arabia.

Most reports on Internet censorship in countries that practice nationwide Internet filtering have focused on the censorship of political content, adult websites, social networking, and media sharing sites such as Facebook and YouTube. This focus is logical since the blocking of these websites is likely to have a far greater impact on users than the blocking of relatively obscure sites. For the sake of thoroughness, however, we should not ignore the large numbers of sites that are blocked in censored countries simply because they have been misclassified by the manufacturer of the blocking program being used.

The author thanks Adam Senft and Helmi Noman of Citizen Lab for helping to collect data for this report, and thanks the anonymous tester in the United Arab Emirates who helped to test the accessibility of websites through Etisalat.

Bookmark and Share

Post a Comment

Your email is never shared. Required fields are marked *

*
*