The Internet has become the new battle ground between authoritarian regimes and ordinary individuals who want unimpeded access to information. The immense popularity of online activism and citizen journalism enabled by social media has instigated state level players to partially or completely block access to the Internet. In return, individuals and organizations have been employing various anti-censorship tools to circumvent these restrictions. In this paper, we claim that censorship is futile as not only has it been ineffective in restricting access, it has also had the side effect of popularising blocked content. Using data from Alexa Web rankings, Google Trends, and YouTube Statistics, we quantify the ineffectiveness of state level censorship in Pakistan and Turkey and highlight the emergence of the Streisand Effect. We hope that our findings will, a) prove to governments and other players the futility of their actions, and b) aid citizens around the world in using legal measures to counteract censorship by showing its ineffectiveness.
2. Setting the stage
5. Conclusion and future work
The popularity of the Internet is a double-edged sword: it opens up a world of information and promotes free speech but at the same time, this popularity incites authoritarian regimes and other actors to repress access to it. As a result, at present more than 60 countries around the world censor the Internet in one form or another (Burnett and Feamster, 2013). These forms include all-out blocking (Dainotti, et al., 2011), partial blocking at different levels (Nabi, 2013; Verkamp and Gupta, 2012; Aryan, et al., 2013), performance degradation/throttling (Anderson, 2013), and content manipulation (Bamman, et al., 2012; Verkamp and Gupta, 2013). To stem this suppression tide, researchers and activists are actively developing an arsenal of various anti-censorship tools (Houmansadr, et al., 2011; Wustrow, et al., 2011; Dingledine, et al., 2004; Feamster, et al., 2002; Burnett, et al., 2010; Clarke, et al., 2002; Invernizzi, et al., 2013). Cognizant of this, censors are trying to stay ahead of the curve (Winter and Lindskog, 2012; Schuchard, et al., 2012; Geddes, et al., 2013). This has led to a cat and mouse game between censors and anticensorship practitioners, with no end in sight, similar to the struggle between encryption and decryption (Barto, 2013).
Fortunately, this tug of war has not dampened the desire of users around the world to gain unimpeded access to the Internet. On the one hand, they are readily using various methods to circumvent restrictions (Nabi, 2013) and on the other they are helping others to do the same through ingenious mechanisms, such as putting up addresses of open DNS servers as wall graffiti (Zmijewski, 2014). In a similar vein, in spite of restrictions and the risk associated with bypassing them, the usage of social media and blogs has been escalating (Caren and Türkoğlu, 2014). At the other end of the spectrum, all of this online activism has induced some governments to pass “Internet constitutions” to enshrine freedom of expression, Internet neutrality, and online privacy (Mari, 2014).
Essentially, at the very core, the goal of a censor is to restrict access to content that is deemed detrimental to a certain vested interest. This blackout often backfires and causes the content to go viral following the Streisand Effect (Greenberg, 2007). The Streisand Effect can be defined as the inadvertent popularity of any material as a result of its suppression (Details in §2.1). In recent years a number of instances of this phenomenon have been recorded. These include the swell in the number of Twitter users after the Web site was blocked in Turkey (Woollacott, 2014), the popularity of the online activism portal Avaaz’s posters to ban South Africa’s lion bone trade (Solon, 2013a), and the re-posting of the “Station hertzienne militaire de Pierre-sur-Haute” entry on Wikipedia after the latter was forced by French Intelligence to delete it (Geuss, 2013). While the Internet is replete with examples of this phenomenon, the evidence is mostly anecdotal and not in the context of state-level censorship.
To remedy this, we present, to the best of our knowledge, the first formal study of the Streisand Effect in state-level censorship. With Pakistan and Turkey as a case study, using data from Google Trends, YouTube Video Statistics, and Alexa Web rankings, we show that censorship has, a) not affected the ranking of Web sites such as YouTube, and b) it has had the side effect of causing restricted content to go viral. The goal of this paper is two-fold: to dissect the Streisand Effect in the context of state-level censorship and also, possibly more importantly, to aid citizens around the world in using legal measures to counteract censorship. For instance, for the past two years activists have been fighting a case in a provincial High Court in Pakistan to unblock YouTube. The ineffectiveness of the ban has even prodded the lead judge to remark at one point: “What the government is saying is a joke. Every child is accessing YouTube. It’s a fraud against the people of Pakistan.” (Shah, 2014) Unfortunately, most of the information about censorship circumvention and the Streisand Effect is circumstantial and anecdotal, to which the law is blind. Therefore, a formal study with concrete numbers can fortify the case for the futility of Internet censorship around the world. It has already proven effective in countries like Turkey where the ban on Twitter was lifted recently after the constitutional court ruled it as illegal (Reuters, 2014).
The rest of the paper is organized as follows. In §2 we give background information about the Streisand Effect and the various data sources used in the paper. Part 3 presents our findings. We discuss the implications of the findings in §4. Part 5 concludes the paper and also discusses future directions.
2. Setting the stage
In this section, we give the reader background information about the Streisand Effect, Alexa Web rankings, Google Trends, and YouTube Video Statistics. In addition, we give details of Internet censorship in our two target countries: Pakistan and Turkey.
2.1. The Streisand Effect
The Streisand Effect is named after the singer Barbara Streisand who in 2003 unsuccessfully tried to have an aerial picture of her Malibu beach house taken down from the Web site of an environmental activist (Greenberg, 2007). She subsequently lost the US$50 million lawsuit but the ensuing media coverage brought more than a million visitors to the aforementioned site. Since then, the Streisand Effect has become synonymous with unintentional virality of any information, online or otherwise, as a consequence of any attempt to censor, suppress, and/or conceal it. This phenomenon has manifested itself in a diverse array of settings. Examples include unsuccessful lawsuits and injunctions (Solon, 2013b; Jacobson, 2009), personal liberty (Cacciottolo, 2012), and even scientific research (Heussner, 2008). Over the years, the effect has been leveraged by individuals and organizations alike to repel censorship. For instance, the American Library Association and others have been organizing a Banned Books week since 1982 to raise awareness about banned and restricted books (Doyle, 2010). Similarly, attorneys frequently advise their clients not to pursue libel cases that can potentially raise interest in the issue rather than subdue it (Legal Eagle, 2012). Above all, the Streisand Effect has been successfully used by numerous individuals around the world to take the fight to government mandated censorship (Economist blog standard, 2008; Woollacott, 2014; Greenberg, 2007). It is important to highlight that while the positive effects of the effect are an effective tool against censorship, it is by no means a panacea. To counteract online censorship around the world, we still require an ecosystem of activists and tools. We discuss this in detail in §4.2.
With a population size almost half of South America and an almost comparable number of absolute Internet users (Nabi, 2013; Internet World Stats, 2011), Pakistan has experienced Internet censorship in many forms over the years. Popular Web sites such as Google, Facebook, Flickr, Wikipedia, and YouTube have borne the brunt of restrictions in recent years. At times, the side effects of this censorship have affected connectivity outside the country as well (Brown, 2010). From a technical perspective, Web sites are blocked at both the DNS level as well as the HTTP level (Nabi, 2013). These restrictions are imposed through technology from Netsweeper at the IXP level (Citizen Lab, 2013). In addition, citizens are also spied upon through FinSpy (Marquis-Boire, et al., 2013). Fortunately, Pakistan also has an extremely vibrant social activism community which has legally challenged Internet censorship and surveillance within the country (Shah, 2014; Malik, 2014). Table 1 shows three major censorship events from the country under consideration in this paper.
Table 1: Censorship events in Pakistan. Date Event ID Description 22 February 2008 PK1 YouTube censored in retaliation to the movie “Fitna” 19 May 2010 PK2 Facebook, YouTube, Flickr, and Wikipedia blocked due to “Everybody Draw Muhammad Day” 17 September 2012 PK3 YouTube blocked in reaction to another controversial movie “Innocence of Muslims”
An Internet penetration of close to 49 percent makes Turkey one of the most connected countries in the world (eMarketer, 2013). It also has the highest Twitter penetration in the world (Baydar, 2014). In addition, 72 percent of the Internet users access newspapers and magazines online (Turkish Statistical Institute, 2012) enabling them to be politically informed. Unfortunately in the last few years, Turkey has gone from relatively benign online censorship to “one of the world’s most determined Internet censors” (Parkinson, et al., 2014). In fact, the Turkish Parliament recently passed legislation that allows the authorities to block any Web site while bypassing the courts (BBC News, 2014). In addition, Internet service providers are bound to keep tabs on a user’s online activity for two years for use by the authorities. Similar to Pakistan, Turkey also has a very active online community whose activism recently prompted its top court to deem the ban on Twitter as illegal (Reuters, 2014). Four main censorship events from Turkey are listed in Table 2.
Table 2: Censorship events in Turkey. Date Event ID Description 8 May 2008 TR1 YouTube blocked due to videos offensive to Mustafa Kemal Ataturk 27 January 2014 TR2 SoundCloud blocked to stop access to leaked audio tapes of the Turkish Prime Minister and his main political rival in a possible graft case 20 March 2014 TR3 Twitter blocked to suppress access to leaked recordings allegedly implicating the Turkish Prime Minister and his inner circle in various corruption scandals 27 March 2014 TR4 YouTube censored to block access to audio recordings of government officials discussing a possible military strike inside Syria
2.4. Alexa Web rankings
Alexa  is a Web analytics company that provides traffic analysis and rankings for top level domains. Traffic rankings — which are updated daily — are calculated by using crowd-sourced data from its browser toolbar. A Web site’s rank is a function of the number of unique visitors and page views. Alexa’s Web portal displays rankings via three views: 1) global; 2) country-wise; and, 3) categorywise. In addition, Alexa also exposes a Web Information Service API that allows the user to query historical data through Amazon Web Services.
2.5. Google Trends
Google Trends  is a time series index of search term volume entered into Google in a geographic region (Choi and Varian, 2012). The index is calculated by normalizing the total query volume for a specific term by the total number of queries in that region. The maximum value is 100 and indexes can be searched back to 1 January 2004. Google Trends have found wide traction for trend prediction, from economic indicators (Choi and Varian, 2012) to disease epidemics (Ginsberg, et al., 2009). Similar to Alexa, Google Trends has both a Web-based portal as well as a back-end API.
2.6. YouTube video statistics
Statistics for a YouTube video can be enabled by the uploader for public consumption. These statistics can include a time series of views, duration of views, and shares. Instead of sharing raw numbers, the portal only exposes generated charts of regular time series or CDFs. In addition, no external API is available to access this information.
This section presents the results of our analysis. We first analyze the ranking of YouTube (§3.1), followed by the popularity of individual topics (§3.2). We then drill down into individual pieces of restricted content (§3.3) and also examine the popularity of anti-censorship tools (§3.4).
3.1. Web site ranking
YouTube has primarily borne the brunt of censorship in Pakistan. Most notably, it was first briefly banned in 2008 and since September, 2012 has experienced an indefinite blockage (Nabi, 2013). To gauge whether this censorship has had any effect on the usage of the Web site in the country (or more specifically its ranking), we make use of data from Alexa Web rankings. Unfortunately, Alexa no longer allows the user to query historical data for a specific country and only displays the current ranking on its Web site. To work around this, we use the Wayback Machine  from the Internet Archive — which caches previous versions of Web sites — to access previous rankings. Table 3 presents the ranking of YouTube in Pakistan for all Alexa snapshots from the Wayback Machine. It is clear that the ban on YouTube in Pakistan has not affected its rankings or indirectly its usage in the country. Specifically, in spite of the indefinite ban since 2012, the Web site has been consistently amongst the top 11 sites in the country.
Table 3: Alexa ranking for YouTube in Pakistan. Date Ranking April 2009 5 February 2010 5 March 2010 4 July 2011 4 August 2011 4 September 2011 4 May 2013 9 August 2013 10 June 2014 10 May 2014 10 July 2014 11
Similar to Pakistan, Turkey has also blocked YouTube multiple times in the last few years. The longest ban lasted 30 months, from May 2008 to October 2010. It also remained blocked between March 2014 and June 2014. In spite of this, Alexa rankings data in Table 4 shows that YouTube remained amongst the top 10 most visited Web sites in Turkey during both censorship periods.
Table 4: Alexa ranking for YouTube in Turkey. Date Ranking April 2009 8 June 2009 8 February 2010 5 September 2011 3 October 2011 3 April 2012 3 May 2012 3 May 2013 3 November 2013 3 March 2014 4 June 2014 4 July 2014 4
Limitations: While the data we collected clearly shows a censorship ineffectiveness trend, the wide gaps in the data do not permit us to claim that the behaviour holds throughout. Nonetheless, it shows that at least for certain months users were not deterred from accessing YouTube during the various blackout periods in both countries. To examine whether there is a correlation between content popularity and censorship, we next analyze data for specific events.
3.2. Topic trends
Having shown that censorship has not affected the ranking of YouTube in Pakistan and Turkey, we now dissect individual censorship events to study the manifestation of the Streisand Effect. To this end, we use data from Google Trends as a metric to gauge the popularity of restricted content. We first examine data from Pakistan followed by Turkey. As listed in Table 1, YouTube was blocked in Pakistan in February 2008 (PK1) for a few days in retaliation to the controversial movie “Fitna”; the trailer for which was uploaded to YouTube in January of the same year. In addition to Pakistan, several other countries, including Indonesia, also blocked YouTube in response. Somewhat amusingly, Pakistan restricted access to the Web site via BGP misconfiguration. This made the Web site globally inaccessible for a large part of the Internet for nearly two hours (Nabi, 2013). Figure 1 shows the Google search volume for the term “Fitna” for users within Pakistan in 2007 and 2008. The dotted green line represents the month in which the video was uploaded while the red dashed line marks the month when YouTube was blocked in Pakistan. The blue line represents the actual search volume data. The data shows that the popularity of the video spiked after it was restricted in Pakistan.
Figure 1: Search volume for “Fitna” in Pakistan. The green dotted line marks the upload date on YouTube and the red dashed line marks its censorship.
We also see a similar trend in Turkey. Since December 2013 a Twitter user with ID “Haramzadeler” has been using the Web site to share audio clips and documents implicating the ruling Turkish government in massive corruption scandals. The collection also includes recordings of the Turkish Prime Minister and his friends and family. In reaction, the Turkish government blocked access to SoundCloud, YouTube, and Twitter. In particular, SoundCloud was blocked in January 2014 (TR2 in Table 2). Figure 2 shows the Google search volume for “Haramzadeler” in Turkey. The results show that searches for the term peaked as a result of the blockage of SoundCloud in January even though Haramzadeler started uploading content months before that.
Figure 2: Search volume for “Haramzadeler” in Turkey. Audio uploaded on SoundCloud: green dotted line, censorship: red dashed line.
Limitations: From the two examples above, it is fairly straightforward to conclude that there is a strong correlation between censorship and the popularity of the censored content. It also suggests that this is not merely a case of the classic “correlation implies causation” fallacy. Nonetheless, in other cases this is harder to claim. For instance, Figure 3  shows Google Trends data for Pakistan from May 2010 for “Everybody Draw Muhammad Day” (PK2 in Table 1). This event resulted in the blockage of YouTube for a year. The finest granularity at which Google Trends data is available is on a weekly scale, therefore it is hard to tell if the content went viral after YouTube was blocked due to the fact that the content upload and censorship took place during the same week and hence the overlap between the green and red vertical lines. We see a similar trend in 2012, when YouTube was banned again due to another controversial movie “Innocence of Muslims” (PK3 in Table 1). As before, the green dotted line marks the time when the video was uploaded, the red dashed line marks the point at which the censorship was enforced, and the blue line shows the search volume from within Pakistan in Figure 4. Again, from this data it is hard to conclude if the censorship caused the popularity or vice versa.
Figure 3: Search volume for “Everybody Draw Muhammad Day” in Pakistan. Video upload: green dotted line, censorship: red dashed line.
Figure 4: Search volume for “Innocence of Muslims” in Pakistan. Video upload: green dotted line, censorship: red dashed line.
3.3. Affected content
So far we have been able to show that, a) censorship has not affected Web site rankings in Pakistan and Turkey, and b) topics spiked in popularity after they were censored. The latter only applies to Google search popularity of the content, not the content itself. Therefore, we now analyze YouTube statistics to gauge whether we can pinpoint the Streisand Effect at play for the actual content. This task is complicated by two factors: a) YouTube only serves graphs for statistics, not the underlying data, and b) most of the contentious videos have now been taken down. As a result, we only focus on one particular clip which was uploaded to YouTube in February 2014, wherein the Turkish Prime Minister allegedly discusses construction permits for an area earmarked for a forest with his friend, a business tycoon and a real estate developer. This video is amongst the set of content which resulted in the censorship of both YouTube and Twitter (TR3 and TR4 in Table 2) in Turkey in March 2014. Figure 5 plots the daily views of the said video. It is clear from the graph that even though the video was uploaded in February, its popularity spiked in March, after YouTube was censored.
Figure 5: Daily video viewership statistics for an alleged conversation between the Turkish Prime Minister and a business tycoon. The x-axis is time and the y-axis is daily number of views. Graph was directly copied from YouTube.
Limitations: While the example under consideration looks promising, it is the only one for which we were able to obtain data. Therefore, a single video is insufficient to extrapolate to other events.
Our analysis thus far shows that censorship has not deterred users from accessing restricted content. In fact, in certain cases the censorship has had the side effect of popularising the content. Bearing this in mind, the next natural question is to ask how users are accessing the restricted content. In case of Pakistan, in prior work (Nabi, 2013) we learned that users were actively making use of anti-censorship tools, such as VPNs and Tor to circumvent censorship. To determine whether there is a correlation between censorship and the use of anti-censorship tools, we again rely on data from Google Trends. Specifically, we note the search volume for “Spotflux”, “Tor”, “Ultrasurf”, and “Hotspot Shield”. We also add two generic circumvention terms viz. “Unblock” and “Proxy”. The data is visualized in Figure 6, the two red dashed lines mark PK2 and PK3, respectively. We see that the popularity of these tools spiked as a result of the two censorship events. Note that Spotflux was not available before PK2, that is why there is no spike in its search volume. We also conducted a similar study for Turkey, with the results presented in Figure 7. In this case, the four red lines represent TR1, TR2, TR3, and TR4, respectively. Again, we see that the popularity of these tools is positively correlated with the four censorship events. Unlike Pakistan, where Tor is popular in the entire time series, in Turkey its usage seems to have started only after TR3 and TR4.
Figure 6: Search volume for anti-censorship tools in Pakistan. The red lines mark PK2 and PK3, respectively.
Figure 7: Search volume for anti-censorship tools in Turkey. The red lines mark TR1, TR2, TR3, and TR4, respectively.
Using limited and sparse data from multiple sources we have been able to show that not only does censorship not work but it also inadvertently causes restricted content to become popular. This is a promising line of work but before Internet activists around the world can claim a convincing victory against the spectre of censorship, we need to analyze more comprehensive data. Towards this end, we now discuss how additional data might be useful and we also call upon organizations, such as Alexa and Google, who might hold this data to share it with the research community so that they can comb through it.
4.1. Open data against censorship
In §3.1, using data from Alexa Web Rankings, we argued that Web site level censorship has not diminished the usage of YouTube in Pakistan and Turkey for certain months. But the lack of monthly data throughout the events deters us from claiming the consistency of the trend. Similarity, in §3.2, using Google Trends data, we showed that certain pieces of content went viral after being censored, arguing for a cause and effect relationship between the latter and the former. While this trend was clear for PK1 and TR2, the weekly granularity of the data prevents us from verifying this for PK2 and PK3. Fine-grained daily data would be sufficient for teasing out the presence of the Streisand Effect. In the same vein, data about the usage of other videos on YouTube would be effective in inspecting the effect for individual pieces of content. In addition, data from ISPs and content distribution networks would be useful in examining access patterns during periods of censorship at a finer granularity. We hope that all of these organizations that are sitting on top of these silos of data will make it publicly available, similar to the filtering data  from the OpenNet Initiative.
4.2. The Streisand Effect as a means to an end
The Streisand Effect in a large number of instances showcases the futility of online censorship. Primarily, it shows that banning content on the Internet generates more interest in the content and increases its circulation. Cognizance of this phenomenon is slowly driving public policy and discourse. A recent example of this is the aftermath of the May 2014 European Court of Justice ruling in the Google Spain SL, Google Inc. v. Agencia Española de Protección de Datos, Mario Costeja González case. Under this “right to be forgotten” ruling, individuals have the right to request search engines to remove personal data from search results (European Commission, 2014). In the ensuing debate, EU countries have been weighing the implications and practicality of the EU-wide “General Data Protection Regulation”. Italy, for instance, in one of its proposals assumed that individuals want an umbrella removal of their personal information, which potentally requires a search engine such as Google to reach out to third parties, e.g., individual social network users or people with personal Web pages, to cascade information deletion. In reaction, Poland opined that the assumption is ill-founded and the required multimodal removal effort will most likely trigger the Streisand Effect as it will draw more attention to the requested content (Council of the European Union, 2014). Under Poland’s proposition, users should be able to choose which information to remove and from where. For example, a user might want a certain piece of information to be removed from Google search results but still be available on a blog. Therefore, awareness of the effect will be influential in defining digital rights policy in the EU. In a similar vein, we believe that governments that directly censor information on the Internet need to be mindful of inadvertent consequences of their actions. For example, countries like Pakistan and Turkey, regularly ban online content to safeguard the religious and social sensibilities of their citizens. Our examination of this category of censorship in this paper has shown that it has largely been ineffective. Knowledge of this observation should prompt state-level censors to rethink their position and explore alternative mechanisms to deal with such issues — for example, by directly engaging with their citizens and considering their points of view.
While the Streisand Effect is a handy instrument to keep censorship in check, it is only one of the many means to an end, not an end in itself. The end being an open, universally accessible Internet. To begin with, the effect does not manifest in all instances of censorship. The rampant presence of censorship around the world is a testament to this (Burnett and Feamster, 2013). Secondly, the goal of this paper is not to argue that online activists need to become complacent and actually embrace online censorship as it, due to the Streisand Effect, achieves the goal of proving its own ineffectiveness. On the contrary, anti-censorship efforts need to be beefed up in terms of political activism and campaigning, design of appropriate tools, and identification and measurement of Internet restrictions to keep up with the increase in country-level censorship (Franceschi-Bicchierai, 2013). Specifically, once content has been blocked, anti-censorship tools such as VPNs and proxies need to be employed. Without these tools the category of the Streisand Effect presented in this paper would not originate in the first place. Furthermore, the initial study and measurement of Internet censorship enables the design of tools to counteract it. For instance, ignoring TCP RST packets as a method to bypass the Great Firewall of China came about as a result of studying its modus operandi (Clayton, et al., 2006). Finally, the efforts of Internet freedom activists in recent years have been effective in raising awareness. In some cases — for instance, in Turkey (Reuters, 2014) — they have even succeeded in helping content become unblocked through legal recourses. Overall, the Streisand Effect would not exist if all of the above efforts were abandoned. In essence, the aim of this paper is to raise awareness about the Streisand Effect in the context of state-level censorship rather than to argue for it as an all-encompassing anti-censorship mechanism.
5. Conclusion and future work
By making use of data from Google Trends, YouTube statistics, and Alexa Web rankings, we were able to not only show that censorship is ineffective but rather that it also has the inadvertent effect of making restricted content go viral. Specifically, we show that the national ranking of YouTube in Pakistan and Turkey has remained largely the same in spite of multiple censorship events. In addition, Google Trends and YouTube Statistics numbers show that blocked videos spiked in popularity after they were censored. This paper was just an initial window into the interplay of the Streisand Effect and state-level censorship. In the future, we aim to extend our study to other countries. In addition, we hope to make use of other datasets to augment our current findings. Finally, Google Trends data also breaks down search volume by region. It would interesting to work out say if censorship circumvention is an urban or rural phenomenon in Pakistan and Turkey.
About the author
Zubair Nabi is a Research Staff Member in the Big Data Systems and Analytics Group at IBM Research, Ireland. His interests include data intensive computing, technology for the developing world, and Internet accessibility. His work has been featured in MIT Technology Review, CNET, Asian Scientist, and SciDev.net, among others. Zubair has an M.Phil. in computer science from the University of Cambridge.
E–mail: zubairn [at] ie [dot] ibm [dot] com
4. Note that the x-axis appears as fine-grained weekly dated, unlike the monthly data in Figure 1.
Collin Anderson, 2013. “Dimming the Internet: Detecting throttling as a mechanism of Censorship in Iran,” CoRR, at http://arxiv.org/abs/1306.4361, accessed 22 October 2014.
Simurgh Aryan, Homa Aryan, and J. Alex Halderman, 2013. “Internet censorship in Iran: A first look,” paper presented at the Third USENIX Workshop on Free and Open Communications on the Internet, at https://www.usenix.org/conference/foci13/internet-censorship-iran-first-look, accessed 22 October 2014.
David Bamman, Brendan O’Connor, and Noah Smith, 2012. “Censorship and deletion practices in Chinese social media,” First Monday, volume 17, number 3, at http://firstmonday.org/article/view/3943/3169, accessed 22 October 2014.
doi: http://dx.doi.org/10.5210/fm.v17i3.3943, accessed 22 October 2014.
Jeff Barto, 2013. “Encryption and decryption — A never ending battle” (2 October), at https://community.thawte.com/blog-posts/encryption-and-decryption-%E2%80%93-never-ending-battle, accessed 22 October 2014.
Yavuz Baydar, 2014. “The Internet bill: Is freedom of expression under threat in Turkey?” Al Jazeera (8 February), at http://m.aljazeera.com/story/201427122110808964, accessed 22 October 2014.
BBC News, 2014. “Turkey passes law tightening control of Internet” (5 February), at http://www.bbc.com/news/world-europe-26062038, accessed 22 October 2014.
Martin Brown, 2010. “Pakistan hijacks YouTube,” Dyn Research, at http://research.dyn.com/2008/02/pakistan-hijacks-youtube-1/, accessed 22 October 2014.
Sam Burnett and Nick Feamster, 2013. “Making sense of Internet censorship: A new frontier for Internet measurement,” ACM SIGCOMM Computer Communication Review, volume 43, number 3, pp. 84–89
doi: http://doi.acm.org/10.1145/2500098.2500111, accessed 22 October 2014.
Sam Burnett, Nick Feamster, and Santosh Vempala, 2010. “Chipping away at censorship firewalls with user-generated content,” SENIX Security ’10: Proceedings of the 19th USENIX Conference on Security, p. 29, and at https://www.usenix.org/conference/usenixsecurity10/chipping-away-censorship-firewalls-user-generated-content, accessed 22 October 2014.
Mario Cacciottolo, 2012. “The Streisand Effect: When censorship backfires,” BBC News (15 June), at http://www.bbc.com/news/uk-18458567, accessed 22 October 2014.
Neal Caren and Didem Türkoğlu, 2014. “Political controversy has helped fuel Twitter’s growth in Turkey,” Washington Post (28 March), at http://www.washingtonpost.com/blogs/monkey-cage/wp/2014/03/28/political-controversy-has-helped-fuel-twitters-growth-in-turkey/, accessed 22 October 2014.
Hyunyoung Choi and Hal Varian, 2012. “Predicting the present with Google Trends,” Economic Record, volume 88, supplement s1, pp. 2–9.
doi: http://dx.doi.org/10.1111/j.1475-4932.2012.00809.x, accessed 22 October 2014.
Citizen Lab, 2013. “O Pakistan, we stand on guard for thee: An analysis of Canada-based Netsweeper’s role in Pakistan’s censorship regime” (20 June), at https://citizenlab.org/2013/06/o-pakistan/, accessed 22 October 2014.
Ian Clarke, Scott G. Miller, Theodore W. Hong, Oskar Sandberg, and Brandon Wiley, 2002. “Protecting free expression online with Freenet,” IEEE Internet Computing, volume 6, number 1, pp. 40–49.
doi: http://dx.doi.org/10.1109/4236.978368, accessed 22 October 2014.
Richard Clayton, Steven J. Murdoch, and Robert N. M. Watson, 2006. “Ignoring the Great Firewall of China,” In: George Danezis and Philippe Golle (editors). Privacy Enhancing Technologies: 6th International Workshop, PET 2006, Cambridge, U.K., June 28–30, 2006, Revised Selected Papers. Lecture Notes in Computer Science, volume 4258, pp. 20–35.
doi: http://dx.doi.org/10.1007/11957454_2, accessed 22 October 2014.
Council of the European Union, 2014. “Proposal for the right to be forgotten” (3 September), at http://register.consilium.europa.eu/doc/srv?l=EN&f=ST%2012274%202014%20REV%202, accessed 22 October 2014.
Alberto Dainotti, Claudio Squarcella, Emile Aben, Kimberly C. Claffy, Marco Chiesa, Michele Russo, and Antonio Pescapé, 2011. “Analysis of country-wide internet outages caused by censorship,” IMC ’11: Proceedings of the 2011 ACM SIGCOMM Conference on Internet Measurement Conference, pp. 1–18.
doi: http://doi.acm.org/10.1145/2068816.2068818, accessed 22 October 2014.
Roger Dingledine, Nick Mathewson, and Paul Syverson, 2004. “Tor: The Second-generation onion router,” SSYM ’04: Proceedings of the 13th Conference on USENIX Security Symposium, volume 13, p. 21.
Robert P. Doyle, 2010. Banned books: Challenging our freedom to read. Chicago: American Library Association.
Economist blog standard, 2008. “Authoritarian governments can lock up bloggers. It is harder to outwit them” (26 June), at http://www.economist.com/node/11622401, accessed 22 October 2014.
eMarketer, 2013. “Young drive Web use in Turkey” (20 September), at http://www.emarketer.com/Article/Young-Drive-Web-Use-Turkey/1010231, accessed 22 October 2014.
European Commission, 2014. “Factsheet on the ‘Right to be Forgotten’ ruling,” at http://ec.europa.eu/justice/data-protection/files/factsheets/factsheet_data_protection_en.pdf, accessed 22 October 2014.
Nick Feamster, Magdalena Balazinska, Greg Harfst, Hari Balakrishnan, and David Karger, 2002. “Infranet: Circumventing Web censorship and surveillance,” Proceedings of the 11th USENIX Security Symposium, pp. 247–262, and at https://www.usenix.org/legacy/events/sec02/feamster/feamster_html/index.html, accessed 22 October 2014.
Lorenzo Franceschi-Bicchierai, 2013. “Google: Government censorship attempts increasing,” Mashable (25 April), at http://mashable.com/2013/04/25/google-government-censorship/, accessed 22 October 2014.
John Geddes, Max Schuchard, and Nicholas Hopper, 2013. “Cover your ACKs: Pitfalls of covert channel censorship circumvention,” CCS ’13: Proceedings of the 2013 ACM SIGSAC Conference on Computer & Communications Security, pp. 361–372.
doi: http://doi.acm.org/10.1145/2508859.2516742, accessed 22 October 2014.
Megan Geuss, 2013. “Wikipedia editor allegedly forced by French intelligence to delete ‘classified’ entry,” Ars Technica (6 April), at http://arstechnica.com/tech-policy/2013/04/wikipedia-editor-allegedly-forced-by-french-intelligence-to-delete-classified-entry/, accessed 22 October 2014.
Jeremy Ginsberg, Matthew Mohebbi, Rajan Patel, Lynnette Brammer, Mark Smolinski, and Larry Brilliant, 2009. “Detecting influenza epidemics using search engine query data,” Nature, volume 457, number 7232 (19 February), pp. 1,012–1,014, and at http://www.nature.com/nature/journal/v457/n7232/full/nature07634.html, accessed 22 October 2014.
doi: http://dx.doi.org/10.1038/nature07634, accessed 22 October 2014.
Andy Greenberg, 2007. “The Streisand Effect,” Forbes (11 May), at http://www.forbes.com/2007/05/10/streisand-digg-web-tech-cx_ag_0511streisand.html, accessed 22 October 2014.
Ki Mae Heussner, 2008. “Injunction to silence MIT student hackers backfires,” ABC News (12 August), at http://abcnews.go.com/ Technology/story?id=5564423, accessed 22 October 2014.
Amir Houmansadr, Giang T.K. Nguyen, Matthew Caesar, and Nikita Borisov, 2011. “Cirripede: Circumvention infrastructure using router redirection with plausible deniability,” CCS ’11: Proceedings of the 18th ACM Conference on Computer and Communications Security, pp. 187–200.
doi: http://doi.acm.org/10.1145/2046707.2046730, accessed 22 October 2014.
Internet World Stats, 2011. “Internet usage and population in South America,” at http://www.internetworldstats.com/stats15.htm, accessed 22 October 2014.
Luca Invernizzi, Christopher Kruegel, and Giovanni Vigna, 2013. “Message in a bottle: Sailing past censorship,” ACSAC ’13: Proceedings of the 29th Annual Computer Security Applications Conference, pp. 39–48.
doi: http://doi.acm.org/10.1145/2523649.2523654, accessed 22 October 2014.
Seth Jacobson, 2009. “Twitter claims new scalp as Trafigura backs down,” The Week (13 October), at http://www.theweek.co.uk/business/19243/twitter-claims-new-scalp-trafigura-backs-down, accessed 22 October 2014.
Legal Eagle, 2012. “The Streisand Effect” (17 January), at http://skepticlawyer.com.au/2012/01/17/the-streisand-effect/, accessed 22 October 2014.
Haseeb Malik, 2014. “Petition filled against FinFisher’s surveillance deployment in Pakistan” (7 April), at http://propakistani.pk/2014/04/07/finfisher-surveillance-technology-operates-under-pta/, accessed 22 October 2014.
Angelica Mari, 2014. “Brazil passes groundbreaking Internet governance bill,” ZDNet (26 March), at http://www.zdnet.com/brazil-passes-groundbreaking-internet-governance-bill-7000027740/, accessed 22 October 2014.
Morgan Marquis-Boire with Bill Marczak, Claudio Guarnieri, and John Scott-Railton, 2013. “For their eyes only: The commercialization of digital spying,” Citizen Lab and Centre for Global Security Studies, Munk School of Global Affairs, University of Toronto (1 May), at https://citizenlab.org/storage/finfisher/final/fortheireyesonly.pdf, accessed 22 October 2014.
Zubair Nabi, 2013. “The anatomy of Web censorship in Pakistan,” paper presented at the Third USENIX Workshop on Free and Open Communications on the Internet, at https://www.usenix.org/conference/foci13/workshop-program/presentation/Nabi, accessed 22 October 2014.
Joe Parkinson, Sam Schechner, and Emre Peker, 2014. “Turkey’s Erdogan: One of the world’s most determined Internet censors,” Wall Street Journal (2 May), at http://online.wsj.com/news/articles/SB10001424052702304626304579505912518706936, accessed 22 October 2014.
Reuters, 2014. “Turkey lifts Twitter ban after court ruling,” New York Post (3 April), at http://nypost.com/2014/04/03/turkey-lifts-twitter-ban-after-court-ruling/, accessed 22 October 2014.
Max Schuchard, John Geddes, Christopher Thompson, and Nicholas Hopper, 2012. “Routing around decoys,” CCS ’12: Proceedings of the 2012 ACM Conference on Computer and Communications Security, pp. 85–96.
doi: http://doi.acm.org/10.1145/2382196.2382209, accessed 22 October 2014.
Bina Shah, 2014. “Trying to dam a digital sea,” New York Times (10 April), at http://www.nytimes.com/2014/04/11/opinion/shah-trying-to-dam-a-digital-sea.html, accessed 22 October 2014.
Olivia Solon, 2013a. “Avaaz counters accusations of clicktivism and takes on the lion bone trade,” Wired (17 June), at http://www.wired.co.uk/news/archive/2013-06/17/lion-bone-clicktivism, accessed 22 October 2014.
Olivia Solon, 2013b. “From Beyonce to The Oatmeal: The Streisand Effect at its most glorious,” Wired (31 December), http://www.wired.co.uk/news/archive/2013-12/31/streisand-effect, accessed 22 October 2014.
Turkish Statistical Institute, 2012. “Information and communication technology (ICT) usage survey on households and individuals, 2012” (16 August), at http://www.turkstat.gov.tr/PreHaberBultenleri.do?id=10880, accessed 22 October 2014.
John-Paul Verkamp and Minaxi Gupta, 2013. “Five incidents, one theme: Twitter spam as a weapon to drown voices of protest,” paper presented at the Third USENIX Workshop on Free and Open Communications on the Internet, at https://www.usenix.org/conference/foci13/technical-sessions/papers/verkamp, accessed 22 October 2014.
John-Paul Verkamp and Minaxi Gupta, 2012. “Inferring mechanics of Web censorship around the world,” paper presented at the Second USENIX Workshop on Free and Open Communications on the Internet, at https://www.usenix.org/conference/foci12/workshop-program/presentation/verkamp, accessed 22 October 2014.
Philipp Winter and Stefan Lindskog, 2012. “How the Great Firewall of China is blocking Tor,” paper presented at the Second USENIX Workshop on Free and Open Communications on the Internet, at https://www.usenix.org/conference/foci12/workshop-program/presentation/Winter, accessed 22 October 2014.
Emma Woollacott, 2014. “Streisand Effect takes hold as Turkey bans Twitter,” Forbes (21 March), at http://www.forbes.com/sites/emmawoollacott/2014/03/21/streisand-effect-takes-hold-as-turkey-bans-twitter/, accessed 22 October 2014.
Eric Wustrow, Scott Wolchok, Ian Goldberg, and J. Alex Halderman, 2011. “Telex: Anticensorship in the network infrastructure,” SEC ’11: Proceedings of the 20th USENIX conference on Security, at https://www.usenix.org/conference/usenix-security-11/telex-anticensorship-network-infrastructure, accessed 22 October 2014.
Earl Zmijewski, 2014. “Turkish Internet censorship takes a new turn,” Dyn Research (30 March), at http://www.renesys.com/2014/03/turkish-internet-censorship/, accessed 22 October 2014.
Received 15 September 2014; revised 24 September 2014; revised 29 September 2014; accepted 30 September 2014.
This paper is in the Public Domain.
R̶e̶s̶i̶s̶t̶a̶n̶c̶e̶ censorship is futile
by Zubair Nabi.
First Monday, Volume 19, Number 11 - 3 November 2014