The Chinese government plans to monitor its people with a comprehensive social credit system.” The goal is to promote honesty and sincerity in order to promote economic and social progress. In the process, those who betray trust are to be severely punished.
China is currently busy creating a digital data monster with tentacles extending into every aspect of life. This is causing concerns about the rampant frenzy to collect data and how it will be handled. The Chinese social credit system is officially scheduled to go into operation in 2020. From then on, not one of the country’s approx. 1.5 billion inhabitants will be able to escape the state’s rating system.
Education in “goodness” Zhang Zheng, director of the China Credit Research Center at Peking University, is an important thought leader and theoretician of the Chinese social credit system. His mindset is rooted in his socialization because the economics professor had initially studied mathematics and natural sciences, which requires a rational and analytical way of thinking. However, dealing with human beings and the problems of society requires a broader, more differentiated approach, which is often difficult for dedicated natural scientists. Social sciences are more than just ones and zeros, black and white, right and wrong, good and evil, but the Chinese social credit system is based precisely on this simplified dualistic way of thinking.
Zheng is convinced that the Chinese social credit system, i.e., socialization as a “good” person with the help of digital tools, will become a sustainable cornerstone for the moral order of Chinese society. This system is intended to improve the morals of society. Whether the everyday morals of the people or the business ethics of companies, the system is supposed to that the rules are followed. This has particularly obvious consequences on individuals: good citizens would be rewarded and favored, while bad ones would be sanctioned with severe restrictions in daily life.
Structure and function
The Chinese social credit system is based on centralized databases containing such records as medical and court files, online shopping, posts on social networks, internet search queries, travel plans, and purchases with credit cards or payment apps. These records are then analyzed and weigh this cluster of data to come up with a single score. Companies and institutions will have no choice but to make their data available to the system. However, there won’t be much need to put pressure on Chinese companies, since there are already voluntary systems in place such as Alibaba’s Sesame Credit (with over 450 million active users), Tencent (operator of the successful Chinese messaging, social media and mobile payment app WeChat), and Baidu. China’s private internet companies have indicated that the Communist Party may use their compiled data and cutting-edge technologies because, in return, they will gain access to previously inaccessible government databases.
Looking at Sesame Credit, not only payment behavior but also “habits or preferences” and “personal networks” can influence creditworthiness. According to Li Yingyun, head of development at Sesame Credit, someone who plays video games ten hours a day is classified as a sluggish person, but those who buy diapers frequently are likely to be a parent and are therefore willing to accept a higher degree of responsibility. Ambitious gamers risk a lower score, while those who are responsible get a higher one. It’s also worthwhile to pick friends with high scores because these can help increase your score. However, if your friends have low scores, you risk losing points. If you are looking for a partner, you can advertise with a high score, because Sesame Credit cooperates with Baihe, China’s largest online dating agency. This means, however, that people with low scores will inevitably remain single.
Pilot operation already running
Companies are not the only ones that are already heavily collecting, processing, and evaluating data. Some three dozen Chinese cities are already experimenting with different social credit systems. For example, Rongcheng, a city of about 670,000 inhabitants on the east coast, has been operating a social credit system since 2014 regarded as a showcase project for a China-wide system. With their Honest Shanghai App, Shanghai operates another popular system, which has also implemented facial recognition. To register, the individual’s is captured with the mobile camera and compared and verified with the electronic identity card. A short time later, users get their first score. This score is updated at the end of each month. The criteria and factors used for a high or low rating are confidential. However, the system takes into account about 3,000 pieces of data per person from almost a hundred government data sources (Rob Schmitz, “What’s Your ‘Public Credit Score’? The Shanghai Government Can Tell You“, NPR.org, 03.01.2017).
Even if individual factors evaluated in the pilot projects are confidential, the Chinese social credit system generally concentrates on the evaluation of four key parameters:
Commercial activities: commercial activities form the basis of the system, because one of the goals of the Chinese government is to use the system to improve the trust in the commercial sector among citizens, but also between citizens and business. So if you pay your bills on time, you will have a clear advantage. Incidentally, such credit rating systems are also common in the West (for example, Schufa in Germany and FICO in the US). The Chinese, however, go one step further: those who travel without a ticket or who get into debt with spending are, in many cases, no longer allowed to travel by express train or plane. Last year alone, this penalty was imposed about 6.7 million times, according to the official figures of the Supreme Court.
We have had the social credit system in our village for several years now. No matter what we do, we think about our credit points. We support the village where we can. We clean a lot and sweep the public areas. Putting garbage or even grass in front of your own door is not allowed. If someone doesn’t follow these rules, they’re considered dishonest. If the village head asks for anything, we do it. Those who keep everything clean and in order are regarded as role models. — cited in Axel Dorloff, “Sozialkredit-System: China auf dem Weg in die IT-Diktatur“, Deutschlandfunk, 09.09.2017.
Social behavior: whether online or off, social behavior plays an important role in the assessment. With a reward and punishment mechanism, the system aims to train residents to behave positively, at least as the government sees it. In Rongcheng, whoever helps others or gets involved in city projects will, for example, get 5-10 additional points. A similar system is in place in Shanghai: those who help older inhabitants or the poor can earn additional points, too, but whether this represents moral progress remains questionable.
Administrative activities: the system will also simplify administrative procedures, as unauthorized requests for public assistance will result in a deduction of points. This applies in particular to the submission of petitions critical of the government. Those who criticize the Communist Party in the social media should not be surprised if they end up on the blacklist. Requests from people below a certain score will be postponed or even ignored. On the other hand, people with above-average scores already enjoy preferential treatment.
Criminal prosecution: law enforcement is already integrated in Rongcheng. If you run a red light, you will immediately lose 5 points; if you drive drunk or are involved in a brawl, you will immediately be blacklisted. The score serves as a kind of criminal record: the inhabitants of Rongcheng have to regularly present their score for job promotions, for membership in the Communist Party, when applying for a bank loan. Nothing happens anymore without a good score.
Rewards and punishments
The rewards and punishments for high or low scores currently vary from system to system. In Rongcheng, everyone starts with 1,000 points, which then increases or decreases depending on the behavior of the person concerned. The highest rating is AAA, which is at least 1,050 points; at the other end of the scale is D, which is fewer than 600 points. Persons with at least an A rating are on a red list, while those below are on a blacklist. Those on the red list are given preferential treatment for admissions to schools, for social benefits, or even when purchasing insurance. Those in the C Group are checked regularly and are subject to certain restrictions. This could, for example, result in the reduction of welfare payments. Those who appear in the lowest D Group no longer qualify for management positions, lose certain benefits and lose their creditworthiness. Another important aspect is the public emphasis on ethical role models or the condemnation of those who “betray trust”. Usually, names, photos, identity numbers, and in some cases even private addresses are published. The majority will hardly be bothered by this at the moment because about 90% of the inhabitants in Rongcheng have an A (Simina Mistreanu, “China Is Implementing a Massive Plan to Rank Its Citizens, and Many of Them Want In“, Foreign Policy, 03.04.2018).
At Alibaba, a score of over 600 leads to the possibility of taking out a small loan of around 5,000 yuan (around $700) when making purchases in its online shop. For scores 650 and higher, one no longer needs a deposit to rent a car, and you might enjoy the benefits of VIP treatment at certain hotels and airports. From 700 points, additional documents can be dispensed with on a trip to Shanghai, and for a person with at least 750 points, the procedure for applying for a Schengen visa is faster on the Chinese side. Currently, Sesame Credit does not yet seem to be imposing penalties (Rachel Botsman, “Big Data Meets Big Brother as China Moves to Rate Its Citizens“, Wired, 21.10.2017).
I’m being punished for issuing a credit guarantee for someone else. The loan wasn’t repaid and I was punished. When I wanted to buy a plane ticket, I couldn’t get one. As a result, I found out that I can no longer buy tickets. That was in November 2016. I can’t buy plane tickets or express train tickets. — cited in Axel Dorloff, “Sozialkredit-System: China auf dem Weg in die IT-Diktatur“, Deutschlandfunk, 09.09.2017.
The wide range of rewards should not deceive readers about the immense risks of this system. A totalitarian surveillance system is currently being established in China, which, depending on political needs, could quickly turn China into a huge prison. People on blacklists and with travel restrictions report that it is very difficult to be removed from these lists (also read Simina Mistreanu, “China Is Implementing a Massive Plan to Rank Its Citizens, and Many of Them Want In“).
However, the impact may not be limited to China. Even if a politically flavored social credit system is rather unlikely in democratic states, this does not mean that companies operating in democratic states do not want to adopt such a business model. Although China is the salient example of such a system, similar approaches can be seen elsewhere in the world. Companies have been assessing individual creditworthiness for a long time. For example: are you wondering why you can no longer get an Uber? Well, chances are you have a dismal passenger rating. By the way, Uber knows who among their customers has had a one-night-stand (Bradley Voytek, “Rides of Glory“, Uber Blog, 12.03.2012). The Danish company Deemly demonstrates how a “light” social credit system could also be marketed in Western countries. It evaluates the trustworthiness of individuals based on the evaluation of their activities on social platforms. In this context, the “Nosedive” episode in the “Black Mirror” series, a popular critique of technology and its social impact, seems to be right on the money. Besides, it should not be forgotten that internationally active Chinese companies such as Alibaba collect data not only from Chinese citizens but from all their customers (including geodata). With the rewards offered, customers are even voluntarily submitting their data.
Democratic freedoms have deteriorated in Venezuela. The government has been characterized as an authoritarian regime, closing spaces for public discussions and free expression, while systematic violations of human rights have intensified. According to IPYS Venezuela, the elections held in Venezuela in recent years have suffered from a lack of fair conditions. The institutionality and the State of Rights have been broken, given the lack of autonomy and independence of the public powers, all dominated by the strength of the United Socialist Party of Venezuela, which has accompanied Hugo Chávez and currently maintains Nicolás Maduro.
Venezuela is experiencing a complex humanitarian emergency, intensified by hyperinflation, the absence of transparency in public management, and the weakness of democratic institutions. These conditions have negatively impacted the quality of life of citizens, as well as the conditions for the protection of human rights. Within this context, Venezuelans have been deprived of the right to decent housing and have very limited access to public services.
Internet blackouts in Venezuela have been documented by IPYS Venezuela, which have left citizens in rural, suburban and urban areas of the country without internet connectivity. According to IPYS Venezuela, these internet blackouts have harmed citizens’ rights to access information and freedom of expression. Freedom House scores Venezuela 20⁄25 in obstacles to access the internet (where a larger ranking is worse).
Research conducted by Mariengracia Chirinos in terms of public policies on internet access reveals that, between 2007 to 2017, a vision of political and social control prevailed in Venezuela, in favor of the defense of national sovereignty and the “defense of the country”. This however, she notes, contradicts the principles of inclusion, diversity, openness, competitiveness and freedom that should guide the process of formulating internet access policies.
Public policies around internet access have been limited by regulatory processes (which follow the model of a closed society), affecting market competitiveness and incentives for investment. This has had a negative impact on technological advancements in the telecommunications sector, which are far from the standards of ECLAC (2016) and the OECD (2016). Between 2017 to 2018, this contributed towards connectivity issues across Venezuela.
IPYS Venezuela reports that digital rights were at risk throughout 2017 in light of several restrictive regulations. Police persecution manifested through arbitrary arrests of citizens based on their opinions expressed online through social networks, various portals of digital media and civil society organizations were attacked, and web portals were selectively blocked. Threats have been made against journalists, while official structures for online surveillance and police monitoring have been proposed.
The turning point came with the approval of the Anti-Hate law. Last November, the National Constituent Assembly (ANC) – a body created outside of the national constitution and which functions as a “superpower” with all of the ruling parliamentarians – approved the “Law against Hatred, for Peaceful Co-existence and Tolerance”. This regulation empowers authorities to block websites that are deemed to spread hate or incite violence. If messages that are considered to “incite hatred” are not removed by website owners within 6 hours, they may be subjected to a fine. The law also includes prison sentences, ranging from 10 to 20 years, for those who do not comply with censorship requests by authorities. Similarly, Article 27 of the Law on Social Responsibility in Radio, Television and Electronic Media sets conditions for the prohibition of content that does not acknowledge the legitimacy of authorities or which fosters citizen anxiety.
Reported cases of internet censorship
Pervasive levels of internet censorship have been carried out in Venezuela since 2014, largely monitored and documented by local civil society groups IPYS Venezuela and Venezuela Inteligente. Their study (between 2015 and 2016) showed that 43 websites were systematically blocked by one or more Venezuelan ISPs. The types of websites that appeared to be blocked the most include: sites related to the parallel market of the dollar (44%), media (19%), blogs criticizing Chavez (12%), games of chance and online bets (9%), collaboration tools or shorteners (5%), personal communication tools (5%), gore (2%), anonymization and circumvention sites (2%), and hosting services (2%). Movistar was found to block sites the most, with 41 blocked domains, corresponding to 35 different websites. The types of sites blocked by Movistar – but which weren’t blocked by CANTV – include: parallel dollar market, chavismo criticism blogs, hosting services, collaboration tools or shorteners, and digital media.
Data presented in August 2017 by Venezuela Inteligente, as follow-up to their previous study with IPYS Venezuela, shows that of the blocked sites, 36% of them were related to currency exchange rates, 32% were media, 16% games of chance and online bets, 12% social networking or communications tools, and 4% of them were blogs critical of the government. 24% of all blocked sites were international with international audiences, while 76% of them had (mostly) local audiences.
Between 2017 to 2018, IPYS Venezuela documented seven cases of internet censorship, involving news websites, currency exchange websites and other sites discussing corruption and economic information. Today, these seven news portals remain blocked by CANTV, Movistar and Digitel, according to OONI Probe network measurement data collected by IPYS Venezuela and Venezuela Inteligente.
Last year, Venezuela Inteligente reported, through the VEsinFiltro project, how three private online streaming broadcasters – Vivoplay, VPI, and Capitolio TV (site now defunct) – were blocked simultaneously by all major ISPs (primarily by means of DNS) as a result of broadcasting live street protests. VPI and Capitolio TV resorted to livestreaming on YouTube, instead of on their own sites, to circumvent the block. The Maduradas portal was also blocked by means of DNS.
Media websites blocked in 2018 include El Pitazo, La Patilla and El Nacional. These censorship events were (temporarily) implemented by both private and state providers, who blocked the sites at their own discretion without a court order, violating due process. These media outlets were blocked by means of DNS tampering and HTTP blocking, primarily by CANTV, Movistar, Movilnet, and Digitel.
Authorities of the National Telecommunications Commission have previously ordered the blocking of websites that disseminate “destabilizing” information or form a “media war” against the government. However, no court order or other legal justification was provided for the censorship events that occurred over the last year. Furthermore, the National Telecommunications Commission has repeatedly ignored public information requests regarding recent internet censorship events.
Measuring internet censorship
To measure internet censorship in Venezuela, the researchers ran OONI’s network measurement software (OONI Probe) on a daily basis across multiple local vantage points. OONI Probe is free and open source software designed to measure various forms of network interference.
The main OONI Probe tests that were ran as part of this study include:
OONI’s Web Connectivity test is designed to measure whether websites are blocked by means of DNS tampering, TCP/IP blocking, or by an HTTP transparent proxy. This test is automatically performed both over the vantage point of the user and from a non-censored control vantage point. If the results from both vantage points match, then the tested website is most likely accessible. If the results however differ, then the measurement is flagged as anomalous. OONI’s current methodology only confirms the blocking of a website if a blockpage is served. In cases where ISPs do not serve blockpages, the relevant network measurements are analyzed over time, examining whether the specific types of failures persist and what causes these failures (i.e. ruling out false positives).
The testing was mostly limited to the URLs included in the Citizen Lab’sglobal and Venezuelan test lists. These lists consist of a variety of different types of URLs that fall under 30 categories and that are tested for censorship by network measurement projects like OONI. Throughout the course of this research, the researchers updated the Venezuelan test list to ensure that reportedly blocked sites were being tested. Overall, around 1,410 URLs, included in both the Citizen Lab’s global and Venezuelan test lists, were measured as part of this study.
In an attempt to identify which equipment was used to implement internet censorship in Venezuela, the researchers ran OONI’s HTTP Invalid Request Line and HTTP Header Field Manipulation tests. Both tests are designed to measure networks with the aim of identifying the presence of middleboxes. OONI’s HTTP Invalid Request Line test does this by sending an invalid HTTP request line to an echo server listening on the standard HTTP port. If a middlebox is present, the invalid HTTP request line will be intercepted by the middlebox, potentially triggering an error that will be sent back to OONI servers. In the past, this has enabled the identification of censorship equipment in various countries around the world. OONI’s HTTP Header Field Manipulation test, on the other hand, attempts to identify middleboxes by sending HTTP requests with non-canonical HTTP headers. If a middlebox is present, it will likely normalize the headers or add extra headers, enabling the identification of its presence in the network. In addition to OONI Probe tests, the researchers also performed additional network measurement tests via Raspberry Pi deployments in Venezuela.
To monitor the accessibility of popular instant messaging platforms over time, the researchers ran OONI’s WhatsApp, Facebook Messenger, and Telegram tests. These tests are designed to measure the reachability of the WhatsApp, Facebook Messenger, and Telegram apps and web interfaces through DNS lookups and by attempting to establish TCP connections to their endpoints.
In light of increased censorship events over the last years, the researchers decided to monitor the accessibility of censorship circumvention tools as well. Many circumvention tool sites were included in the Citizen Lab’s global test list, which the researchers measured via OONI’s Web Connectivity test. But we also ran OONI’s Vanilla Tor and Tor Bridge Reachability tests, which are designed to measure the blocking of the Tor network and Tor bridges.
Once network measurement data was collected from all of these tests, OONI data was subsequently processed and analyzed based on a standardized set of heuristics for detecting internet censorship and traffic manipulation. The researchers analyzed all OONI Probe network measurements collected from Venezuela between 20th February 2014 to 10th August 2018.
The main findings though that the researchers present in this study are based on:
Networks from which most of the recent measurements were collected from, namely: Digitel (AS264731), CANTV (AS8048), Movistar (AS6306) and Movilnet (AS27889).
Recent censorship findings that are currently more relevant.
Censorship findings that have been persistent over time (i.e. sites that remained blocked over time and which presented the highest ratio of anomalies).
Acknowledgement of limitations
The first limitation of this study is associated with the testing period. This study includes an analysis of thousands of network measurements collected from Venezuela over the last four years, between 20th February 2014 to 10th August 2018. Censorship events that may have occurred before and/or after the analysis period are not examined as part of this study.
Another limitation to this study is associated to the amount and types of URLs that were tested for censorship. OONI’s Web Connectivity test was run to measure the accessibility of 287 URLs that are more relevant to the Venezuelan context and 1,123 internationally relevant sites. All of these URLs were selected in collaboration with community members over the last years. The researchers acknowledge the URL selection bias and that the testing sample of URLs might exclude many other sites that are blocked in Venezuela. The researchers therefore encourage researchers and community members to continue reviewing and contributing to these test lists to help improve future research and analysis.
Since block pages weren’t detected in Venezuela (at least for none of the tested URLs), censorship findings are presented with caution, acknowledging that false positives may be present. This is the primary reason why the researchers mainly present findings that (a) presented consistent anomalies over time (suggesting blocking) and (b) IPYS Venezuela and Venezuela Inteligente were able to verify locally in terms of (in)accessibility.
Finally, while network measurements were collected from multiple ASNs in Venezuela, OONI’s software tests were not run consistently across all networks. To share more recent and relevant findings, the researchers mainly focus on ASNs from which measurements were collected the most over the last months: Digitel (AS264731), CANTV (AS8048), Movistar (AS6306) and Movilnet (AS27889).
Following Venezuela’s 2015 elections, civil society groups IPYS Venezuela and Venezuela Inteligente reported (through the use of OONI Probe) on the blocking of a number of websites, including currency exchange websites, blogs expressing political criticism and media-related sites.
The researcher’s latest OONI findings show that such websites are currently blocked by multiple Venezuelan ISPs and have remained blocked all along. Measurements collected from Venezuela also suggest that a number of other sites (such as el-nacional.com, lapatilla.com, elpitazo.com and armando.info) have more recently been blocked as well.
As part of the following sections, the researchers share OONI data pertaining to the blocking of news outlets, sites expressing political criticism, currency exchange sites and zello.com. The following data is based on recent measurements collected from four Venezuelan networks: Digitel (AS264731), CANTV (AS8048), Movistar (AS6306) and Movilnet (AS27889). The researchers also confirm the blocking of the Tor network by state-owned CANTV.
Independent media websites are blocked in Venezuela (primarily by means of DNS tampering), as illustrated in the following table (based on recent OONI measurements).
El Pitazo is an independent news outlet run by Venezuelans that started off as a YouTube channel in 2014, expanded to a radio program, and eventually created a media website. They aim to share information with the most economically disadvantaged populations of Venezuela and to shed light on issues that are otherwise censored by state-owned media. El Pitazo is one of the few media outlets that has a presence in all states in Venezuela, and whose news agenda is focused on issues of community complaints, conflicts, and acts of corruption that affect citizens and are of public interest.
A few months ago (in April 2018), Venezuela Inteligente and IPYS Venezuela reported that two of El Pitazo’s domains (elpitazo.com and elpitazo.info) were blocked by CANTV, Digitel, Movistar, Movilnet and Intercable by means of DNS. Recent OONI data not only shows that these domains remain blocked across ISPs, but that a third domain of El Pitazo (elpitazo.ml) has been blocked as well.
CANTV, Digitel and Movistar primarily appeared to block El Pitazo domains by means of DNS, while most measurements collected from Movilnet presented HTTP failures, suggesting potential HTTP blocking. The presence of both DNS lookup errors and HTTP failures may suggest that ISPs employ both DNS and HTTP blocking techniques, or that HTTP failures are caused as a result of DNS blocking techniques not being implemented properly. Alternatively, they could be caused by a congested network, server-side blocking, or if the site in question went down during testing due to a DDoS attack. But these possibilities is ƒrather unlikely, as El Pitazo domains run behind Cloudflare, so they should be quite resistant to failures.
Two months after El Pitazo domains were blocked, IPYS Venezuela reported that independent news outlets La Patilla and El Nacional were blocked as well.
La Patilla was founded in 2010 by the former CEO of Globovision (private Venezuelan TV channel) and is ranked as one of the most visited websites in Venezuela (ahead of other major news websites). Currently, lapatilla.com is accessible, but was temporarily blocked between 6th to 10th June 2018. OONI data collected on 6th June 2018 shows that the site was accessible on Movistar (AS6306), but blocked by state-owned CANTV (AS8048). Lapatilla.com was tested multiple times on CANTV and all measurements presented the same HTTP failures and “generic timeout errors”, suggesting HTTP blocking. CANTV though appears to have unblocked the site by 11th June 2018, as corroborated by all subsequent measurements.
El Nacional is Venezuela’s largest independent newspaper. Having run stories on corruption, official brutality, electoral fraud, protests and other stories critical of the government, the newspaper has received significant government pressure over the last months. Similarly to La Patilla, el-nacional.com primarily appears to be censored by means of HTTP blocking, as suggested by HTTP failures (and “generic timeout errors”) presented in recent OONI measurements. OONI data suggests that the site’s blocked by CANTV and Movilnet, but accessible on Digitel and Movistar.
HTTP failures indicative of blocking have been inconsistent or even intermittent at times. This may suggest that internet censorship is not implemented in a centralized way (i.e. by the same people) or in a way that doesn’t affect all traffic. Venezuela Inteligente and IPYS Venezuela reported that investigative journalism site armando.info was inaccessible as well. This site is known for its critical and extensive reporting on corruption and has been tested fairly regularly across ISPs over the last two years. Most OONI measurements collected up until 12th August 2018 suggested that the site was accessible. But on 13th August 2018, OONI Probe testing revealed that the site was suddenly inaccessible on CANTV, presenting HTTP failures.
To investigate further, IPYS Venezuela and Venezuela Inteligente coordinated a measurement campaign, engaging locals across Venezuela to test armando.info with OONI Probe in various networks and regions of the country. In the evening of 13th August 2018, armando.info was tested on CANTV, Movistar, CIX and Intercable in the following regions: Caracas, Carabobo, Táchira, Aragua, Bolívar, Lara, Portuguesa and Monagas (the table on the right summarizes the results of their testing).
What’s clear from recent OONI Probe measurements (collected on 13th August 2018) is that the potential blocking of armando.info is certainly inconsistent. We can see from the table, for example, that measurements collected from CANTV alternated between being accessible and presenting HTTP failures. And these failures weren’t triggered consistently over time and across regions.
The first CANTV measurements (presenting HTTP failures) in the early evening of 13th August 2018 were collected from Caracas, while the last CANTV measurements presented in the table (showing accessibility) were collected from Táchira. The other accessible CANTV measurement at 6:16 pm was collected from Carabobo. This is particularly interesting, as it may suggest that CANTV doesn’t roll out the same censorship across its network, or that network or configuration issues impacted the accessibility of armando.info.
Venezuela Inteligente and IPYS Venezuela (who are based in Caracas) report that their experience in attempting to access armando.info (on CANTV, Movistar and Digitel) is also inconsistent. As of 13th August 2018, there are moments when they can access the site and there are moments when they can’t. While the armando.info site was inaccessible, as documented by OONI Web Connectivity tests, the server was reachable and accepted TCP connections even as the HTTP exchange failed.
It therefore remains unclear whether armando.info is (or was) intentionally blocked. However, it’s worth highlighting that armando.info uses Google’s “Project Shield“, so server-side issues are unlikely a reason for the observed network anomalies. Further monitoring and testing is required.
Back in 2016, IPYS Venezuela and Venezuela Inteligente reported that a number of blogs critical of the government were blocked. Recent testing shows that the following two sites are currently blocked across ISPs, primarily by means of DNS tampering:
vdebate.blogpost.com (site now defunct) is the blog of an organization whose mission is to “work for the recovery of democracy in Venezuela”. In collaboration with other organizations and volunteers, they defend the human, political and civil rights of Venezuelans.
ovario2.com is a blog that covers Venezuelan issues, expressing political criticism.
Previous measurements collected from CANTV show that alekboyd.blogspot.co.uk (a blog covering corruption and other political issues) was blocked by CANTV by means of DNS tampering, up until (at least) 5th April 2018. The blog though has since been unblocked and is currently accessible.
Venezuela is experiencing the worst economic crisis in its history. The country heavily depends on its oil (it has the largest oil reserves in the world), the revenue of which supported its social programmes and food subsidies. But when the price of oil fell, these programmes became unsustainable and the country plummeted into a food crisis.
Venezuela has established different exchange rate systems for its national currency (the bolivar), with government control on the price of basic goods, which is very high. In light of hyperinflation, coupled with the devaluation of the bolivar in the black market, many Venezuelans are opting for dollars rather than bolivares. But according to the Venezuelan government, this deepens the country’s economic crisis.
Miami-based DolarToday is run by the Venezuelan diaspora and is widely used to track the plummeting black market value of the bolivar. It was first reportedly blocked in 2013. In late 2015, Venezuela’s central bankfiled suit in the US againstdolartoday.com, alleging that the site’s managers “committed cyberterrorism” and “sowed economic chaos” in Venezuela. According to recent OONI measurements, dollartoday.com remains blocked on CANTV.
Zello is a mobile app that serves as a walkie-talkie over cell phone networks. Over the last years, it has been popular among protesters in Venezuela, Ukraine and Russia. During Venezuela’s 2014 protests, the app was reportedly blocked for enabling “terrorist acts“. Recent testing suggests that the service remains blocked by (at least) three ISPs (CANTV, Movistar, Digitel).
Blocking of Tor
The Tor network offers online anonymity, privacy, and censorship circumvention. By bouncing communications across a distributed network of relays, Tor hides its users’ IP addresses. In doing so, Tor users not only have online anonymity, but they can also bypass the blocking of sites and services (since they access them from IP addresses allocated to different countries).
As a result, the Tor network has become a target of censorship in several countries around the world (such as Egypt and Iran), where governments attempt to make circumvention harder and improve their online surveillance capabilities. To bypass Tor censorship, Tor bridges have been built to enable users to connect to the Tor network in censored environments. Tor Browser offers built-in (public) bridges that users can enable. If such bridges are blocked, users can request for (private) custom bridges.
According to recent testing and analysis, Venezuela now also blocks access to the major part of the Tor network and to many public obfs3 and obfs4 Tor bridges. State-owned CANTV (AS8048) appears to have started blocking the anonymity network around 20th June 2018, following months of increased censorship, particularly targeting media websites.
OONI’s Vanilla Tor test is designed to measure the reachability of the Tor network from the local vantage point of the user. If the test does not manage to bootstrap a connection within 300 seconds, access to the Tor network is likely blocked. Similarly, OONI’s Bridge Reachability test measures the reachability of (public) Tor bridges by attempting to successfully bootstrap a connection to them. To confirm the potential blocking with more confidence (and rule out false positives), it’s useful to examine measurements collected from the same network over time.
All measurements collected up until 6th June 2018 were successful, showing that the Tor network was accessible in Venezuela. On 20th June 2018, however, Tor testing started to fail and civil society group Venezuela Inteligente reported the blocking of the Tor network and Tor bridges by CANTV.
Most other measurements collected from 20th June 2018 onwards (from the same network on an almost daily basis) have failed as well, strongly suggesting that state-owned CANTV (AS8048) has been blocking access to the Tor network over the last two months. According to recent mid-August scans from CANTV, around 75% of the Tor network appears to be blocked.
The lack of measurements between 6th to 20th June 2018 prevents from determining the exact date when Tor first got blocked. It’s worth noting though that the blocking probably started on 20th June 2018, since that’s when local civil society group, Venezuela Inteligente (who’s been monitoring internet censorship in Venezuela over the last years), first reported on it.
To investigate further, OONI ran tests from a Raspberry Pi connected to CANTV (AS8048) and performed some experiments examining the blocking of Tor relays. Based on the following, OONI was able to successfully confirm that connections to 74% of well-known IP:Port entities of the Tor network were blocked. The blocking was implemented on the reverse path, so it was hard for the client to distinguish it from server-side blocking:
The client could perform a TCP traceroute to all of the hops except for the last one; the client therefore got ICMP TTL Exceeded responses all the way long, but did not receive SYN-ACK.
The server sees SYN and sends SYN-ACK.
If the server rejects SYN with ICMP Port Unreachable – instead of RST – then the client gets the packet and the Linux TCP stack returns the “connection refused” error.
The server can perform a reverse TCP traceroute back to the client’s IP without anomalies.
Anomalous packet loss is observed on “parasitic” reverse TCP traceroutes, when the traceroute is executed using 5-tuple of existing connection. The anomaly seems to be located within the GlobeNet network, a US-based company that provides one of the backbone internet links to Venezuela’s state-owned CANTV.
The “parasitic reverse traceroute” experiment was designed in the following way: a) the client tried to establish 1000+ connections to the TCP port of Tor relay, b) both “blocked” and “non-blocked” relays were tested, c) the relay was replying with a batch of marked SYN-ACKs with varying TTL fields.
The following chart summarizes the percentage of replies from specific routers and latency to them. It highlights that the network anomaly occurs between two GlobeNet routers.
In addition to Tor blocking, Venezuela Inteligente also reported that access to a large amount of obfs3 and obfs4 bridges (i.e. Tor bridges enabling Tor censorship circumvention) was blocked as well, making it practically impossible to circumvent Tor blocking with built-in bridges. OONI’s bridge reachability measurements corroborate these reports, showing the blocking of many Tor endpoints.
Bridge reachability tests run from CANTV (AS8048) in late June 2018 show a failure rate of around 94% to known Tor bridges. Not all of these failures are necessarily caused by blocking, as some bridges might be offline or unreachable at any given moment. The high percentage of connection failures though is highly indicative of blocking targeted to well-known bridges. Repeated testing in mid-August 2018 showed a similar percentage: 88% of running bridges were unreachable from a CANTV vantage point.
Venezuela Inteligente tested a random sample of unlisted, publicly available bridges from BridgeDB, revealing that the failure rate is around 26% and that all testing to private Tor bridges resulted in successful connections, regardless of the type of bridge (including vanilla, obfs3 and obfs4 bridges). Forward TCP traceroutes towards various accessible Tor relays go via GlobeNet, Level3, Telia and Seabone. This also refutes the hypothesis that Tor blocking depends on uplink (assuming that forward and reverse paths match).
It’s worth highlighting that Tor’s website (torproject.org) has remained accessible in CANTV (and other networks), even though access to the Tor network and obfs4 is blocked.
Further testing on 2nd October 2018 revealed that around 97% of public Tor nodes were reachable with TLS handshake from the vantage point of CANTV. This corroborates local reports on Tor being accessible again. While the precise date of unblocking is quite unclear, Tor Metrics suggest that Tor may have been unblocked on 30th August 2018, since we observe a spike in Tor usage, as illustrated below.
Censorship in Venezuela appears to be a symptom of its deep economic and political crisis, which is considered the most severe crisis in the country’s history. This is strongly suggested by the blocking of numerous currency exchange websites, as well as by the blocking of independent news outlets and blogs that discuss corruption and express political criticism.
The recent blocking of the Tor network (which followed the blocking of news websites El Pitazo and El Nacional) may signify that internet censorship is becoming more dynamic in Venezuela, as ISPs are taking extra steps to reinforce censorship and make circumvention harder. The blocking of the Tor network – which offers online anonymity, in addition to circumvention – might also suggest that the government is attempting to improve its online surveillance capabilities.
While Venezuelan ISPs primarily block sites by means of DNS tampering, they also appear to be implementing HTTP filtering, suggesting a variance in the filtering rules adopted by ISPs. And the variance, both in terms of censorship techniques and censored platforms, across regions and ISPs also indicates that internet censorship is not implemented in a centralized way.
The censorship events identified as part of this study (particularly the blocking of news websites and blogs) contradict the rights outlined by the Inter-American Commission on Human Rights (IACHR) in its report on Standards for a Free, Open and Inclusive Internet. Media censorship and the blocking of blogs limit press freedom and the right to freedom of thought and expression. In examining each right outlined by IACHR, questions around the necessity and proportionality of these censorship events are inevitably raised, particularly in terms of how they relate to human rights.
Venezuela’s political and economic environment is fragile and as events unfold, its internet censorship apparatus may evolve. Continuing to monitor censorship events in Venezuela is therefore essential. This study can be reproduced and expanded upon through the use of OONI Probe and OONI data.
Amid conflict and political turbulence, South Sudan has one of the least developed telecommunications and internet systems in the world. Fifteen Internet Service Providers (ISPs) operate in South Sudan, but the lack of fibre-optic cables and the limited availability of public power hinder connectivity. MTN enjoys the greatest share within the mobile phone market, followed by Vivacell and Zain. Earlier this year however, Vivacell’s license was suspended for not paying USD 60 million in fees.
South Sudan’s Transitional Constitution of 2011 guarantees freedom of expression and press freedom under Article 24, with possible exceptions for public order, safety, or morality. The Article also calls on media to abide by professional ethics. Article 32 of the Transitional Constitution guarantees the right to access official information, with exemptions for public security and personal privacy. The regime though regularly violates media freedom protections in practice, and government officials have engaged in rhetoric that contributes to a hostile environment for the press.
In an attempt to verify reports on the blocking of websites and to examine South Sudan’s internet landscape more broadly, OONI did some network measurement tests in South Sudan.
OONI Probe consists of a number of software tests that scan TCP, DNS, HTTP and TLS connections for signs of network tampering. Some tests request data over an unencrypted connection and compare against a known good value. Others check for HTTP transparent proxies, DNS spoofing, and network speed and performance.
To measure the blocking of websites, OONI started off by carrying out some research to identify South Sudanese URLs to test. They subsequently added these URLs to the Citizen Lab’s test list repository on GitHub, since OONI Probe is designed to measure the blocking of URLs included in these test lists. Over the last few months, OONI primarily ran OONI Probe’s Web Connectivity test (among other OONI Probe tests) in two networks: MTN South Sudan (AS37594) and IPTEC Limited (AS36892).
As part of their testing, they measured the blocking of URLs included in the global (including internationally relevant sites) and South Sudanese (including sites relevant to South Sudan) test lists. Once they collected OONI Probe network measurements from South Sudan,they analyzed them with the aim of identifying network anomalies that could serve as signs of internet censorship.
Last year, media outlets Sudan Tribune and Radio Tamazuj, and independent blogs Nyamilepedia and Paanluel Wel, were reportedly blocked in July 2017. OONI recent testing not only corroborates these reports, but also suggests that these sites remain blocked one year later.
The following table links to network measurements pertaining to the recent testing of each of these sites across two ISPs:
OONI findings suggest that MTN (AS37594) blocks TCP/IP connections to these sites, while IPTEC (AS36892) blocks access by means of DNS tampering. It’s worth noting that both MTN and IPTEC block access to both http://sudantribune.com and http://www.sudantribune.com.
South Sudanese authorities blocked these sites for publishing “subversive content” and stated that the bans would not be lifted until those institutions “behaved well”. Sudan Tribune and Radio Tamazuj are foreign-based media outlets accused of hostile reporting against the government.
Paanluel Wel is a leading blog for the Dinka tribe, known for spearheading tribal political interests for the Dinka people and inciting hatred and violence against the Nuer people and other tribes. Nyamilepedia, on the other hand, is a leading blog for the Nuer tribe, known for promoting Nuer political interests and spearheading hatred against the Dinka and other Nuer who left the rebellion to join the Dinka-led government.
Many other URLs presented network anomalies (such as HTTP failures) as part of our testing, but such anomalies were most likely caused due to poor network performance and transient network failures. This suggests that South Sudanese internet users may encounter challenges in accessing sites in various points in time, even if they’re not intentionally being blocked.
It’s worth highlighting, however, that many of the URLs that OONI tested (including internationally popular and local sites) were found to be accessible in South Sudan during this study. These include sites related to conflict resolution and peacekeeping, such as the United Nations Mission in South Sudan (UNMISS) site.
These measurements clearly show that the Mikrotik HTTP transparent proxy was present last year in the network path to the above sites through South Sudan’s 4G Telecom (AS327786) network. It remains unclear though if this proxy is still in use, since measurements haven’t been collected from this network in recent months.
It’s worth noting that this equipment may potentially be used for implementing internet censorship and/or for caching (the Mikrotik HTTP proxy has this feature) to improve connectivity. Given though that most of these sites were accessible (and the ones that weren’t presented different errors, sometimes triggered as part of anti-DDoS protection), it may be the case that this proxy was primarily deployed for improving connectivity and network performance.
Internet censorship does not appear to be pervasive, but limited to sites that authorities deem to publish “subversive content” and incite violence. This is evident through the blocking of Nyamilepedia and Paanluel Wel, the leading blogs of the Nuer and Dinka tribes who are known to incite violence. OONI data also corroborates the blocking of media outlets Sudan Tribune and Radio Tamazuj, both of which are hosted outside of South Sudan. Local journalists and media organizations though face different (non-digital) forms of censorship.
Nonetheless, the fact that South Sudan has already started implementing internet censorship raises questions as to whether its internet censorship apparatus will expand as internet penetration levels increase and political events unfold. Further research and testing is therefore required to better understand the country’s internet landscape and monitor any new censorship events.
Throughout the testing period, between January 2017 to May 2018, more than 1,000 URLs presented network anomalies. 178 of which consistently presented a high ratio of HTTP failures, strongly suggesting that they were blocked. Rather than serving block pages (which would have provided a notification of the blocking), Egyptian Internet Service Providers (ISP) appear to primarily block sites through the use of Deep Packet Inspection (DPI) technology that resets connections.
In some cases, instead of RST injection, ISPs drop packets, suggesting a variance in filtering rules. In other cases, ISPs interfere with the SSL encrypted traffic between Cloudflare’s Point-of-Presence in Cairo and the backend servers of sites (psiphon.ca, purevpn.com and ultrasawt.com) hosted outside of Egypt. Latency measurements over the last year and a half also suggest that Egyptian ISPs may have changed their filtering equipment and/or techniques, since the latency-based detection of middleboxes has become more challenging.
The chart at the right illustrates the types of sites that presented the highest amount of network anomalies and are therefore considered to more likely have been blocked.
To examine the impact of these censorship events, AFTE interviewed staff members working with some of the Egyptian media organizations whose websites got blocked. They reported that the censorship has had a severe impact on their work. In addition to not being able to publish and losing part of their audience, the censorship has also had a financial impact on their operations and deterred sources from reaching out to their journalists. A number of Egyptian media organizations have suspended their work entirely, as a result of persisting internet censorship.
“Defense in depth” tactics for network filtering
Security experts are probably familiar with the “defense in depth” concept in which multiple layers of security controls (defense) are placed throughout an IT system, providing redundancy in the event that a security control fails. In Egypt, ISPs seem to apply “defense in depth” tactics for network filtering by creating multiple layers of censorship that make circumvention harder.
Back in 2016, OONI uncovered that state-owned Telecom Egypt was using DPI (or similar networking equipment) to hijack users’ unencrypted HTTP connections and inject redirects to revenue-generating content, such as affiliate ads. The Citizen Lab expanded upon this research, identifying the use of Sandvine PacketLogic devices (Sandvine is a company based in Waterloo, Ontario, Canada) and redirects being injected by (at least) 17 Egyptian ISPs.