Media censorship in Azerbaijan through the lens of network measurement – July 2021 report

On July 1, Azerbaijan Internet Watch launched a new report titled “Media censorship in Azerbaijan through the lens of network measurement”. The report was prepared in partnership with the Open Observatory of Network Interference (OONI) summarising key findings from network measurements conducted between January 2020 to May 2021. The full report can be accessed here.

About the report

In light of reports on the blocking of websites in Azerbaijan, the Open Observatory of Network Interference (OONI) and Azerbaijan Internet Watch (AIW) formed a partnership to collaborate on researching internet censorship in the country. Over the past year, OONI and AIW have collaborated on collecting and regularly analyzing censorship measurements from Azerbaijan, while providing timely updates through reports. In this report, we share findings from our analysis of OONI network measurements collected from Azerbaijan between 1st January 2020 to 1st May 2021. The aim of this study is to document and increase the transparency of internet censorship in Azerbaijan through the analysis of empirical network measurement data.

Key findings

  • Blocking of independent news media and circumvention tool websites. Throughout the testing period, several independent news media and circumvention tool sites presented HTTP failures caused by connection timeouts. This suggests the potential use of Deep Packet Inspection (DPI) by ISPs in Azerbaijan.
  • Attempts to block Tor and Psiphon. ISPs in Azerbaijan attempted to block Tor and Psiphon amid the 2020 Nagorno-Karabakh war. However, both attempts appear to have been quite ineffective. 
  • Temporary blocking of social media amid 2020 Nagorno-Karabakh war. Between September 2020 to November 2020, several social media websites presented the same HTTP failures (as news media and circumvention tool sites), while the testing of WhatsApp and Telegram presented signs of TLS level interference.
  • Variance of censorship across networks. ISPs in Azerbaijan appear to be adopting similar censorship techniques. However, censorship varies from network to network, as different ISPs block different websites and apps at different moments in time.

Blocked news media websites

Several independent news media websites presented signs of blocking in Azerbaijan throughout the analysis period.

These domains include:

  1. `azerbaycansaati.tv`
  2. `criminal.az` 
  3. `www.24saat.org` 
  4. `www.abzas.net` 
  5. `www.azadliq.info`
  6. `www.azadliq.org`
  7. `www.gununsesi.info`
  8. `www.gununsesi.org`
  9. `www.kanal13.tv`
  10. `www.meydan.tv`

OONI data also suggests that the site (`www.occrp.org`) of the Organized Crime and Corruption Reporting Project (OCCRP) and the site (`www.rferl.org`) of RadioFreeEurope/RadioLiberty (RFE/RFL) were blocked in Azerbaijan as well. The blocking of the OCCRP site reportedly began in September 2017, following the publication of a major investigation (“Azerbaijani Laundromat”) into corruption, bribery, and money laundering in which powerful figures were allegedly involved. The blocking of the RFE/RFL website also reportedly began in 2017, following an Azerbaijani court order which RFE/RFL described as “another blatant attempt at silencing its reporting in the country”

Blocking of social media amid 2020 Nagorno-Karabakh war

Amid the 2020 Nagorno-Karabakh war, OONI data shows that access to several social media websites and apps was blocked in Azerbaijan. The following chart, limited to social media websites that presented signs of blocking between February 2020 to May 2021, aggregates OONI measurement findings collected from 4 AS networks in Azerbaijan.

Blocking of social media websites in Azerbaijan based on OONI data (collected between March 2020 to May 2021), https://explorer.ooni.org/search?since=2020-01-01&probe_cc=AZ&test_name=web_connectivity&only=anomalies

As is evident from the above chart, most of these social media websites primarily presented signs of blocking during the 2020 Nagorno-Karabakh war (between 27th September 2020 to 10th November 2020), but were found accessible when tested (on several networks in Azerbaijan) in the months before and after the war. Notably, most anomalous measurements presented HTTP failures (because the HTTP requests timed out), similarly to the blocking of news media websites (discussed previously). This provides a stronger indication that these social media websites were blocked, particularly since ISPs often use the same censorship technique(s) to block a variety of different websites. 

Blocked circumvention tool sites

Numerous circumvention tool websites presented signs of potential blocking when tested (on up to 3 AS networks) in Azerbaijan between February 2020 to May 2021, as illustrated through the following chart.

Blocking of circumvention tool websites in Azerbaijan based on OONI data (collected between January 2020 to May 2021), https://explorer.ooni.org/search?since=2020-01-01&probe_cc=AZ&test_name=web_connectivity&only=anomalies

Similar to the blocking of news media and social media websites, we observe that the testing of circumvention tool websites often resulted in HTTP failures caused by connection timeouts. This consistency in terms of failures, observed on several AS networks over the period of a year, strongly suggests blocking of these circumvention tool websites. As testing coverage increased from January 2021 onwards, we observed an increased volume of anomalous measurements, most presenting the same HTTP failures.

Conclusion

Press freedom appears to be quite limited in Azerbaijan, as suggested by the blocking of several independent news media websites in the country. These media websites presented signs of blocking throughout their testing (on several local AS networks) between January 2020 to May 2021 (corroborating past reports on the blocking of media websites in Azerbaijan), with recent OONI measurements suggesting that their blocking remains ongoing

Potentially in an attempt to prevent the circumvention of media censorship, ISPs in Azerbaijan appear to have blocked access to a number of circumvention tool websites over the last year as well. It remains unclear, however, if the apps of these circumvention tool sites were also blocked (as they were not tested as part of this study); and even if they were, it’s possible that local internet users may have been able to use them nonetheless, given that circumvention tools often include in-built circumvention techniques for evading censors. 

Amid the 2020 Nagorno-Karabakh war, ISPs in Azerbaijan appear to have attempted to block the Tor and Psiphon circumvention tools. Yet, these attempts were likely ineffective, given that both tools have in-built circumvention techniques and fallback options for circumventing blocks. In Tor measurements, we observe that most ISPs did not block all tested Tor directory authorities, suggesting that it was possible to use Tor nonetheless (as also indicated by the spike in Tor usage from Azerbaijan during that period). Similarly, many Psiphon measurements during this period were successful, suggesting that it may have been possible to use the Psiphon VPN on many networks.

Several social media websites (such as `www.facebook.com` and `www.youtube.com`) and apps (primarily WhatsApp and Telegram) presented signs of blocking between September 2020 to November 2020, which coincides with the 2020 Nagorno-Karabakh war. It is, therefore, possible that ISPs may have increased efforts to block circumvention tools (during this period) in an attempt to prevent the circumvention of social media censorship.

Interestingly, we observe similar censorship techniques adopted by different ISPs in Azerbaijan, but variance in terms of which internet services are blocked by ISPs over time. In other words, we see ISPs blocking websites and apps in similar ways (seemingly using the same censorship techniques), but different ISPs block access to different websites and apps (and sometimes this varies at different moments in time). 

Throughout the testing period, independent news media and circumvention tool websites presented HTTP failures caused by connection timeouts, suggesting the potential use of Deep Packet Inspection (DPI) by ISPs in Azerbaijan. Similarly, when social media websites were temporarily blocked amid the 2020 Nagorno-Karabakh war, their testing also presented HTTP failures caused by connection timeouts. This suggests that most ISPs in Azerbaijan block websites using similar (if not the same) censorship techniques.

Both WhatsApp and Telegram presented signs of TLS level interference on several different AS networks in Azerbaijan amid the 2020 Nagorno-Karabakh war. In the case of WhatsApp, the HTTP requests to `web.whatsapp.com` succeeded, while the HTTPS requests failed (during the TLS handshake), which could be an indication of SNI-based filtering. In the case of Telegram, we see that both HTTP and HTTPS requests to `web.telegram.org` timed out. 

As media censorship (and the blocking of circumvention tool websites) appears to be ongoing in Azerbaijan, there is a need for further testing to evaluate these censorship events in more depth over time. The temporary blocking of social media amid the 2020 Nagorno-Karabakh war also suggests that new censorship events can emerge in Azerbaijan, as political events evolve. 

This study was carried out through the use of open methodologies, free and open source software, and open data, enabling independent third-party verification of our research findings. We encourage researchers to expand upon this study by running OONI Probe and analyzing OONI measurements from Azerbaijan.    

OONI measurements indicate censorship remains

In its most recent measurement report, the Open Observatory of Network Interference [OONI] concludes that “while social media censorship in Azerbaijan appears to have been lifted, the media censorship remains.” These and other findings are based on the recent measurement report produced in partnership with OONI. 

Here are some highlights.

Blocked websites

The news websites that presented signs of blocking in Azerbaijan (between December 2020 to February 2021) include:

🛑 azerbaycansaati.tv – at the time of blocking azerbaycansaati.tv in 2017, the Government of Azerbaijan claimed “a number of articles published” by the news website “included calls aimed at ‘forcible change of the constitutional order,’ ‘organization of mass riots,’ and other illegal activities.” 

🛑 www.24saat.org – a more detailed report about how 24saat.org was blocked can be found in this report, published by Qurium in 2017. 

🛑 www.abzas.net – DDoS attacks against abzas.net commenced on January 12, 2017, and lasted for eight days. During five full consecutive days, the website remained inaccessible until it was finally migrated to VirtualRoad.org’s secure hosting infrastructure.

🛑 www.azadliq.info – as a hosting provider for azadliq.info Qurium published this report about initial signs of blocking against this online news platform. The website was attacked numerous times according to documentation and forensic reports by Qurium. The technology deployed in these DDoS attacks was Allot and Sandvine DPI gear.

🛑 www.azadliq.org – the news website which represents the Azerbaijan Service for Radio Free Europe, was blocked on March 27, 2017. 

🛑 www.gununsesi.org – signs of DPI technology used in blocking gununsesi.org were once again documented by Qurium.

🛑 www.kanal13.tv – was among blocked websites in 2017 while its editor prosecuted [charges were dropped three years later.] 

🛑 www.meydan.tv – was also among the websites that were blocked in 2017 together with azerbaycansaati, azadliq.info and others. 

🛑 www.occrp.org – in response to the leaks about Azerbaijan Laundromat published by the Organized Crime and Corruption Research Project [OCCRP], the government of Azerbaijan suspended access to OCCRP’s website.

There is no official data on the number of blocked websites in Azerbaijan. The Ministry of Communication, High Technologies and Transportation has so far failed to provide accurate lists. This in itself is a violation of Article 13.3.6 of the Law on Information, Informatisation and Access to Information, which requests the Ministry to prepare a list of blocked websites if it has blocked access to a resource and the court upheld this decision.

In July 2018, the Prosecutor General’s Office launched criminal investigations against four news websites: criminal.az, bastainfo.com, topxeber.az and fia.az. The former two were accused of “knowingly spreading false information,” while the latter two were accused of “spreading unfounded, sensational claims in order to confuse the public.” Criminal.az is an independent website, known for its coverage of crime-related news, while bastainfo.com is affiliated with the opposition party Musavat. The latter two are run-of-the-mill online news websites.

In addition to the usual suspects, video streaming service Vimeo appeared to be briefly blocked during the testing coverage:

Circumvention

Several circumvention tool websites appear to have been interfered with in Azerbaijan during the testing period, as illustrated below:

The good news are that access to social media sites and apps was restored during the testing period. The following chart shows that while WhatsApp and Telegram were blocked in November 2020, both apps (along with Facebook Messenger) have been accessible in recent months:

How you can help?

If you are interested in contributing to these tests you are welcome to try the following instructions

Azerbaijan September – November 2020 OONI Measurement Results

On September 27, the government in Azerbaijan introduced a series of restrictions on Internet access as a result of military operations in Nagorno Karabakh. Users in Azerbaijan were left with limited Internet access while access to a number of social media platforms, as well as communication apps, were restricted. 

Azerbaijan Internet Watch, and OONI collected data based on OONI measurements from Azerbaijan. Below is the report looking at data between September 2020 to November 2020. 

According to collected OONI data, Azerbaijan experienced the blocking of independent media websites as well as signs of potential circumvention tool site blocking. However, further OONI Prote testing is required to confirm these censorship events. Azerbaijan Internet Watch continues to monitor the situation. 

Starting on September 27, a number of social media services indicated signs of blocking in Azerbaijan. Among these were:

Whatsapp: between 27th September 2020 to 11th November 2020 , all OONI Probe WhatsApp tests showed signs of WhatsApp blocking. This is illustrated through OONI Probe WhatsApp measurements collected from multiple local networks in Azerbaijan. Previously according to the same measurements, no such instances were documented. The measurements indicated that attempted connections to WhatsApp’s registration service and web interface (web.whatsapp.com) failed. In some cases HTTP request to web.whatsapp.com succeeded while HTTPS request failed as illustrated below. 

This could be an indication of SNI based filtering of WhatsApp. We also observe that the tls_handshakes field presented failures, further suggesting that access to WhatsApp was blocked in Azerbaijan by means of SNI based filtering.

This pattern was seen across multiple ISPs in Azerbaijan between 27th September 2020 to 11th November 2020.

Telegram: Similarly to the testing of WhatsApp, OONI measurements presented signs of Telegram blocking in Azerbaijan between 27th September 2020 to 11th November 2020. But unlike WhatsApp, a few Telegram tests during this period were successful. This is illustrated through OONI Probe Telegram measurements collected from multiple local networks in Azerbaijan between 1st September 2020 to 27th November 2020. The following table illustrates OONI measurements collected from the testing of Telegram and WhatsApp on 5 different networks in Azerbaijan between September 2020 to November 2020.

We not only observed similarities between the date range of potential blocking (of WhatsApp and Telegram), but we also saw potentially similar censorship techniques, as the HTTP requests to Telegram Web (web.telegram.org) timed out as well. Similarly to WhatsApp, we observed a timeout in the TLS handshake, suggesting TLS level interference of Telegram as well. As a result, it is possible that internet users in Azerbaijan couldn’t use the Telegram and WhatsApp mobile apps (on the tested networks) during this time period (between 27th September 2020 to 11th November 2020), even though connections to the tested app endpoints succeeded.

Social media sites: apart from WhatsApp and Telegram, several social media websites presented signs of blocking as well, starting from early October 2020. These include:

It is however, important to note these sites’ limited testing coverage limits Azerbaijan Internet Watch and OONI’s confidence with respect to their blocking, and they have not been tested more recently especially after the conflict ended.

AIW and OONI continue monitoring the situation with blocking.