[tor-project] OONI Monthly Report: March 2022


OONI’s March 2022 status report is shared below.

# OONI Monthly Report: March 2022

Throughout March 2022, the OONI team worked on the following sprints:

  • Sprint 60 (1st - 13th March 2022)
  • Sprint 61 (14th - 31st March 2022)

Our work can be tracked through the various OONI GitHub repositories: https://github.com/ooni

Highlights are shared in this report below.

## Published report on new blocks in Russia amid war in Ukraine

We published a research report on new censorship events in Russia amid the war in Ukraine, based on OONI data analysis.

You can read the report here: https://ooni.org/post/2022-russia-blocks-amid-ru-ua-conflict/

Below we share some key findings:

  • Media censorship. Russia recently started blocking access to independent Russian news media websites (such as Dozhd and New Times) and foreign news media websites (such as BBC, Deutsche Welle, Russian version of Voice of America, and RFE/RL service websites).

  • Blocking of a website about captured and killed Russian soldiers. Russia blocked access to 200rf.com, which was created by representatives of the Ministry of Internal Affairs of Ukraine to share information about captured and killed Russian soldiers in Ukraine.

  • Centralized throttling of Twitter. OONI data suggests that access to twitter.com was throttled in Russia between 26th February 2022 to 4th March 2022. As the throttling of twitter.com seems to have stopped across all ISPs in Russia at the same time (~08:00 on 4th March 2022 UTC), it appears to have been centralized. Moreover, interference to twitter.com appears to have changed from throttling to blocking (through the injection of a RST packet) on 4th March 2022.

  • Blocking of Facebook and Twitter. OONI data shows that access to facebook.com and twitter.com was blocked in Russia by 4th March 2022 (primarily through the injection of RST packets).

  • Decentralized censorship. Every Internet Service Provider (ISP) in Russia is responsible for implementing government-mandated blocks independently. As a result, we observe variance in how internet censorship is implemented across Russia, as blocks are not implemented on all networks in the country, and different ISPs adopt different censorship techniques. Some ISPs implement blocks through the use of multiple techniques at the same time, making circumvention harder.

  • Different censorship techniques. To block websites, OONI data shows that Russian ISPs adopt the following censorship techniques (beyond throttling):

  • DNS manipulation, redirecting in some cases to blockpages

  • HTTP man-in-the-middle, serving blockpages

  • TLS man-in-the-middle

  • Injection of a RST packet after the ClientHello during the TLS handshake (most prevalent)

  • Timing out the session after the ClientHello during the TLS handshake

  • Closing the connection after the ClientHello during the TLS handshake

These findings received press coverage from Wired (https://www.wired.com/story/russia-splinternet-censorship/), Business Insider (https://www.businessinsider.com/splinternet-in-pictures-what-russian-internet-looks-like-right-now-2022-3), VPNCompare (https://www.vpncompare.co.uk/ooni-russia-censorship-propaganda/), and Protocol (https://www.protocol.com/russian-internet-crackdown).

This research report also resulted in a donation to OONI from Pantheon, who (along with several other open web leaders) published a statement in support of OONI’s mission: https://pantheon.io/blog/pantheon-rises-support-open-web

## OONI Probe Mobile

We released OONI Probe Android 3.6.0: https://github.com/ooni/probe-android/releases/tag/v3.6.0

This release includes the following improvements:

  • New Tor Snowflake experiment
  • New DNS Check experiment
  • RiseupVPN test temporarily disabled
  • Links to OONI blog posts and research reports
  • User-initiated tests start off minimized
  • Improved progress bar and test runtime counter
  • Measurement engine synced with OONI Probe CLI 3.14
  • Bug fixes and improvements

We followed-up with an OONI Probe Android 3.6.1 release (https://github.com/ooni/probe-android/releases/tag/v3.6.1) with a bug fix.

On OONI Probe Android, we also worked on:

On OONI Probe iOS, we worked on:

## OONI Probe Desktop

We released OONI Probe Desktop 3.7.0:

Leading up to this release, we made several release candidates for extensive testing and quality assurance (https://github.com/ooni/probe-desktop/releases).

Notably, OONI Probe Desktop 3.7.0 includes the following improvements:

## Expanding censorship measurement methodologies

We wrote the third prototype of Websteps (codenamed “winter 2022”), our new experiment for measuring the blocking of websites. Rather than modifying the existing implementation, we chose to fork relevant packages of the probe-cli repository into a separate repository (https://github.com/bassosimone/websteps-illustrated/), which allowed us to iterate faster and perform more changes without the risk of breaking the existing codebase or splitting each change into easy-to-review, small patches.

We achieved quite a lot of progress in this fork, ranging from improvements to the underlying libraries to significant progress in terms of the final version of Websteps. The repository itself contains extensive documentation about what changed. Here we only provide some highlights:

  • We wrote a complete specification of Websteps that not only describes the measurement methodology, but also the analysis algorithms;

  • We tested this Webstep implementation on a few networks in China, Iran, and Italy;

  • We improved the conceptual model of Websteps, which now allows us to easily express what we are measuring, with more clarify with regards to the measurement algorithm;

  • We developed a new integration testing framework for Websteps that (rather than relying on simulating censorship using Linux networking capabilities as we currently do) is based on re-running actual measurements collected on censored networks;

  • We significantly improved the DNS implementation by adding support for PTR, CNAME, and NS records, fixing bugs and adding performance improvements;

  • We implemented a DNS Ping extension for Websteps that only runs after specific DNS measurements to collect extra information (e.g. multiple DNS replies) that is rather useful to collect evidence of DNS-based censorship;

  • We wrote a Websteps measurement library in Python that allows us to process and visualize measurements and manage integration test cases.

As a next step, we are engaging external researchers for an additional round of review of the experiment. Following this review (and any further changes based on feedback), we will start the process of integrating the changes into the OONI Probe CLI repository.

## Launch of Circumvention Tool Reachability Dashboard

We launched a new dashboard (https://explorer.ooni.org/chart/circumvention) that presents findings from all censorship circumvention tool test results (including the new Snowflake test) on OONI Explorer.

This new Circumvention Tool Reachability Dashboard presents aggregate views of real-time OONI data collected from the reachability testing of circumvention tools around the world. Through this interactive dashboard, users can check whether circumvention tools work and monitor their potential blocking around the world based on real-time OONI data. As the dashboard presents charts based on aggregate views of OONI data, these charts can potentially support relevant outreach and advocacy efforts.

## OONI Measurement Aggregation Toolkit (MAT)

Throughout March 2022, we made significant progress on the Measurement Aggregation Toolkit (MAT) in preparation for the public launch in April 2022. Based on community feedback and extensive testing, we worked on making several improvements (https://github.com/ooni/explorer/issues?q=is%3Aissue+label%3AMAT+is%3Aopen).

Notably, we added support (to both OONI Explorer Search and the MAT) for filtering Web Connectivity measurements based on website categories (https://github.com/ooni/explorer/issues/288). This enables OONI Explorer and MAT users to filter measurements based on website categories (such as news media) and to easily discover thematic website censorship, such as the blocking of news media websites, human rights websites, and many other categories of websites (we use the Citizen Lab test list category codes).

To ensure compatibility between the MAT and the OONI Explorer Search Tool, we added support for domain-based filtering (https://github.com/ooni/explorer/issues/664) in the MAT (as opposed to filtering by URL). This enables users to get aggregate views of all measurements pertaining to a domain, rather than having to explore measurements separately for each URL corresponding to a domain.

We made a series of improvements to the tables and charts generated through the MAT, and we improved the labels and copy used in various features (such as buttons and drop-down menus) of the platform. We also worked on adding code-level documentation to highlight critical parts of the codebase. Following all the improvements, we created a preview link for further testing leading up to the launch.

## OONI backend

During March 2022, OONI backend activities mainly involved the reprocessing of legacy OONI measurements to improve data quality. As part of this, we are working on improving the measurement re-processor (https://github.com/ooni/backend/issues/569). To monitor the progress, we developed new dashboards on Jupyter.

As part of the migration to Clickhouse, we implemented fetching measurements from the spool directory served by Nginx and developed speed improvements for the API’s get_measurement. We benchmarked ways to upload database tables from Clickhouse to S3. This could be used in the future to automatically publish database tables to external researchers.

We made improvements to our URL prioritization algorithm (https://github.com/ooni/backend/issues/562) to ensure that popular social media URLs (which are frequently blocked around the world) are always prioritized for testing globally. As part of our work to move probes to the new test helper, we created a monitoring dashboard, we added performance metrics to the API, and we forwarded some of the production traffic to the new hosts.

## Improving OONI data analysis capabilities

As part of our efforts to consolidate block page fingerprints that we find with those collected by the Citizen Lab (https://github.com/ooni/backend/issues/516), we opened a pull request on the Citizen Lab repository for filtering annotations with DNS signatures from Russia (https://github.com/citizenlab/filtering-annotations/pull/1). These were extracted as part of our OONI data analysis efforts when investigating the emergence of new blocks in Russia following the invasion of Ukraine.

## Creating an online OONI Training Course for Advocacy Assembly

During March 2022, we worked on completing all the materials for the online OONI training course that we are creating for Small Media’s Advocacy Assembly project (https://www.advocacyassembly.org/).

In particular, we:

  • Created 11 screencasts for various OONI tools;
  • Wrote copy for 10 training slides (which are included in several chapters of the course);
  • Created quizzes for 3 training chapters;
  • Coordinated with partners on creating case study videos.

## Test list updates

In response to censorship events, we contributed updates to the following Citizen Lab test lists:

We also reviewed and merged several test list pull requests contributed by community members.

## Notable community use of OONI Probe and OONI data

### ISOC blog posts on blocks in Russia

As part of their “Internet Perspectives: Ukraine and Russia” blog series (https://pulse.internetsociety.org/blog/internet-perspectives-ukraine-and-russia), Internet Society (ISOC) published the following 2 blog posts based on OONI data:

### Article on network congestion and blocks in Iran

Digiato (an Iranian technology news portal) published an article discussing network congestion in Iran, where Iranian experts discuss intermittent blocks in the country, citing OONI data (https://digiato.com/article/2022/02/23/why-iran-internet-is-congested). The article also features a screenshot of a chart generated from OONI’s Measurement Aggregation Toolkit (to demonstrate the intermittent blocks).

## Community activities

### OONI presentation at Spacemesh Meetup

Spacemesh is a cryptocurrency organization that also has a social causes program (https://spacemesh.io/blog/spacemesh-social-causes-program/). As part of this program, they focused on “Fighting Censorship” and engaged their global community with using OONI Probe (https://spacemesh.io/blog/winning_together-raise-the-flag/#challenge-overview). In particular, they encouraged their community to install and run OONI Probe, and to practice several OONI Probe hands-on exercises (“challenges”) that we shared with them. We were also invited to give an OONI presentation for Spacemesh.

On 8th March 2022, OONI’s Maria presented OONI at a Spacemesh meetup. The presentation can be viewed here: https://vimeo.com/688867593

Spacemesh subsequently wrote about OONI in their weekly newsletter: https://smesher.substack.com/p/weekly-inspiration-for-humans-in

### OONI Community Meeting

On 29th March 2022, we hosted the monthly OONI Community Meeting on our Slack channel (https://slack.ooni.org/), during which we discussed the following topics:

  1. Updates from the OONI team (new research and OONI Probe releases)

  2. Cambodia’s National Internet Gateway

  3. Addressing false positives in OONI measurements

  4. Simplifying the OONI FAQ

## Userbase

In March 2022, 40,833,632 OONI Probe measurements were collected from 3,155 AS networks in 167 countries around the world.

This information can also be found through our measurement stats on OONI Explorer (see chart on “monthly coverage worldwide”): https://explorer.ooni.org/

~ OONI team.