wiki:org/meetings/2018MexicoCity/Notes/OONIRapidResponse

How can we respond to censorship events around the world?

  • They often take place at the weekend or out-of-hours
  • It’s important for OONI to do this, because there are lots of people involved and need someone to collect information
  • Data can help to make change
  • Timeliness is everything
  • Currently no alert system in place, so censorship events are discovered through analysis of specific countries, or when it’s reported in the news. Primary route remains the community - partners, #KeepItOn campaign mailing list (currently the best resource), WhatsApp groups

Steps taken

  1. Determine if measurements are available
  2. Are the measurements useful? If it’s a web site, is it even being tested?
  3. Do we know people on the ground? If not, we need to identify locals who might be willing to help, then rapidly train them on how to do the testing.
  4. Get new info and analyse.

#KeepItOn have used info in their reports Work with lawyers in Uganda and other countries

Some events are predictable, eg forthcoming elections in Cameroon. Have been co-ordinating with local groups to set up tests.

This cannot scale to meet community needs, there are too many censorship events around the world. We need more training / education in OONI, Tor. It’s difficult to train over WhatsApp.

Need closer collaboration with community, so that they can train locally and OONI isn’t a bottleneck.

Anomaly detection is being done on a local basis in certain places. It’s an important part of the rapid response need, can be done in conjunction with the Access Now helpdesk and other similar resources.

The pipeline, for some of the events, should allow for pre-emptive work so that it can be more easily captured on election day. OONI have a calendar for doing that, but some of the censorship can take place after the election or at another unpredicted time.

There is a plan to create a censorship alert system. Tor can work with research groups if OONI provides data and research questions to be analysed / interpreted.

It would be useful to know the specific context in each different country.

OONI Explorer is not currently a live representation of what’s happening. It will be revamped in the next few months.

False positives: There’s a danger that journalists could pick up on these and report on them, which would be bad for everyone.

OONI is checking complementary data when something appears blocked - eg Does the site have a high global failure rate? Is the site limiting access to users in a country?

Can we crowdsource information? People could submit evidence / stories themselves when an event is detected. For example, there is a Metrics timeline on the wiki.

Idea: Can you get a bot in a WhatsApp channel that cuts into a conversation when a blocking event is discussed? Could be a natural language problem - “Is Twitter blocked for you? No” would trigger the bot. It could be better to have the ability to trigger a bot when something appears blocked. ResistBot in the US is an example of a successful bot.

Anomaly detection: Keep working on it to make censorship events more predictable. When something happens that wasn’t predicted, find a way to predict it next time.

OONI has a global list, and a country-specific list. Mobile probe only tests the global list at this point, but that is due to change.

Test lists are imperfect. They can’t be complete, and it’s difficult to get people to contribute over long periods of time.

Brave collect browsing history in a self-claimed privacy-preserving way, could OONI learn from that?

How should OONI work with Tor? Collaborate with: Tor Metrics (for comparative measurements), Comms (for publicising), Community team (for training)

Educational challenge: Teaching people how to use Tor could lead to Tor itself being blocked. Educational material needs to help people understand how to deal with more than a site being blocked.

Censorship measurement training should include training about how the internet is censored, what the different kinds of censorship mean, etc.

Tor Browser usability: Should there be a button in TB to try every method to connect, if vanilla TB fails? How could that work? What happens next time the user connects? Can they revert to vanilla TB if an esoteric method is chosen?

Censorship alerts: Tor Metrics records context for false positives, so they can learn from it.

OONI was originally set up to record anomalies, not interpret them. Now, they will say something is censorship if it is considered intentional blocking with a block page returned. Otherwise, they will report (with / via partners) that certain anomalies happened over a given period of time.

Last modified 2 months ago Last modified on Oct 1, 2018, 12:01:16 AM