search

ooni / backend

Everything related to OONI backend infrastructure: ooni/api, ooni/pipeline, ooni/sysadmin, collector, bouncers and test-helpers
BSD 3-Clause "New" or "Revised" License
50 stars 29 forks source link

Fix error annotation in Web Connectivity measurements #454

Open agrabeli opened 4 years ago

agrabeli commented 4 years ago

A data analyst at the Citizen Lab highlighted that many OONI Web Connectivity measurements display an "error" annotation (instead of the "anomaly" annotation).

For example, see: https://explorer.ooni.org/measurement/20191118T063541Z_AS15802_ocKS0RbSg8GE1oLahQKYeAozvsxM3HglG8f3xMRkrWVOd4ajtN?input=http://gayguide.net/

It seems that this failure is annotated due to a failure on the control. However, the raw data clearly shows that a blockpage was served on the local network.

How can we circumvent cases when the control fails, but there is a clear sign of blocking on the local network?

This may be possible in countries like Indonesia, where block pages are served and we have their fingerprints.

agrabeli commented 4 years ago

This bug seems to be caused because OONI Explorer presents an error, irrespective of whether there is a known block.

In cases like Indonesia or UAE, where they serve block pages and we have their fingerprints, it would be great if we can find a way to circumvent the control failure and display "confirmed blocked" nonetheless.

This though should probably happen in the API analysis logic, since OONI Explorer merely fetches measurements from the OONI API.

@bassosimone says:

This seems an issue with the pipeline. When processing or reprocessing a measurement the pipeline should check whether the body contains a known fingerprint regardless of whether the control measurement failed. [And also adds] if the pipeline is doing that already, then this is an issue with either the API or explorer. But let's check with the pipeline first.

sarathms commented 3 years ago

@agrabeli This issue seems to be resolved at least for the example measurement linked in the description which now shows a confirmed blocked page.

Can we considered this issue fixed then?