search

ooni / probe

OONI Probe network measurement tool for detecting internet censorship
https://ooni.org/install
BSD 3-Clause "New" or "Revised" License
751 stars 142 forks source link

Consider implementing control for more experiments #1796

Open bassosimone opened 2 years ago

bassosimone commented 2 years ago

Facebook and whatsapp are currently down. This situation may generate false positives in the apps. We are able to see backend-wise that it's a huge outage, but a user would not see this fact in the app. So, maybe having a control measurement could be a way to communicate clearly to users they're not seeing an anomaly.

An opposite approach could be to have the backend immediately score and return also information from other probes to a probe, so the probe knows it's looking at a large outage. Is this easy and/or practical to do?

Maybe there are combined approaches when we can do both?

This summarises a conversation I just had with @agrabeli on this topic. We should discuss this topic further!

agrabeli commented 2 years ago

I generally think it could be useful to include control measurements in all OONI Probe tests because they can serve as a baseline for comparison, which is not only useful for detecting censorship (when the results differ), but also for inferring cases when services are down (when anomalies are seen in both the control + experiment measurements).

Alternatively, this could be done in the backend (to check for global accessibility of a tested service), but based on the current structure, you would only get these results on OONI Explorer (rather than in the client).