4 ms·

The same rise was seen in Turkey last month. The phenomenon is caused by failed reconnect due to DPI-based censorship. See Annex A: https://turkeyblocks.org/201

by indice 10y ago

The same rise was seen in Turkey last month. The phenomenon is caused by failed reconnect due to DPI-based censorship. See Annex A: https://turkeyblocks.org/2016/12/18/tor-blocked-in-turkey-vpn-ban/ https://turkeyblocks.org/2016/12/18/tor-blocked-in-turkey-vp...

indice 10y ago
Turkey Blocks' research and terminal capture suggest each failed attempt causes about 100 broken Tor connections, explaining the spike in metrics where one would really expect a drop.
baseio 10y ago
I don't think DPI is the answer here: o In the case of Turkey, they had more than 10k direct users, the DPI-based censorship yielded a 50k spike, that's a 1:5 ratio, whereas in the case of the UAE it's a freaking 1:30 ratio and it doesn't stop from increasing. o In the case of Turkey, the spike was followed by a spike in bridge use, most using the obfs4 pluggable transport. There's however no such apparent spike for UAE, in fact it's only a nearly constant 300 obfs4 users https://metrics.torproject.org/userstats-bridge-combined.html?start=2017-01-13&end=2017-01-22&country=ae https://metrics.torproject.org/userstats-bridge-combined.htm...
- indice 10y ago
  Actually bridge use has just gone up like Turkey. It took a couple of days after your comment for the UAE charts to get updated. Wait a week more and we'll see if the same noisy zig-zag pattern appears in unbridged connections that showed up in Turkey. That's the smoking gun.
mirimir 10y ago
Right. Tor Metrics for UAE with "Show possible censorship events if available" enabled: https://metrics.torproject.org/userstats-relay-country.html?start=2016-10-23&end=2017-01-21&country=ae&events=on https://metrics.torproject.org/userstats-relay-country.html?... And for Turkey: https://metrics.torproject.org/userstats-relay-country.html?start=2016-10-23&end=2017-01-21&country=tr&events=on https://metrics.torproject.org/userstats-relay-country.html?...

by indice 10y ago

indice 10y ago

Turkey Blocks' research and terminal capture suggest each failed attempt causes about 100 broken Tor connections, explaining the spike in metrics where one would really expect a drop.

baseio 10y ago

I don't think DPI is the answer here: o In the case of Turkey, they had more than 10k direct users, the DPI-based censorship yielded a 50k spike, that's a 1:5 ratio, whereas in the case of the UAE it's a freaking 1:30 ratio and it doesn't stop from increasing. o In the case of Turkey, the spike was followed by a spike in bridge use, most using the obfs4 pluggable transport. There's however no such apparent spike for UAE, in fact it's only a nearly constant 300 obfs4 users https://metrics.torproject.org/userstats-bridge-combined.html?start=2017-01-13&end=2017-01-22&country=ae https://metrics.torproject.org/userstats-bridge-combined.htm...

indice 10y ago
Actually bridge use has just gone up like Turkey. It took a couple of days after your comment for the UAE charts to get updated. Wait a week more and we'll see if the same noisy zig-zag pattern appears in unbridged connections that showed up in Turkey. That's the smoking gun.

mirimir 10y ago

Right. Tor Metrics for UAE with "Show possible censorship events if available" enabled: https://metrics.torproject.org/userstats-relay-country.html?start=2016-10-23&end=2017-01-21&country=ae&events=on https://metrics.torproject.org/userstats-relay-country.html?... And for Turkey: https://metrics.torproject.org/userstats-relay-country.html?start=2016-10-23&end=2017-01-21&country=tr&events=on https://metrics.torproject.org/userstats-relay-country.html?...