[liberationtech] Images of Blocking in Different Countries?

[Joss Wright]

On Aug 15, 2012, at 2:46 AM, "Eric S Johnson" <crates at oneotaslopes.org> wrote:

> There are persistent reports that China?s cybercensorship can
> sometimes vary (a little) by ISP, but I?ve never seen this (I?ve only
> been to ~13 of the 34 PRC-defined provinces), and Alkasir hasn?t ever
> detected any such variations. 

I carried out a few experiments last year, looking at regional
variations in the responses of Chinese DNS servers for domains that had
been reported as blocked by Herdict. (Using a list of DNS servers pulled
from the APNIC WHOIS database.)

There are plenty of caveats to the approach I took: DNS server location
doesn't map reliably onto where the user is, DNS poisoning can occur at
border routers, DNS servers can return different responses to different
people, my methods for detecting poisoning were quite crude, etc.

Given all that, I found a lot of variation in the DNS responses across
China. I published a paper about it in FOCI'11 last year, but that
mainly focused on the ethical issues of censorship research:
http://static.usenix.org/events/foci11/tech/final_files/Wright.pdf 

There are some preliminary results and visualisations in this
presentation:
http://www.slideshare.net/josswright/finegrained-censorship-mapping

(Apologies for my terrible GIS skills...)

The approach was quite crude, but does support the hypothesis of
filtering being at least partially decentralised. (Which makes sense for
such a massive project.) 

I'm currently looking at quite a promising approach for a much more
interesting set of experiments based on IP scans rather than DNS. Watch
this space. :)

Joss