[tor-talk] Tor Reject Exit Policy to Syria, Iran, Belarus, Saudi Arabia, Moldavia, Qatar, etc?

Roger Dingledine arma at mit.edu
Sun Oct 2 20:29:47 UTC 2011


On Sun, Oct 02, 2011 at 09:23:15PM +0200, Fabio Pietrosanti (naif) wrote:
> i made some dirty stats by looking at all "reject" lines in Exit
> Policies from cached-descriptors.
> 
> Then i sorted them, uniq'ed them and made a geoip look on top of them.

Here are some suggestions to get better accuracy:

A) You're overcounting relays by looking at every descriptor in your
cached-descriptors file. One simple fix, assuming your Tor is just a
client, is to look at the fingerprint line in the router descriptor,
and discard all but the newest descriptor for a given fingerprint.

B) Most relays reject their own IP address in their exit policy (since
it's common that there's some service that trusts connections from that
IP address, e.g. a linksys router that lets you log in). So you should
read the IP address out of the router line, and ignore reject lines
for that IP address.

I expect once you correct for A and B your numbers will look very
different. All it takes is one guy in SA running a relay and having his
IP get changed 15 times and you suddenly have 15 reject lines to SA in
your list.

The next step is that you'll want to think about bandwidth and uptime
of relays -- SA has a heck of a lot of Tor users, and every once in a
while one of them clicks the "make me a relay" button, but it doesn't
stay up very long or push much bandwidth.

I guess how much you should read out of tiny transient relays
depends on what you're trying to learn by these stats.

> It's interesting to notice that the amount of IP blocked in Tor Exit

Another thing you should know is that some of these addresses
you've got aren't IP address, but rather netblocks. So
reject 7.0.0.0/8:*
is quite different from
reject 7.0.0.0:*

> If someone would like to check it, eventually using some python magic
> with a parsers of cached-descriptors fine, it would be interesting to
> see the results and/or resulting tool to make the analysis.

You might like
https://gitweb.torproject.org/tor.git/blob/HEAD:/contrib/exitlist

--Roger



More information about the tor-talk mailing list