Re: Tor a carrier for Botnet traffic?

[Scott Bennett <bennett@xxxxxxxxxx>]

     On Tue, 1 Sep 2009 14:11:04 -0300 Freemor <freemor@xxxxxxxxx> wrote:
>On Tue, 1 Sep 2009 11:38:42 -0500 (CDT)
>Scott Bennett <bennett@xxxxxxxxxx> wrote:
>
>
>>      You raise a good point.  Nevertheless, *we* aren't the ones doing
>> the banning, rather the ISPs are (e.g., Comcast in the U.S.), and the
>> ISPs are certainly not going to ban Windows.
>>      There's been quite a few times that, when I've called an ISP to
>> try to get them to fix a problem on their end, I've been told that
>> they don't support FreeBSD, which, of course, is irrelevant.  I don't
>>...
>[snip]
>
>You bring up a good point Scott about the Tor website referring to Tor
>as a server. The truth is that it is much closer to a router or proxy.
>The problem as I see it however is that ISPs that have AUP against
>servers usually include proxies as a no-no.. in fact I've seen some so

     That is true, but just avoiding the term "server" will help in many
cases, I think.  And if we can insist that if our tor nodes are proxies,
then their big routers are also running proxies, we may at least befuddle
them from threatening us further for a while.  It's hard to say without
trying it, I suppose.

>broad as to claim that using ping or trace-route are "hostile" and
>grounds for disconnection.

     Ugh.  Even worse than what I've run into.  The AUPs of some ISPs
do make it impossible to run tor in relay mode without getting disconnected,
and we probably can't get around all of them.  It's really outrageous that
they want us to pay, but then cut us off if we actually use the service.
Anything we can do to reduce the number of would-be nodes prevented from
actually being nodes by the ISPs' interference, though, would be a good
thing.
>
>So I agree that the wording on the Tor site probably should be changed.
>I however doubt it will do much to convince errant ISPs. from what I
>have seen the thing that sets off hostile attitudes by some ISPs (I'm
>in Canada mind) is too much outbound traffic. This is especially true
>for ISPs that consistently oversell their bandwidth. (a la Rogers).
>
     Comcast and Time-Warner apparently oversell their capacity.  The hell
of it is that they could get a lot of leeway just by running Quality of
Service software on their routers, but they appear to be unwilling.  If
there were more competition available, it wouldn't be such a problem.
     Meanwhile, here's something to ponder.  Someone whose Internet
connection is not allowed to have anything listening on any ports on
that interface might still be able to run a limited tor node, provided
it is never used as an entry node for a circuit.  If each such node
were to open at least one TCP connection to NumEntryGuards entry guards
and then always keep at least one connection open to each such chosen
entry guard, then the limited nodes could publish modified descriptors
that listed the entry guards in to which they were connected.  A client
building a circuit could then choose that node as a non-entry hop for a
circuit, although one of the chosen entry guards would have to precede
it in the circuit.  In principle, at least, tunneled directory services
and hidden service directory service could be provided via such a setup
as well, although the rate capacities and latencies involved might render
these not worthwhile.  This type of limited node could optionally provide
exit service as well, so it would not be restricted to middle node duty
alone.  (Increasing NumEntryGuards to a somewhat larger number than 3 might
also be a good idea in this situation.)
     To reduce the anonymity set partitioning that could result from this
setup, it might be necessary to have a chosen guard recognize when
a limited node were connected to it, so that it could refuse to extend
a circuit through the limited node until it had a certain minimum number
of limited nodes connected to it simultaneously.  Once that threshhold of
established connections from limited nodes were reached, the entry guard
could begin allowing circuits to be built through the limited nodes that
were connected to it.
     I've obviously neglected a lot of the grubby details above, e.g.,
the specifics of how a limited node responds to an entry guard going down
or becoming unreachable.  I just wanted to throw this out as a blue-sky
idea for discussion to find out whether it might be a) actually workable
and if so, b) worth the bother to do it.  Because it would require yet
more changes to the directory format and protocol, I'm sure it would not
be trivial to implement.  I've been wracking my brain, trying to think of
some way that people in my own situation vis-a-vis ISP problems might still
be able to contribute something to tor network operation, and thus far,
that's about the only idea I've come up with that *might* be useful.  If
anyone comes up with other ideas, please toss them into the discussion ring,
too.


                                  Scott Bennett, Comm. ASMELG, CFIAG
**********************************************************************
* Internet:       bennett at cs.niu.edu                              *
*--------------------------------------------------------------------*
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *
**********************************************************************