On Thursday 25 September 2003 02:55, Tracy R Reed wrote:
> On Wed, Sep 24, 2003 at 08:31:55PM -0500, Pascal spake thusly:
> > Interesting story on Slashdot today.  I wonder how hard it would be to
> > implement in Freenet?
> >
> > http://yro.slashdot.org/article.pl?sid=03/09/24/132216
>
> Not hard at all and I think it is an excellent idea.

Agreed. However, a patch to the MTA would be required to give it a different 
mechanism for look-ups, e.g. one that looks up from a text file. Either that, 
or we would need a DNS proxy that would do that for the MTA, but this would 
probably need to be patched into the DNS server that the MTA is using.

> The big issue is you
> would have to make the entire zone file available on one whack which lots
> of people will see as irresponsible because it gives spammers a list of
> open relays. However they may end up crashing those boxes resulting in
> overall less spam, who knows.

Not if it's done properly. Set it up so that each file is under a separate 
key. Therefore, if a node 1.2.3.4 is an open relay, you would insert a 
file/key called SSK@<public key>/mydnsrbl//1.2.3.4, and put some information 
in it, such as the last time the relay was checked and confirmed for 
openness.

That way one could not just go and harvest a complete list of open relays - 
they would have to know what the relay being checked is to look up whether to 
allow relaying, i.e. perfect for RBL checking, but totally useless for 
spammers wanting to get a complete list of known open relays.

> This could be an excellent incentive to
> admins to run a freenet node on their network and it is a desperately
> needed legitimate use for freenet.

Agreed. But what are the chances of spammers having enough bandwidth to DoS 
the entire Freenet? There are certainly a LOT more zombied machines out there 
than there are Freenet nodes.

> Anyone know how to go about setting up such a service?

See above. Would almost certainly require an MTA patch, and it would probaby 
require a rather heavy node for checking. Fred is not great at handling lots 
of simultaneous requests constantly. I have done some tests on this, and it 
takes approximately a P3/700, 1024 connections, 256 threads, 256 routing 
table size to mostly saturate a 256 Kb/s upstream connection, if there is no 
local load. This goes up by approximately a factor of 2-2.5 if there is a 
local load of 16-32 simultaneous local connections.

All this means that if you are running a rather heavy mail server, you would 
need a Freenet node running on hardware that is an order of magnitude bigger. 
And a transient node probably wouldn't cut it, either, because the routing 
would not be settled enough. If the argument is to save bandwidth, the 
chances are that spam still uses less bandwidth on a network than a 
reasonable Freenet node would, depending on the load put on by the mail 
server. I am sure there will be SOME network admins who will decide to 
upgrade their mail server to a dual Opteron with 2 Gb of RAM just so they can 
run a Freenet RBL, but for most of them, the situation is not sufficiently 
desperate. It will get there, it just hasn't happened yet. There have been 3 
major RBLs that were shut down in the past month or two. I am sure there will 
be more, as the fewer RBLs there are, there are fewer to distribute the 
attacking resources across. We rather seem to be in a downward spiral on this 
issue. My big concern is that entire Freenet is potentially DoS-able with 
somebody who has that much bandwidth available.

> Ideally we could get one of the existing services to do it.

Then the source node would be traceable and could be attacked. Attacking 1 
node is still a lot easier than attacking the whole network.

> I really wouldn't want to get
> into the business of probing mail servers and keeping a list etc.  but I
> could if I had to.

It's not exactly difficult. Set up a script that will scan all IP addresses on 
port 25, and start creating a cache of IP addresses in a database. Run a 
second script that takes the live mail servers and tries sending an email to 
some email address you use as a testing drop box. To get better and more 
reliable results, you'd need a list of all registered domains, so you could 
look up their MX. Then try to get through by forging the MAIL FROM and/or 
From: headers to the domain (or even specific known to work email address) 
that the mail server is supposed to handle.

No doubt there would be a lot of incompetent mail server admins who would 
argue that their relays are not open (it happened before, some were even 
stupid enough to voice their stupidity and lack of understanding of what an 
open relay is in public rants and online articles), but I don't really have 
much sympathy for them.

Anyway, this is getting very off topic for a Freenet list.

Gordan
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to