Grex Oldcoop Conference

Item 222: Where to stop spam: before or after it reaches grex?

Entered by gelinas on Sun Dec 5 05:55:38 2004:

At the grex-walk debrief today, the conversation turned to the prevalence
of spam and the various ways of combatting it.  It occurred to me to ask:
is it time to change our philosophy of how best to reduce the flow?

Historically, we (grex, cyberspace communications, inc) have worked to
prevent the spam ever crossing our link.  Of late, that effort has seen
decreasing success.

The off-the-shelf tools for stopping spam do not, as near as I can tell,
work before the mail crosses the link:  The tools are separate from SMTP
and have to receive the message to judge its fate.

If we are willing to sacrifice the band-width, there are several anti-spam
programs we could make available to our community.  Is it time to sacrifice
the bandwidth?
18 responses total.

#1 of 18 by gelinas on Sun Dec 5 05:57:20 2004:

If/when we move to a co-location site, we should get more bandwidth, so the
sacrifice would not be as painful.

Also, the solution decided upon would be implemented on the new machine, not
on the current grex.


#2 of 18 by prp on Sun Dec 5 21:27:04 2004:

Why not go with both?

I'm currently using three: Grex, Comcast, and Mac OS X Mail.  I don't know
how much Grex stops, but Comcast gets about fourteen/week and Mail gets 
about four/week.

It might be nice kept a list of stuff that got stopped and sent a weekly
summary.


#3 of 18 by prp on Sun Dec 5 22:06:00 2004:

Does Voyager offer any spam filtering service?  If so, could the mail be
run through it before going to Grex.  Co-location would mean changing ISP's.
Do the canadites offer spam filtering?


#4 of 18 by gelinas on Mon Dec 6 04:46:19 2004:

To run our mail through our ISP's spam filters, we'd have to have all of our
mail delivered to their mail servers.  I've not thought that this is something
we would want to do.


#5 of 18 by tod on Mon Dec 6 06:52:08 2004:

I think the least administrative burden would be to run spamcop and an
aggressive global procmail setting which diverts positives to /dev/null
This should be further explored.  Glad folks on the A2 Grexwalk thought of
it. ;)


#6 of 18 by dpc on Mon Dec 6 14:09:42 2004:

How much bandwidth are we talking about sacrificing?


#7 of 18 by tod on Mon Dec 6 17:40:30 2004:

No more bandwidth than currently being utilized.  Its more about CPU
utilization.


#8 of 18 by cross on Wed Dec 8 03:27:14 2004:

Realistically?  Not that much.  Realize that most spam gets through
grex's spam blocks; the amount we block is negligable compared to the
amount we get.  So filtering it on this side of the DSL link isn't
really much of a bandwidth hit.  But I've been saying this for years
and been mostly ignored.


#9 of 18 by dpc on Wed Dec 8 14:48:43 2004:

I'm not in favor of ignoring people with good suggestions.


#10 of 18 by tod on Wed Dec 8 18:57:17 2004:

I second Mr.Cahill's motion.


#11 of 18 by krj on Fri Dec 10 23:03:38 2004:

The two approaches are not mutually exclusive; setting up 
a system to deal with the spam after it arrives on Grex does not 
preclude efforts to stop it from being transmitted here.


#12 of 18 by cross on Sat Dec 11 00:25:51 2004:

Regarding #11; That's true as well.  There's no reason we can't try
and have the best of both worlds.


#13 of 18 by gelinas on Sat Dec 11 04:17:44 2004:

I find that two spam filters are worse than one, not better, when the goal
is individual control over the definition of spam.


#14 of 18 by cross on Sat Dec 11 19:25:12 2004:

Well, the thing is, some things you just *know* are spam.  Or virii.  Or
something of the like.  There's no reason for the system to not filter
those.  Users are then free to filter whatever they like.


#15 of 18 by prp on Sun Dec 12 02:15:41 2004:

Individual control would mean not blocking anything from getting to
Grex, but seperating the junk mail from the rest, so people can take
a look and see what is getting diverted.

Although I suppose it would be possible to do this at the ISP end.


#16 of 18 by other on Sun Dec 12 16:42:36 2004:

Sure.  Just add a header identifying the spam as such at the ISP and
filter for it locally.


#17 of 18 by devnull on Tue Dec 14 05:37:40 2004:

Individual control has disadvantages, though, too.

If you block spam before the SMTP session completes, then you can give the
sender an error message, so that if there's a false positive, the sender
is likely to know that the message was rejected.  Most approaches to individual
filtering hope that the reciever bothers to go looking for false positives,
which the reciever may not actually bother to do.  (You could in theory
send a bounce message after the fact, with the individual approach,
but then you're hoping the return addresses are valid, and you might end up
effectively acting as a relay to send spam to some third party if the sender
address was forged, and forging return addresses is trivial.)

Also, there is a technique known as greylisting can only be done on the smtp
server recieving the mail.  (The basic idea is that many (but not all)
spammers do not bother to retry if they get a temporary failure, and real
mailers do retry unless they are broken.  So you can keep track of which
machines try to contact your SMTP server, and if you haven't heard from them
recently, you give a temporary failure, perhaps for an hour or so, after
which you let mail through.)

greylisting might reduce the amount of bandwidth used by spammers, too, if
it gives the temporarily failure at the rcpt to: point in the smtp session.

Of course, it is possible for an smtp server to be told about the preferences
of individual users.


#18 of 18 by jesuit on Wed May 17 02:15:21 2006:

TROGG IS DAVID BLAINE


There are no more items selected.

You have several choices: