Do you use greylisting?

Submitted by root on Sun 2 Sep 2007

Tags: , ,

 

Huh?  <-> 42%472 votes
No  <-> 31%352 votes
Yes  <-> 25%284 votes
Total 1109 votes

Posted by daemon (146.231.xx.xx) on Mon 3 Sep 2007 at 09:30
[ Send Message | View Weblogs ]

I would, but our central IT does greylisting upstream, so there'd be little point in us slowing things down...

Cheers
:wq

[ Parent ]

Posted by remofritzsche (217.71.xx.xx) on Mon 3 Sep 2007 at 10:34
[ Send Message | View Weblogs ]

I considered to use it, but I'm not sure how much it realy performs. The world of spam changes realy fast. In our company we have greylisting and a lot of spam, too.

What do you think? Is it useful? Is there any percentage of restrained spam mails?

Remo

A programmer is just a tool which converts coca cola into code.

[ Parent ]

Posted by Wayne (82.144.xx.xx) on Mon 3 Sep 2007 at 15:08
[ Send Message | View Weblogs ]
Only when the server is being hammered

[ Parent ]

Posted by mario (201.242.xx.xx) on Wed 5 Sep 2007 at 23:31
[ Send Message | View Weblogs ]
I use postgrey and spamhaus, a delay of 5 or more minutes is acceptable.
I like the spamtrap option that policyd has, would be useful to block AOL when the dictionary attacks begin (20-30 tries to unexistent accounts).
A year ago one user told me that from 7 spams a day he went to 1 or 2 a week.

[ Parent ]

Posted by Anonymous (82.154.xx.xx) on Mon 10 Sep 2007 at 20:47
my friend, it made the diference between thousands of spam messages to just a dozen of them per mailbox... i think that is by itself a statement...

cheers,

João Carneiro

[ Parent ]

Posted by mwr (24.158.xx.xx) on Sat 15 Sep 2007 at 20:31
[ Send Message | View Weblogs ]
I'm not as concerned with stopping every spam as much as I am with stopping lots of it with little effort and little (or no) side effects. We just put greylisting on our new mail server, and here's what we found. 09/11 results are a bit smaller since we installed the mail server late that afternoon:

DateDelivered without DelayGreylistedGreylisted, but Re-sent
2007/09/1122213630
2007/09/128593367
2007/09/138882709
2007/09/1484536313

So we greylisted anywhere from 30-60% of our incoming email on those days. Aside from the first day, less than 3% of it ever got resent. Grepping through those logs, it looked like it was all spam to mailman administrative accounts, to moderated lists, or to my personal account. Everything else was delayed by an hour or less, and about 30% of it was delayed by 15 minutes or less. I'm sold.

[ Parent ]

Posted by kaerast (82.47.xx.xx) on Mon 3 Sep 2007 at 15:55
[ Send Message | View Weblogs ]
It's useful when your mail server gets lots of email, usually from the same people. However, when you've got a baby mail server handling small amounts of mail or getting just one or two emails from lots of different people then it's less useful.

[ Parent ]

Posted by GoodTimes (209.120.xx.xx) on Tue 4 Sep 2007 at 14:36
[ Send Message | View Weblogs ]
but it is still useful

i only have a handful of users and i found that greylisting reduced the amount of spam i received by almost half

i still need spamassassin and such running to get the rest, but this takes care of a lot of the ...lightweight... spam out there

the negative is that it can delay new email, but i find that a small price to pay when i can look at my inbox for several days in a row without one piece of spam having gotten through

and remember, if you delay spam, then you have more of a chance that DCC might have identified that message as spam and increase the chances that when it is redelivered (if it is redelivered) you'll id it as spam

aaron

[ Parent ]

Posted by Anonymous (213.164.xx.xx) on Wed 5 Sep 2007 at 08:57
At the cost of users asking where their e-mail is..

[ Parent ]

Posted by Anonymous (62.206.xx.xx) on Wed 5 Sep 2007 at 10:25
there is not guranteed delivery time. so waiting for the first time, when a new triplet is established is it worth.
most users know about our greylisting and they tell us the senders information in cases of emergency for whitelisting.
greylisting itself cuts down the spam to more than the half. its our second step,first we use dns-blocklists and 3rd spamassassin. so our spam situation is quite good and we have close to none false-positives.
problems seen with greylisting:
1. speed, but solved via whitelisting as written above.
2. misconfigured mailservers, retrying to fast/often using non-rfc intervals.
3. misconfigured mailservers, with unallowed chars in the hostnames, wich is not only a greylisting problem...
4. users of small companies often send directly, using dial-up addresses that change often and are filtred via dns-blocklists...althought they could use their providers mailserver, what could reduce whitelisting issues...

[ Parent ]

Posted by GoodTimes (69.17.xx.xx) on Wed 5 Sep 2007 at 12:13
[ Send Message | View Weblogs ]
What are you all using to greylist? I'm using DCC with sendmail. Is there something better? i haven't yet figured out whitelisting with this (i haven't tried that hard either) since I got it setup. It was actually kind of a pain to setup the first time as the package had some errors (which have since been reported and fixed)

aaron

[ Parent ]

Posted by simonw (212.24.xx.xx) on Wed 5 Sep 2007 at 13:41
[ Send Message | View Weblogs ]
Postgrey.

Guess a "I wouldn't start from there" kind of response.

[ Parent ]

Posted by Steve (82.32.xx.xx) on Wed 5 Sep 2007 at 21:52
[ Send Message | View Steve's Scratchpad | View Weblogs ]

When it comes to greylisting I've fallen back to the simpler "sender + recipient" pairs; that takes care of server-farms which have different IP blocks to send mail from.

I found prior to making that switch I had extremely delayed mail from senders such as googlemail - they'll first try sending from one IP, then later another, and later still a completely different one!

Steve

[ Parent ]

Posted by GoodTimes (69.17.xx.xx) on Thu 6 Sep 2007 at 21:52
[ Send Message | View Weblogs ]
how'd you do that?

aaron

[ Parent ]

Posted by Steve (82.32.xx.xx) on Thu 6 Sep 2007 at 21:56
[ Send Message | View Steve's Scratchpad | View Weblogs ]

I guess it depends on what you're using, for me using qpsmtpd on Debian I just add this to qpsmtpd/plugins:

greylisting black_timeout 60 recipient 1 sender 1 remote_ip 0

The ones mean "include this data" when making the test, the zero means "don't include it". So here I include both sender address, recipient address, but not the IP.

Steve

[ Parent ]

Posted by simonw (84.45.xx.xx) on Fri 7 Sep 2007 at 22:21
[ Send Message | View Weblogs ]
Postgrey uses the /24 (class C?) as the third part of the triplet.

This handles most of these cases fine, it also ships with a whitelist.

I've only come across very few exceptions, mostly very broken mail servers, or people with issues on the Postgrey mailing list. I remember part of IBM managed to ship email internally, via Lotus Notes, and defeat greylisting.

But also depends how busy your email server is, Postgrey whitelists at 5 successful deliveries (by default), so most of Google's mail servers were white-listed in a few days.

I would think sender/recipient pairs lead to a lot of spam passing greylisting, spam tending to have unimaginative data.

[ Parent ]

Posted by Anonymous (213.11.xx.xx) on Tue 11 Sep 2007 at 14:49
I don't use greylisting as I do not consider it as good solution to spam.
It requires action from the users, who most of the times don't understand the warn message or don't even want to read it, considering it's spam!
Additionnaly I guess it can slows a little mail delivery...
I realy don't like this kind of policy but I have to said it's an efficient way to get rid of almost all spam (I maybe of ham to!).

[ Parent ]

Posted by daemon (146.231.xx.xx) on Tue 11 Sep 2007 at 15:43
[ Send Message | View Weblogs ]

Any properly configured mail server shouldn't have any trouble with greylisting behavior, so you shouldn't have any lost ham.

Also, it shouldn't require any action from the users, and you shouldn't have any users getting warning messages about the fact that their mail's delayed, unless of course your mail server is blocking messages for far too long -- a few minutes is usually enough, rather than the hours that would be necessary to illicit a warning to the original sender.

Cheers.
:wq

[ Parent ]

Posted by Anonymous (89.181.xx.xx) on Wed 12 Sep 2007 at 16:52
I use greylisting as a second necessary condition for rejection in my small mailserver: mail goes through a variety of more or less rigid checks in Exim's ACLs but it isn't denied based on the result of those checks; suspect senders are simply sent to the greylist, while all other skips greylisting.
This way I can reject spam safe and accurately, and accept mail from most broken/weird servers (unless they have messed up retry times or something, but I'm yet to find such server).
I also mark mail originating from countries with high levels of spammers as suspect without risking losing legitimate mail, which leads to even less spams getting through.

[ Parent ]

Posted by Anonymous (86.49.xx.xx) on Sat 15 Sep 2007 at 12:02
I use milter-greylist with sendmail. I combine it with mimedefang, ClamAV and SpamAssassin.

In greylisting I use rather short delay only 1 minute.
On average day we have 20--40 thousands attempts delayed of which 5--10 percent tries it later. So greylisting seems to be still very efficient.

Several users complained, namely because of mixing delivery times of mails from mailing lists (when after long silence a thread starts, the first mail is delayed). So I put those users on whitelist.
Also I had to whitelist several IPs of some broken MTAs.

JDe

[ Parent ]

Posted by Anonymous (66.92.xx.xx) on Mon 17 Sep 2007 at 15:31
I tried it but it screwed up the mailman lists on my machine.

[ Parent ]

Posted by Anonymous (74.92.xx.xx) on Thu 20 Sep 2007 at 01:17
While not Debian specific, it really kick spammers where it hurts.

Read more about it here:

http://www.openbsd.org/cgi-bin/man.cgi?query=spamd

[ Parent ]

User Login

Username:

Password:

[ Advanced Login ]

Register Account

Quick Site Search

Poll Archive

View Prior Polls