Fighting spam with "sendmail aliases" in postfix (spampots?)
Hi
I have a mail server with some domains (about 200). I'm taking them from a
sendmail and putting them on a postfix-ldap + courier-ldap + amavisd +
spamassassin + clamav (thanks to perdition, the pop/imap proxy, I am doing
this and nobody notices). Everything goes well, but I have a doubt.
On the sendmail server I have some "aliases", I mean, some accounts from what
I receive mail no matter which domain is sent to (being a domain of this
machine). One utility of this was that I received all "postmaster@domain.com"
without having to configure anything.
But another utility was the spam honeypots, or spampots, or whatever you call
it, (that is, some addresses I'm sure are going to receive spam), and this
served to prove the anti-spam filter. For example, man@domain.com: no one of
my customers have this account, so every mail on this mail account is spam.
If the mail passed the anti-spam filter, I can feedback spamassassin with it
(using sa-learn).
I have some others like this: comercial, info, webmaster, etc, etc. What was
very good in Sendmail is that this aliases were only active if they were not
in the virtual user table, that is, I receive mail to "comercial@domain.com"
only if "domain.com" don't have this account.
This was pretty useful to keep trained bayesian filters in spamasssassin, and
I increased efficiency killing spam.
But now with postfix, to get this working I have 2 possibilities: create
accounts and redirect them to me if customer doesn't want it, or put all
domains in $mydestinations, and deliver them as local and not as virtual...
I think creating all accounts is very uncomfortable, but maybe I miss some
points on security about $mydestinations...
I'd like to hear your opinion about two methods, or if you know a better way,
or if you think spam-pots are the wrong way to fight spam, etc.
I'd be grateful with any hint, opinion, link or whatever respecting this.
Thanks in advance :)
Reply to: