Robert wrote: |Hello Kurt,
Howdy. |Monday, November 17, 2003, 11:03:24 AM, you wrote: | |KB> I've just put up 2.60 on FreeBSD, and consider myself a bit of a |KB> newb, but I've been reading the man page for spamassassin, |and found |KB> a section on spamtrapping. We've got 40 or so addresses of former |KB> employees that would be ideal for this purpose, but I'm not much |KB> interested in Razor/DCC, etc. What I'd really like to do |is feed the |KB> spamtrapped stuff into the Bayes database. Is this a reasonable |KB> thing, and if so, is there a way to do it? | |You don't indicate how your email is being delivered and/or |retrieved. My |methods work for POP3 on my system -- YMMV Hmmm... As I mentioned, the FreeBSD box is a gateway to the Exchange box, with no local mailboxes. I did forget to mention that the FreeBSD box is running Postfix, but the combo is working well. So, there's no POP3 involved - it's a straight SMTP gateway. |Concern: Are you 1000% sure that EVERY email sent to those addresses |will be spam? I have an old email address for someone no |longer with the |organization that was discontinued 6 or 7 years ago, and we |still get one |or to valid emails to that address each year. I'm therefore |unable to use |that address as an automatic spamtrap. You do NOT want to |poison your own |Bayes database with even one or two messages a year. I understand and appreciate your warning. These are low-level accounts from ex-employees who didn't normally have business contacts outside of the company. I've not seen any business-related emails to them in the time that this system as been active (I implemented an empty distlist on my Exchange server for these addresses early last year, which dumps emails to those addresses in a separate accounts mailbox, and have had no complaints in that time.) I actually have more potential spamtrap addresses, but the others were for people who might still be receiving legitimate email, so I wasn't planning on including those. |Method: Auto-forward any such addresses to a common spamtrap |mailbox. Run |sa-learn --spam --mbox $directory/inbox against that mailbox every so |often (daily? hourly?). Delete the inbox file after each sa-learn run. This looks possible, but I'll probably have to research more about how postfix operates, since this means adding at least one local mailbox to the gateway, with aliases for the spamtrap addresses to it. Anyone have ideas on how to do this? Currently, no mail gets saved locally - even the daily security/cleanup reports get sent from the FreeBSD box to our Exchange server, so that I can look at them. | |Alternate (no special spamtrap needed): Auto-forward any/all unknown |email address (eg: [EMAIL PROTECTED]) to a common "check these |out" mailbox. Add rules to your local.cf to kick the SA score |above your |auto-learn threshold for these specific email addresses. They'll be |auto-learned. This approach looks less likely for me, but it is an interesting approach. The downfall of it is that there are a lot of fat-fingered people in the world, and a mistyped address attached to a legitimate email could be problematic. | |Note: Both methods assume system-wide Bayes. It'll do no good to apply |these to one Bayes database which is not used by the majority of email |users. Yup. That's what I've got, AFAICT. | |Bob Menschel Kurt ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
