|SPAM still delivered to user
on Tuesday, June 06, 2006 at 5:32:56 AM (EST)
Within LS admin I set up the Bayesian filter (5 months ago)to help catch incoming Junk mail. I've set the action to deliver the mail to a separate user "SPAM" and to prefix the subject with the phrase "MERL SPAM".
ALL mail that goes into the SPAM user mailbox (900+ per day) has been classified as junk by the Bayesian filter (since it has the prefix phrase). However, some classified junk mail (also prefixed) still gets delivered to proper LS users. This problem has started ocurring in the last month.
Most of the junk sender addresses appear listed in LS user blacklists (anti-spam settings via webmail options), however, I thought the email would NOT make it to the user if classified as junk, and go straight to the SPAM user. Is this correct?
If a LS user flags an email as junk within the webmail client, does the Bayesian filter get updated OR is the sender address ONLY addes to the LS user blacklist?
My guess is that the Bayesian filter should get updated. If so then surely any subsequent classified junk mail should go to the SPAM account?
Any ideas welcome
on Tuesday, June 06, 2006 at 9:38:24 AM (EST)
So the spam flag in the headers says "yes"? Is this happening for all users or just a few users? The Anti-spam mail rule might need to be turned on in Web Mail. :)
By the way, e-mail address blacklists aren't exactly the most effective way to block e-mail because they can be so easily forged. For example, I can easily send an e-mail from my email address, but when I log in to an SMTP server, I can say I'm actually "President@whitehouse.gov". Because SMTP servers trust the e-mail address in the helo/rcpt/data command, it can show up as anything. It's not really a limitation of Lan Suite, but in SMTP servers in general (and why there's a movement to ditch the SMTP server standards and come up with a more secure way to send e-mail).
on Tuesday, June 06, 2006 at 10:46:07 AM (EST)
Thanks for the reply. Although I'm not that expert on headers/SMTP, I can't see any entry obviously related to a spam flag in ANY of the classified messages (either the ones in the SPAM account or my own user account). I have just set the Bayesian filter to add a -LNS-Spam-Check header to message, I guess that's something else though?
This is how I set up our email system, the blacklist usage appears automatic. Please comment if you think the set-up should be different.
We receive all our business email (for @merl-ltd.co.uk) from a 'catch-all' POP3 mailbox hosted by our external webhosting company. We do not subscribe to any mail filtering on their SMTP servers. All of the mail comes into LS and is delivered "according to the address" to respective users.
As mentioned, I have the Bayesian filter enabled (10% confidence level - no false positives to date), sending classified junk to a user ("Spam") mailbox.
I performed a lot of training early on by manually selecting junk and not junk messages from various users WITHIN WEBMAIL.
Also, within the webmail program, ALL users are set up with both "checking methods" enabled in the Anti-Spam Settings. Action on the junk mail is to "move to the" INBOX.
From what I've observed, the webmail 'personal' blacklists are populated automatically whenever an email is selected as JUNK.
Is this set-up correct? Do the "checking methods" need to be employed for the anti spam settings within the Webmail for each user, or is enabling the Bayesian filter within LS configuration enough? Is our prefix phrase attached to the messages when they reach LS and BEFORE delivery to any user? If so then the message should go to the spam user mailbox, surely?