Submitting spam to an auto-learn account

[sillyme]

Based on one of your earlier posts I have created an account to use as a training for the Bayesian filter. The account is called good@ and has an alias of spam@. I have created a spam folder in it. Users forward spam that has not been flagged to spam@. I am using the content filter to strip the from header field (from=""), set the spam flag and redirect to good@. In good@ the spam is placed into the spam folder. Any wanted emails that were flagged as spam are forwarded to good@ by the user and ends up in the inbox.

My question is that if the user has several spam emails to forward do they have to do them one at a time or can they send them as a group. Sent as a group one email is created and all the individual emails are attachements.

Thanks

[Code Crafters]

First of all just to comment on your setup a bit it would be simpler if you had 2 accounts spam@yourdomain.com and good@yourdomain.com. If all mail to spam@ recirects to good@ then that's a bit redundant other than to trigger then good@ rule. You also need to make sure that if spam@ redirects to good@ that the good@ rule comes after the spam@ one in content filtering. I'm sure these rules probably work but could probably be simpler set up.

Normally, forwarding emails to another account is only needed for POP3 accounts since IMAP4 and WebMail users normally just store the mail in their own SPAM / non-SPAM folders for Bayesian learning which is much easier to do.

To answer your question anyway, it is better to send the mails in their original form as sending as a group email will also add a harmless container email that will be trained as SPAM and may cause more false positives (non-SPAMs detected as SPAM). Training from within an IMAP4 / WebMail account rather than forwarding is even better as the original headers are fully intact.

[sillyme]

Thats basiclly what I thought but I wanted to make sure.

I thought there would be a problem using two email accounts as the spam would be the inbox and auto learn would treat all inbox's as good mail.

In the content filter here are the rules I set

Spam Training Good
Any SMTP reciepient contains "good@domain.com"
Remove spam flag
edit "from" header to ""
stop filtering

spam training bad
Any SMTP reciepient contains "spam@domain.com"
set spam flag
edit "from" header to ""
stop filtering

IN the User account
Good account has a alias of spam@domain.com and I created a spam folder
under the Filters tab
Check if marked as spam is ticked
and place in directory is set to spam

If there is a better way besides IMAP for pop3 users to assist in the training of the Bayesian filter let me know.

Thanks

[Code Crafters]

Forwarding mail is the only way for POP3 bayesian training. However, You can filter the mail with content filtering rather than user filtering easier as follows:

Spam Training Good
Any SMTP reciepient contains "good@domain.com"
Remove spam flag
Place in directory "non-spam"
Edit "from" header to ""
Stop filtering

Spam Training Bad
Any SMTP reciepient contains "spam@domain.com"
Set spam flag
Place in directory "spam"
Edit "from" header to ""
Stop filtering

Use the user spam@domain.com and/or good@domain.com for bayesian user training with "non-SPAM" and "spam" as the good and bad folders to learn from within that user which avoids using the Inbox for training which is obviously better.