K9 Spam Filter Software

[Jonny M]

Downloaded this after seeing it in the sticky, but it just adds [Spam] and a percentage to the subject. Is there any way to set it up to bounce spam messages back, or does it only add [Spam] to the subject?

Thanks

Also, is there a way to stop it adding [Spam] to messages that aren't spam?

[Jiff Lemon]

Create a spam folder, then a rule to either move the mails to the SPAM folder, marking them as read (with the essential Stop processing more rules) or have the rule delete them (dangerous).

Why Dangerous? you could delete mails that aren't classed as spam.

Initially you need to train it - Open K9, click on recent emails and highlight all the spam, and mark it as SPAM. What happens then (as I understand) is that the mails are analysed, and the words within the mail added to a SPAM database. New mails are then scored against this database and anything over a certain % is marked as spam, to be picked up by the rule.

A similar process happens when you mark mails as Good.

Theres also whitlist and black lists you can use if you really want to start filtering....

[Zathras]

K9 comes with some default rules, though it needs training to be perfect. I actually use K9 in association with Bloomba's SAProxy, based around SpamAssassin, so I have two layers of spam filtering. Takes more setting up and is slower, but I don't care. You can set K9 to add a custom header rather than append Spam to the title - this is better as then you can filter on the header into a spam folder and any 'hams' (good mails misidentified as spam) can be moved back into your inbox without having the 'SPAM' prefix.

You also don't really want to automatically bounce spam back as every so often, even with the best training and rules in the world, you'll get a message that is misidentified.

I'd suggest RTFM with K9 - it's really quite simple to get going with. Training - just double click the systray icon and you see your recent emails and how they were processed. You can then change the status of non-identified spam to spam and hams to 'good' and K9 will learn. I get upwards of 100 spams a day and after training on 6 misidentified emails over two weeks ago I've only had one misidentified message.

[DaBeeeenster]

Cloudmark Spamnet used to be the best by a mile, but now they charge you to use it

[Jiff Lemon]

Well, Just picked up a weekends worth of mail.

170 mails, 169 of them correctly identified as Spam, 1 genuine mail correctly filtered out.

Damn this program is good!

[Zathras]

Aye, I'm very impressed with it. The only failing (and you'll get that with all proxy-style spam filters) is that when it misidentifies spam/ham and you correct it, it doesn't correct the email in your mailer program and move it into the correct folder. However that's not surprising because every mail program under Windows uses different storage methods and data structures and it would be extremely difficult for K9 to find the relevant email.