free antispam for Linux

[Lennart Sorensen]

On Mon, Sep 24, 2007 at 08:55:00AM -0400, John Van Ostrand wrote:
> The summary of all this is that generally most solutions use more than 
> one technique for finding spam. SpamAssassin performs pattern matching 
> in the email and meta data, greylisting and delayed banners check for 
> impatient MTAs, RBLs look for known spammers or for hosts that shouldn't 
> be sending email, OCR decodes images to look for words, razor and pyzor 
> compare messages with a network of MTAs, spam traps capture spam and 
> there are many more techniques.
> 
> One thing that I've learned is that there isn't a fire-and-forget open 
> source solution. Each needs tuning and work to keep up. I tend to have 
> to re-tune my solution (MailScanner) quarterly. I use RulesDuJour to 
> keep spam assassin rules up to date daily, but I have to check for new 
> rule sets manually. MailScanner comes up with new ways to detect spam so 
> upgrading and reviewing it's changelog are valuable. Very rarely I will 
> institute a new MTA method (like delayed banner) or turn on a new 
> MailScanner feature.
> 
> In my case my server processes about 10,000 messages a day. About 5,000 
> are refused before they submit the message. Of the 5,000 remaining about 
> 2,000 are spam.

I have bogofilter setup to auto add words from messages that it already
declare to be spam so it can learn new ones.  If I see messages it
missed in my inbox, I feed them to it to learn.  Similarly if I ever do
find a false positive I feed that to it to learn that it was not spam.
For the most part I don't have to do anything and it just does the right
thing.

--
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists