[wplug] Is Spamassassin good?

[Jonathan Billings]

Mike Sussman wrote:
> I have heard and read many good things about Spamassassin.  The other day, I 
> installed it in my Kubuntu Feisty installation, using Kmail mail client.  
> Spamassassin is implemented as a series of filters, and is not all that 
> convenient because kmail locks up when it is checking mail in my imap 
> account, and it can lock up  for several minutes.  But getting rid of spam is 
> a good thing, and I am willing to put up with inconvenience it it works.

One of the things I've done to speed up the responsiveness of 
spamassassin is to run spamd, and have the filters use spamc instead of 
spamassassin.  This means that the big heavy perl script is already 
loaded, and the only thing you have to load every message is a small 
binary written in C.

> But my surprise is that it does not find very much spam!  It identifies only  
> about 10% of the messages I would regard as spam myself.  I thought that by 
> specifying messages as spam (one filter provided by the installation is 
> named "Classify as spam"), that Spamassassin would learn and start picking 
> more out, but it hasn't been happening that way.  When I look at some 
> messages that are not identified as spam, I can see very low numbers (less 
> than 2, when the default criterion is 5), so I think it will never be finding 
> much spam.  And there was one distressing false positive from a friend of 
> mine that I had to fix with by manually placing his address into the 
> whitelist.
> 
> My question is: am I expecting too much for my type of installation?  Can I 
> expect Spamassassin to find more than, say, 75% of the spam that is sent to 
> me?  If so, how can I learn how to make that happen?


I suggest saving un-tagged spam in a particular folder, and run it 
through 'sa-learn' (read the man page for directions) to teach 
spamassassin's bayesian filters.

You can also download additional spamassassin rules at the Rules 
Emporium (http://www.rulesemporium.com/) which have up-to-date rules for 
catching spam.  There are also rules that will check the domains in the 
email against known spam DNS blacklists.

I no longer run my own anti-spam measures, but when I did, these were 
very effective.

-- 
Jonathan Billings <billings at negate.org>