Re: [vchkpw] [OT] Better Spam Detection

2005-04-26 Thread Kiril Todorov
Andrew Niemantsverdriet wrote:
> pretty much stock spamassassin rules. 
[...]
> How can I improve our detection? What recommendations does the list
> have?

Check out : http://www.rulesemporium.com/

Get the email corpus, it's good for a start feed of the bayes analyzer
in SA.
Feed regulary your bayes filter with spam and ham messages that you receive.



signature.asc
Description: OpenPGP digital signature


Re: [vchkpw] [OT] Better Spam Detection

2005-04-26 Thread Kiril Todorov
Silly me...
Forgot - add the razor2, pyzor and dcc checks as well.

Best Regards,
Kiril.


signature.asc
Description: OpenPGP digital signature


Re: [vchkpw] [OT] Better Spam Detection

2005-04-26 Thread zrr
Train SpamAssasin with more Spam. Check out tmda-cgi.

Best
Zeno

On 4/26/05, Andrew Niemantsverdriet <[EMAIL PROTECTED]> wrote:
> I have been running qmail + vpopmail + qmailscanner + spamassassin +
> clamav for a while now. While some spam is caught our detection rate is
> not great.
> 
> My qmail uses ordb and spamhaus and dsbl for rbl checking. I am using
> pretty much stock spamassassin rules. The only thing that is changed is
> it uses mysql for per user preferences.
> 
> How can I improve our detection? What recommendations does the list
> have?
> 
> Thanks,
>  _
> /-\ ndrew
> 
>


Re: [vchkpw] [OT] Better Spam Detection

2005-04-26 Thread Kyle Wheeler
On Tuesday, April 26 at 10:21 AM, quoth Andrew Niemantsverdriet:
I have been running qmail + vpopmail + qmailscanner + spamassassin +
clamav for a while now. While some spam is caught our detection rate is
not great.
My qmail uses ordb and spamhaus and dsbl for rbl checking. I am using
pretty much stock spamassassin rules. The only thing that is changed is
it uses mysql for per user preferences. 

How can I improve our detection? What recommendations does the list
have?
I agree with most of the list - your options are:
1. enable all of the rules that spamassassin has (and install any 
  necessary software
  a. Razor2
  b. Pyzor
  c. DCC
  d. DNS rules
2. make sure that it is using and developing bayesian databases

That much has gotten me a VERY good spam detection rate. Beyond that:
3. Add rules from some of the rule archives
4. Finally - examine the spam that doesn't get caught to try and 
  determine patterns. Then you can write your own rules.

~Kyle
--
Dealing with failure is easy: Work hard to improve. Success is also easy 
to handle: You've solved the wrong problem. Work hard to improve.
-- Alan Perlis


signature.asc
Description: Digital signature