regarding training of ham and spam to spamassassin

[Mark Sapiro]

On Thu, May 13, 2010 at 06:49:19PM +0530, arun gupta wrote:
> > Dear Sir,
> 
> 
> As per said by you i am pasting one spam mail
> 
> _____________________________________________________________________________--
> From: cdac.in support <pd at cdac.in>
> 
> To: pd at cdac.in
> 
> Subject: setting for your mailbox pd at cdac.in are
> changed
> 
> 
> 
> SMTP and POP3 servers for pd at cdac.in mailbox are changed. Please carefully
> read the attached instructions before updating
> settings.
> 
> 
> 
> http://mung_url_to_stop_hit.invalid/web/setup.zip
> 
> 
>  ________________________________________________________________________________________
> 
> I tried at least 100 times this type of spam mail to spamassassin through
> sa-learn but spamassassin caught as spam after 3 days, please give the some
> inputs.


If you install the "sought" ruleset
<http://wiki.apache.org/spamassassin/SoughtRules>, that message will
hit JM_SOUGHT_1 and get +4 points.

As others have noted, sa-learn trains SA's bayes filter which is a
statistical filter. It is not going to give a high spam probability
based on only a few samples.

-- 
Mark Sapiro <mark at msapiro.net>        Any clod can have the facts;
San Francisco Bay Area, California    having opinions is an art. -
                                      C. McCabe, The Fearless Spectator