SpamAssassin sa-learn issue

[Tim Thorburn <de...@arashidesign.com>]

  Hello,

I've run into an issue with a mailbox where large amounts of spam are 
making their way through filters.  The majority of these spam messages 
receive SA scores of 0.3-1.6; for obvious reasons I cannot adjust the 
filter sensitivity this aggressively.  After reading through the docs 
I've come across the 'sa-learn' command which is supposed to help 
SpamAssassin learn beyond its scoring method - please correct me if I 
misunderstood here.

I'm using SpamAssassin version 3.2.5 on RHEL4 on a VPS should this be of 
assistance.  I've moved a collection of these spam messages into a new 
folder made with a supplied webmail software.  The command I am using 
is: sa-learn --spam /path/to/file  This file currently has 153 spam 
messages; however when the above command is run I receive the following 
message:
Learned tokens from 0 message(s) (0 message(s) examined)

I take this to mean that the learn tool either found nothing of 
importance to scan, or simply found nothing at all to scan.  Are there 
additional parameters I should be including here?  Or is it more a case 
of because these messages receive such a low SA score to begin with, 
SpamAssassin finds nothing wrong?

TIA for any assistance,
-Tim Thorburn

Re: SpamAssassin sa-learn issue

[RW <rw...@googlemail.com>]

On Thu, 12 Aug 2010 08:49:00 -0400
Tim Thorburn <de...@arashidesign.com> wrote:

>   The command I am
> using is: sa-learn --spam /path/to/file  This file currently has 153
> spam messages; 

if the file is a mailbox file you need to specify the format e.g. --mbox


> however when the above command is run I receive the
> following message:
> Learned tokens from 0 message(s) (0 message(s) examined)

Re: SpamAssassin sa-learn issue

[Bowie Bailey <Bo...@BUC.com>]

 On 8/12/2010 8:49 AM, Tim Thorburn wrote:
> Hello,
>
> I've run into an issue with a mailbox where large amounts of spam are
> making their way through filters.  The majority of these spam messages
> receive SA scores of 0.3-1.6; for obvious reasons I cannot adjust the
> filter sensitivity this aggressively.  After reading through the docs
> I've come across the 'sa-learn' command which is supposed to help
> SpamAssassin learn beyond its scoring method - please correct me if I
> misunderstood here.

This should help, but if you are getting large amounts of spam scoring
less than 2 points, there is probably something else wrong.  What
version of SA are you running?  Have you run sa-update recently?  Are
you using the network rules?  Can you post a few sample spams (headers
and all) to a pastebin so we can see the spam and see which rules are
hitting?

-- 
Bowie