Re: Scanning email

Tim Schneider wrote:
> I have trouble seeding data for new email accounts.
>
Are you using the merged groups feature, or not? I am, and it helps this
problem quite a bit. However, new accounts still need some training,
even with merged groups.
> as far as I can tell, dspam does not scan images or other attachements to
> the email message so these new fangled image spams blow right through
> dspam unless it is able to trigger on the header information.
>
No, it has no real way of classifying attachments other than the
attachment names. If you're after this, I believe there is a
SpamAssassin plugin that does OCR on images to try and classify them.
However, I think the setup is somewhat non-trivial, especially if you
want to do things to optimize the performance like caching repeat image
results.

In any case, I find that my DSpam installation now does catch most of
these, it just required more training than regular spam. However, I also
don't have friends who just send me e-mails with only a picture and
nothing else in the message body, so DSpam could essentially be assuming
that any e-mail for me that just has an attached image and nothing else
is spam.
> I'd recommend storing the user preferences in the sql tables, the
> procedure is in the documentation, and then running the sql maintenance
> cronjobs. This helps keep the sql data under control.
>
Indeed, both of these are almost required.

Graeme
Received on Thu Dec 14 14:13:08 2006