Increased SPAM getting through

[Glenn Steen]

On 17/10/06, Mailscanner Admin <mailscanner at mckerrs.net> wrote:
> Has anyone else noticed increased SPAM getting through over the past week or so ? Especially where the spam is fairly obvious yet BAYES_00 is triggered ?
>
> here is one example;
>
> X-mckerrs-MailScanner-SpamCheck: not spam, SpamAssassin (score=0.334,
>         required 3, BAYES_00 -2.60, EXTRA_MPART_TYPE 1.09,
>         HTML_IMAGE_ONLY_24 1.84, HTML_MESSAGE 0.00)
> X-mckerrs-MailScanner-From: pixkmoq at tomieraines.com
>
> my spam score is 3.0 as you can see and the BAYES_00 seems to be the one the brings the score back under 3.

Yes, well... Is this by any chance an image based spam? (I'm thinking
that would be a rethorical question, given the rule hits you have). If
so, look into implementing ImageInfo and perhaps FuzzyOcr
(ongoing/recent threads on this liost point you in the right
direction;-). Bayes has likely not that much to work with, so... that
would probably explain why it fires the _00 rule... And training will
not help you much with that type of spam...
-- 
-- Glenn
email: glenn < dot > steen < at > gmail < dot > com
work: glenn < dot > steen < at > ap1 < dot > se