One of the new tricks spammers are using is placing their spammy text in
an image and trying to bypass Bayesian scanning with junk text. The
pharma and stock spams seem to be particularly fond of this tactic.
I've been using the FuzzyOCR plugin for SpamAssassin
(http://fuzzyocr.own-hero.net/) to combat this. It has worked very well,
but does require a bit of setup to get going.
ISP Lists wrote:
> I've got vpopmail/netqmail built using typical clamav and spamassassin
> (clamd, spamd). I've got XBL filtering and CHKUSER enabled on smtp. I'm
> actively training my Bayes filters. I do not use verified sender or SPF.
> Spamassassin's local.cf look like this:
> required_score 6
> rewrite_header Subject [SPAM]
> report_safe 0
> use_pyzor 0
> use_razor2 1
> use_dcc 0
> dcc_home /var/dcc
> skip_rbl_checks 0
> rbl_timeout 3
> score RCVD_IN_BL_SPAMCOP_NET 2
> use_bayes 1
> bayes_auto_learn 1
> bayes_path /home/spamd/.spamassassin/bayes
> I STILL find a good bit of spam is getting through. (pharma, mortgages,
> stock hype, etc)
> I wonder whether there are other/better anti-spam tools I should use to
> cull the spam more effectively. Suggestions most welcome.