Re: Spam filter reviews?
On Thu, Mar 20, 2003 at 10:26:32AM -0800, Bill Wohler wrote:
> The Bayesian stuff did seem to reduce the spam ending up in my +inbox
> (dozens a day down to 3 or 4), but it still has the rules to generate
> false positives.
One thing I had to do to avoid missing important mail is to zero the
scores for some tests:
score RCVD_IN_MULTIHOP_DSBL 0
score RCVD_IN_RFCI 0
score RCVD_IN_OSIRUSOFT_COM 0
score RCVD_IN_UNCONFIRMED_DSBL 0
score RCVD_IN_SBL 0
score X_OSIRU_SPAM_SRC 0
I get important message from quite a few servers included in those lists
(including one of the servers I manage). Sometimes it's not possible for
the admin to remove it (as in my case). We have a DLS connection, and
there's only ONE company which provides such service in my area.
And guess what? Both home and busines users will get blacklisted IPs,
because the company wasn't competent enough to interact with the
blacklist owners and fix some issues.
> I'm trying out bogofilter now. It obviously still has a lot of learning
> to do, so I'll report back here in a week or so.
I do something like "If bogofilter doesn't catch it, run it through
spamassassin". Works well (most spam is caught by bogofilter, but some
gets through, and SA tags it as SPAM).
J.
Reply to: