<<< Date Index >>>     <<< Thread Index >>>

Re: [OT] Spam solutions (was: Sendmail vs. Exim, and SMTP Advice)



* It was Thu, Jun 17, 2004 at 03:24:39AM -0700 when Roy S. Rapoport said:
> On Thu, Jun 17, 2004 at 06:15:14AM -0400, David Yitzchak Cohen wrote:
> > Treating BCCed mail as junk doesn't guarantee no false positives, and
> > doesn't catch most of the new breed SPAM.  If you ask me, template
> > scanners are the most powerful anti-SPAM tool available that can
> > guarantee zero false positives.  If you have spamcatcher addys, you can
> 
> With the minor caveat that your system currently has about a 50% False
> Negative score :)
> 
> I'll throw in my hat for CRM114, BTW.  You can see its sucess on my system
> at 
> http://www.inorganic.org/~rsr/crm_success.txt
> 
> which is updated hourly.  Summary: Deployed for 82 days.  History for these
> 82 days: 99.45% accuracy, with 99.72% false positive avoidance and 99.74%
> false negative avoidance.  Now that I've been training it well, I believe
> the last 30 days are more representative of what it's capable of -- 16,920
> messages, 2477 tagged as spam.  99.935% FP avoidance and 99.9% FN
> avoidance. 
> 
> All with incredibly minimal hassle setting up and maintaining.

Have you used dspam at all? I've been getting corrupt databases with 2.x
and 3.x-beta versions of dspam using either mysql or berkdb. This is
easy to fix when it happens (I'm assuming it's weird chars in spam
because the corrupt table is dspam_token_data, anyways getting too OT)
but I've thought of trying out CRM114 anyways. Is it as simple as dspam
to get running? Seems like it might be worth a switch, although I get
great results from dspam too. (After a week or two of training I get
_very_ little spam in my inbox, and don't remember the last time I found
a false positive)

-- 
Sami Samhuri

Attachment: pgpWX0rnf9sCm.pgp
Description: PGP signature