Exactly. Using the results of DSPAM within SA or SA to train DSPAM causes problems. However, I am using a multi-layer defense; since no one tool is perfect. I want to deny a significant percentage of Spam at SMTP session time via an ACL; however, at this stage of the game, I do not want any false positives. I use SA as part of the SMTP ACL validation with specific rules disabled or added (such as Bayes is disabled in SA for our system). I also require a higher score; again to prevent a false positive. After I accept the email, I do not want to just drop it if it is spam. I use DSPAM for this final filtering; giving the users complete control.
Tim -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Craig Whitmore Sent: Wednesday, November 16, 2005 11:17 PM To: Timothy Spear; 'Marc Perkel'; [email protected] Cc: Timothy Spear Subject: Re: [exim] Who likes DSPAM? Statisticlly using DSPAM and then using SA with its results will make it alot worse that just using DSPAM SA gets ~95% correct, but DSPAM can get more like 99.995% correct after a while.. Heurestuc Spam systems such as SA (yes HA bas bayesium stuff, but its crap in its accuracy)) need constant updateing with new rules. Stastistic systems such as DSPAM/Death2Spam etc just work. I've love to see anything that will do better than DSPAM on the market (Free or Commercial) Any Anti-Spam people companies out there on this list want to compare it with DSPAM? Thanks Craig http://www.spam.co.nz ----- Original Message ----- From: "Timothy Spear" <[EMAIL PROTECTED]> To: "'Marc Perkel'" <[EMAIL PROTECTED]>; <[email protected]> Cc: "Timothy Spear" <[EMAIL PROTECTED]> Sent: Thursday, November 17, 2005 5:09 PM Subject: RE: [exim] Who likes DSPAM? > Yes, SA is one of my SMTP time filters as part of an Exim ACL. It is set > to > high enough to not have false positives, but blocks about 60% of the SPAM > which make it that far (I have other simple rule filters before SA; such > as > Distributed Check Sum). And just a note: I do not use the result of my > other > filters to train DSPAM. That tends to cause problems. As for the > percentage, > yes DSPAM can do that, although I do not know how. > > Tim > > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On > Behalf Of Marc Perkel > Sent: Wednesday, November 16, 2005 10:06 AM > To: [email protected] > Subject: Re: [exim] Who likes DSPAM? > > > > Timothy Spear wrote: > >>There is no reason to place the result into SA. SA has its own bayes >>implementation (which I think is not as good). I use SA (it is the final >>check before acceptance) during the SMTP session without the bayes filter; >>just rule based. >> >>Tim >> >> >> > > That's interesting. So SA is a prefilter to DSPAM? Do you reject email > at SMTP time with SA and then use DSPAM on what's left? > > Spamprobe has a way of returning a one line score. I can: > > cat message| spamprobe score > > And I get a result with a number line 0.988234 so I can then process > that number. Can DSPAM do that? I'm also having problems compiling it > under Fedora Core 4 - can't find libmysqlclient - but I should go to > their support forum to solve that. > > -- > ## List details at http://www.exim.org/mailman/listinfo/exim-users > ## Exim details at http://www.exim.org/ > ## Please use the Wiki with this list - http://www.exim.org/eximwiki/ > > > -- > ## List details at http://www.exim.org/mailman/listinfo/exim-users > ## Exim details at http://www.exim.org/ > ## Please use the Wiki with this list - http://www.exim.org/eximwiki/ > -- ## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://www.exim.org/eximwiki/ -- ## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://www.exim.org/eximwiki/
