On Tue, Jul 29, 2008 at 05:00:53PM -0500, Len Conrad wrote: > >On Tue, Jul 29, 2008 at 02:36:36PM -0500, Len Conrad wrote: > > > >We're trying to replace a Windows anti-spam on the mailbox servers > > > >with amavisd/sa/clam on the front-end mx. > > > > > > > >We are running in tandem both now in the amavis/sa/clam testing phase. > > > > > > > >The backend mail content-scanner is still catching too many true > > > >spams that get past amavis. > > > > > > > >We uploaded the spams caught by backend to the mx and ran them > > > >through spamc, with these results: > >... > > > > > > here's an example of a true positive msg caught by the backed after > > > amavis passed is as clean: > > > > > > "spamc -c" says against the DATA body checks out as: > > > > > > 70537530.eml 10.3/5.0 > > > > > > > > > but in the 70537530.eml file: > > > > > > X-Spam-Score: -2.27 > > > X-Spam-Level: > > > X-Spam-Status: No, score=-2.27 tagged_above=-20 required=5 > > tests=[AWL=-1.556, > > > BAYES_50=0.001, HABEAS_ACCREDITED_SOI=-4.3, > > HTML_IMAGE_RATIO_04=0.172, > > > HTML_MESSAGE=0.001, MIME_HTML_ONLY=1.457, URIBL_BLACK=1.955] > > > > > > so amavis correctly decided "Passed CLEAN" for -2.27, but the 12+ > > > discrepancy between spamc and amavis scores" > > > > Are you running spamc as the same user amavisd runs as? > > good point, was su'd to root > > > If not - and > >I suspect you're not - different per-user SA settings may be the source > >of these differences. What score do you get when you run spamc on that > >same file as the amavis user? > > I changed the file to vscan:vscan, su'd to vscan and command line > spamassassin returned: > > Content analysis details: (6.8 points, 5.0 required) > > spamc -c running as vscan returns: > > $ spamc -c < /home/harry/declude/20080729/70537530.eml > 10.2/5.0
Very strange. So you're getting 3 different sets of values here - even spamc vs. command-line are disagreeing. Hopefully someone else will have an idea, as I'm fresh out at the moment. > > The other problem I see there, very likely related, is that it > >appears you've let amavisd run with Bayes and auto-whitelisting but > >without any feedback of spam/nonspam, and so it's autowhitelisting > >things it shouldn't and probably also has a polluted Bayes DB - note > >that "AWL=-1.556" indicating that spam had been whitelisted. > > we have local.cf with bayes_auto_learn 1, but I can' see we have > auto-whitelisting configged anywhere. And yet I see that AWL=-1.556 in the report... -- Clifton -- Clifton Royston -- [EMAIL PROTECTED] / [EMAIL PROTECTED] President - I and I Computing * http://www.iandicomputing.com/ Custom programming, network design, systems and network consulting services ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ AMaViS-user mailing list AMaViS-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/