Greg A <[EMAIL PROTECTED]> writes: > I have an issue with the whole Ham/Spam concept in SpamAssassin. I > will define the terms in case you are not aware of the terminolgy > used. > > Ham=good wanted email > Spam=bad unwanted email
Those are not the generally accepted definitions and definitely not the ones used by us. spam = unsolicited bulk/commercial email ham = everything else > SA gives supposed ham lower scores in a few areas, Bayes and > auto-whitelist for instance. Exceedingly few instances in the forthcoming SA 2.60. > I believe we should not care about ham at all and we should not > attempt to identify ham. Anything that is not spam is by default ham > anyway so trying to identify both ham and spam is more difficult and > more error prone and expoitable than just trying to identify spam > alone - in my opinion. We've always tried to whatever works. By "works", I mean whatever helps classify email as either spam or ham. The formula for what works _best_ has changed over time and will probably continue to change. If you attempt to classify email as spam or not, then you've made a ham pile too. That's just how classification works -- you must have at least two piles. Yes and no, whatever. Now, if you want to talk about SA rules that have negative scores, then we've had to adapt somewhat with the increasing sophistication of spammers and SA 2.60 will do a much better job. Only a few hard to forge negative rules (whitelist, BondedSender, some Bayes, and Habeas) are still included. > If you feel the same way simply disable all items in SA that can > downgrade spam and make it look like and train like ham. For instance, > set all bayes tests of 50% and below to a value of zero. Set > auto-whitelist to disabled. Also set any SA test values that give a > large negative score to zero. This should help make your SA install so > that it is not expoitable. Not exploitable, yes, but at the expense of more false positives. SA scores are tuned as a group, you shouldn't disable rules set without a specific reason and definitely not en masse. Now, if you have statistics that show doing this produces an overall favorable result, I'd be interested in seeing them. > I'm sure there are many things you can do to make your SA even more > accurate. I have some other tricks but I won't tell you all that I do > to custom configure my SA, it is best that we all define our own rules > so that spammers always have a moving target. (Look for SA tests in > the headers that would have correctly marked your spam and raise those > scores for instance.) The nice thing is that the SA developers have > made it so that you can customize SA any way you want it. I use SA in > a sitewide corp environment. Okay, this I can agree with. :-) Daniel -- Daniel Quinlan anti-spam (SpamAssassin), Linux, and open http://www.pathname.com/~quinlan/ source consulting (looking for new work) ------------------------------------------------------- This SF.Net email is sponsored by: INetU Attention Web Developers & Consultants: Become An INetU Hosting Partner. Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission! INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
