This is a discussion on (OT) Erratic spam filtering by SpamAssassin within the Slackware Linux Support forums, part of the Unix Operating Systems category; --> Hi all, I receive close to 100 spam mails a day. So I set up spamassassin (2.60-rc2) from source. ...
| |||||||
| FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| Hi all, I receive close to 100 spam mails a day. So I set up spamassassin (2.60-rc2) from source. I receive mail thru fetchmail and filter it thru spamassassin by creating .procmailrc and .forward files as given in SA INSTALL file to receive good mail in /var/spool/mail/srikanth and spam in a "caughtspam" file. Initially, filtering was good. THen after sometime, 10 to11 spam mails leaked through. I then used training of SA by giving "sa-learn" since I have a 20 MB caughtspam file now with close to 2200 spam mails. The funny thing is , each day, before I receive mail, if i train SA by giving "sa-learn --spam --mbox /home/srikanth/caughtspam" and then if I receive mail using fetchmail, filtering is really good and no spam comes through or only one or two comes thru. But if I retrieve mail thru fetchmail without going thru "sa-learn" first, it is the old story again, 10 or 11 spam comes thru. I checked it more than 5 to 6 times by both methods and the result is consistent. Why? Does SA needs to be trained everytime? Does it forget what it previously learnt? I think not , since it says it learnt only from 7 or10 messages and I have not given any forget options to "sa-learn". Can anybody enlighten ? NS Srikanth |
| ||||
| -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 NotDashEscaped: You need GnuPG to verify this message On Thu, 06 Nov 2003 22:30:08 +0500, N S Srikanth <nssrikanth@hotmail.com> wrote: > Can anybody enlighten ? Yeah, folks on the SpamAssassin users mailing list, I think.. It sounds all like coincidence to me... spam isn't consistent enough to get a good sampling, and a few leaking through won't quite do that. Just to be sure, you're training SpamAssassin with ham as well as spam, right? A good mixture, I've heard, is about 60% ham, 40% spam. -- Rob | If not safe, Email and Jabber: | one can never be free. athlonrob at axpr dot net | -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.3 (GNU/Linux) iD8DBQE/qsbghm6KEoOOAe0RAtNSAJ9E/0mb5VKQQrnqJZrBD24bOPnK5QCguMNP mfmlIs0FIZP6kcvfJneUfmk= =4URl -----END PGP SIGNATURE----- |
| Thread Tools | |
| Display Modes | |
|
|