Mailing lists for programs are typically full of problem reports.  Here
is the reverse - the equivalent of a thoroughly happy baby-birth story
in the middle of a book on childbirth and all that can go wrong with it.

I have been running SA 2.55 for 17 days on my own account.  I have RBL
checks turned on and Bayes trained with about 1800 spams and 1800
non-spams, both sets carefully checked manually.  This is detailed at:

  http://www.firstpr.com.au/web-mail/Postfix-SA-Anomy-Maildrop/

This is on Red Hat 7, calling SpamAssassin from Courier Maildrop's
.mailfilter file, with all the mailing list messages previously sorted
so they don't go through SpamAssassin.  Messages which are not deemed to
be spam go through Anomy Sanitizer to defang suspect HTML and
attachments and to filter out messages with executable attachments.

After a week I adjusted the scoring of Bayes mid range and higher scores
upwards.  I have the threshold set to the default 5.0.  Extrapolating
those changes back to include the first week, I have this happy result:

  0 false positives.
  6 false negatives.

  630 spams correctly detected.
 ~350 non-spam emails correctly identified as non-spam.

This is a 99.06% success rate.

The highest score of the non-spam messages was 3.6 - most scores were
negative.  The scores of the false negatives, arranged by Bayes score, were:

Bayes range   Total score

 1-10%        3.7  (Just a few lines of invalid-looking HTML which
                    Netscape 7.02 would not render, pointing to
                    spammer's web sites.)

44-50%        2.8  (Base 64 encoded. I will write about this in a
                    separate message.)

44-50%        4.9  (Simple HTML - "exonerate your debt":
                    X-Spam-Status: No, hits=4.9 required=5.0
                    tests=BAYES_44, CLICK_BELOW, CONSOLIDATE_DEBT,
                    HTML_60_70, HTML_FONT_BIG, HTML_FONT_COLOR_BLUE,
                    HTML_FONT_COLOR_GRAY, HTML_FONT_COLOR_RED,
                    HTML_FONT_COLOR_UNSAFE, HTML_LINK_CLICK_HERE,
                    HTML_SHOUTING3, LOW_PAYMENT, MISSING_MIMEOLE,
                    MISSING_OUTLOOK_NAME, REMOVE_PAGE, SUBJ_YOUR_DEBT )

50-56%        4.2  ("attractive 43 yr old swf", quoted printable:
                    X-Spam-Status: No, hits=4.2 required=5.0
                    tests=BAYES_50, CLICK_BELOW, EXCUSE_3,
                    FROM_HAS_MIXED_NUMS, MISSING_MIMEOLE,
                    MISSING_OUTLOOK_NAME, NEVER_ANOTHER, REMOVE_PAGE )

56-60%        2.4  (Short HTML with two links.)

56-60%        4.4  (Plain text with link to spammer's site.
                    X-Spam-Status: No, hits=4.4 required=5.0
                    tests=BAYES_56, NO_REAL_NAME, PRIORITY_NO_NAME,
                    SEMIFORGED_HOTMAIL_RCVD )



So only two of the spams were scored below the highest non-spam.  I
regard this as highly successful and salute all those who contributed to
Spam Assassin.


  Thanks!

    - Robin





-------------------------------------------------------
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

Reply via email to