VMUNIX Blues

by Mark Mayo

1. Chris Adams March 9th, 2005
   19:55 GMT

   I found it’s easier to simply continue training SpamAssassin - the way I work is simply by having a cron entry run sa-learn on read messages in my inbox and spam folders. When a message is miscategorized simply move it to the appropriate folder and make sure it’s marked as seen:

   find ~/Maildir -type d -maxdepth 1 -mindepth 1 ! -iname .Trash ! -iname .Spam ! -iname .unblocked-spam ! -iname .junk -exec find {}
   -type f -mtime -3 -name \*S \; | sa-learn –ham -f -

   find ~/Maildir/.Spam/cur/ -name \*S | sa-learn –spam -f -

   Note that this approach assumes you use a Maildir-based IMAP server; for a pure IMAP approach I’ve actually used the Perl Mail::Box module to make an IMAPS connection and loop over the appropriate mailboxes - the only drawback to that approach is that it requires you to either run it interactively or store your mail password in a file (yech).
2. mark March 11th, 2005
   9:49 GMT

   That’s a neat tip. Thx! I do use Maildir, so I may try this out and see how well SpamAssassin does with better training. I’ve also tried dspam out recently, and it seemed to be quite good as well. One thing I really like about SpamSieve now that I’ve been using it for a couple weeks is the corpus browser. Really handy to drill down to some words and see how they’re categorized. The instant stats screen is nice too. The big problem with a spam filter that runs on the client, however, is that you always need to leave the client running. Well, it’s a problem if you’re like me and check your email from a variety of computers, anyways. Latest stats:

   Filtered Mail
   522 Good Messages
   8696 Spam Messages (94%)
   468 Spam Messages Per Day

   SpamSieve Accuracy
   1 False Positives
   83 False Negatives (99%)
   99.1% Correct

   Corpus
   1684 Good Messages
   2385 Spam Messages (59%)
   129844 Total Words

   Rules
   1605 Blocklist Rules
   863 Whitelist Rules

Powered by Hemingway 2.5 flavored Wordpress.

Entries RSS Comments RSS