False positives?

started at 20 Dec 2006 by bohica
  • bohica
    20 Dec 2006
    Richard - I posted on one of the blog pages before I remembered the forum. Sorry. In the dspam setup in the blog - have you forgotten to give instructions on how to address false positive retraining or did I miss it?

    Thanks!
  • Richard
    20 Dec 2006
    Tim, don't worry about the commenting....

    You are correct in your assumptions. I haven't encountered any false positives yet. But you can alter the existing training script to do this very easily.

    Create a mailbox folder called NotSpam.

    Rename the Junk folder in the script to NotSpam.

    Change the lines with:


     $DSPAM --user $USER --class=spam --source=error < $j



    into:

     $DSPAM --user $USER --class=ham --source=error < $j



    And you are done, haven't gotten around to it myself but I think this should do it.

    Hope this helps you out. Are you getting any false positives ?
  • bohica
    20 Dec 2006
    Yeah - came in today and everything I received overnight was tagged as SPAM. Of the 20 messages or so - 5 were not spam.... we shall see...

    Would you consider using the etc/postfix/transport method you originally tried so you could forward email to a cleaner address?
  • Richard
    20 Dec 2006
    Nope, I didn't like the forwarding as then you need to put the dspam ID in the mailbody of the message as not all mail clients leave the header in the original message untouched. I hate putting stuff in emails that doesn't belong there, all messages that you would reply to would have something like:

    2,458968cb103465209328925

    in there. I also think that if you start a maildialog you would end up with multiple Id's in your message and I don't think DSpam would handle that properly.
  • bohica
    28 Dec 2006
    All of my mail was being identified as spam. I updated the daily script - with a definition at the top...

    NOTSPAM_VIRTUAL_FOLDER=.NotSpam

    and the section just above the line

    done < $USERFILE

    with the following...

    echo `date` Processing False Positives as Clean
    # check if the user has a .NotSpam folder
    if [ -d $VIRTUAL_BASE$MAILDIR$NOTSPAM_VIRTUAL_FOLDER ]; then
    echo Entered User NotSpam Folder
    # check both new and cur directories for notspam !
    cd $VIRTUAL_BASE$MAILDIR$NOTSPAM_VIRTUAL_FOLDER/new
    for j in *
    do
    # check if the file exists
    if [ -s $j ]; then
    echo `date` Processing `pwd`/$j as Not Spam
    $DSPAM --user $USER --class=innocent --source=error < $j
    fi
    done
    cd $VIRTUAL_BASE$MAILDIR$NOTSPAM_VIRTUAL_FOLDER/cur
    for j in *
    do
    # check if the file exists
    if [ -s $j ]; then
    echo `date` Processing `pwd`/$j as Not Spam
    $DSPAM --user $USER --class=innocent --source=error < $j
    fi
    done
    else
    echo No NotSpam Folder
    fi


    it seems to work... but I will not know for a couple of days. I have not moved on to maildrop until I know that spam is properly identified. Thought I would share...

    I do not know what will happen if you move mail misidentified as spam into the NotSpam folder, and then leave it there to be processed multiple times?

    As for why things were being identified as spam -- the dspam factors were listing things like to: as .99 and things like subject *.is as .99. Wierd... we will see if it is just my configuration. Is there a way to reset the entire dspam stats for a specific user?

    Thanks!
  • Richard
    31 Dec 2006
    Thanks for the update, need to do this myself as well but as I said before, I haven't received anything indicated as spam that wasn't spam yet...

    Keep me posted on the progres.

    For starting all over again, I think you can delete the user from the database and you would start over again.

Reply

You must log in to post.