I have been noticing that the eye-readable text for most spam bears no
resemblance to the Reply-To where they are normally random characters, the
length of the Reply-To my be a combination factor to help differentiate it.
I wonder how good a spam sign it might be to calculate the correlation
between the 2 values. It will certainly catch some ham. Has anyone thought
about training a neural net on the user names from ham/spam corpii to see
what kind of confidence levels might show up to add a 0.1 to 2.5 score.
Here is one I saw today.
Reply-To: "Bill Miller" <[EMAIL PROTECTED]>
Reply-To: "Megan Mock" <[EMAIL PROTECTED]>
Reply-To: "Melinda Gibbs" <[EMAIL PROTECTED]>
Best Regards, Bob
Robert J. Strickler
-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive? Does it
help you create better code? SHARE THE LOVE, and help us help
YOU! Click Here: http://sourceforge.net/donate/
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
