[Spambayes] text in image spam: feature suggestion

skip at pobox.com skip at pobox.com
Tue Aug 22 11:45:13 CEST 2006


    Tony> Sorry if this is already included. Would a simple hash of the
    Tony> image file be useful for eliminating spam? A lot of spams seem to
    Tony> send the same image with completely different text.

Maybe, though maybe not.  They do make changes in the image as well.  I just
saw several of the same one where there were various changes: font,
blinking, etc.  Also, with the pump & dump scams they seem to keep you
"updated" on the rising price of the stock as time goes on.  Any of those
would change the hash.  I am already computing the log of the size (then
truncating to an int).  I suspect that would be less sensitive to minor
changes in the image.

Skip


More information about the SpamBayes mailing list