This may not be a new idea but wouldn't a new RBL based on image spam be
worthy? I've been testing FuzzyOCR recently and although it seems to work it
seems sort of "brute force". From what I understand it converts the images
to a PPM or PNM format and then runs gocr over those images and attempts to
find words form your wordlist in the result.
Couldn't there be an RBL established similar to DCC, Razor or Pyzor that
maintained the hash of known spam images and forego the wordlist detection?
Skipping the OCR tests would lessen the load on mail servers as well.
Granted there would be concessions and a little loss of flexibility to this
method but an image spam is an image spam in most everyone's book.
BTW, I intended to post this to the Fuzzy list but I'm waiting for my
subscription to be approved by the moderator.
Bill
