[Mailman-Developers] Re: Opening up a few can o' worms here...

Ka-Ping Yee ping@zesty.ca
Mon, 29 Jul 2002 19:53:15 -0700 (PDT)


Barry wrote:
> Do you want to attempt to obscure the
> address (e.g. "barry--at--python--dot--org")

Chuq wrote:
> Anything you programmatically obscure will be programmatically
> de-obscured.

I don't think this is entirely true.  This is probably true if you
invent ways of rearranging text that is still readable as an e-mail
address, but one possible good solution is to use an image instead.

If you generate an image containing the entire e-mail address, it
can be made extremely hard to read, even with state-of-the-art OCR.
And it raises the bar so high that it will be a long time before
spammers bother to even try OCR -- it's much too easy to harvest
addresses in other, simpler ways.

Just to give you an idea of what tricks are possible: a GIF can
contain many different colour table entries that map to almost the
same colour; the background can be patterned; the text can be
distorted or blurred; the text can be drawn shadowed or embossed;
the text can be broken into multiple frames of a GIF animation;
etc. etc.  Many of these things can be done subtly enough that the
text is still trivial for anyone to read, yet it becomes phenomenally
difficult for a machine to decode -- to the point where the spammer's
best bet is to take a video capture of the screen, print it out,
run it through a scanner, and run a fancy OCR program on the result.


-- ?!ng