On Fri, Feb 28, 2014 at 12:07 PM, Mansi Gokhale <gokhalemansi12(a)gmail.com>wrote;wrote:
The image idea where users are asked to spot the odd
one out like
demonstrated or find all the similar images like mentioned in
here<https://www.mediawiki.org/wiki/CAPTCHA>
If you display 8 images and the user has to pick one, then even by random
guessing the attacker has a 12.5% chance of passing the captcha. That's not
good at all. Finding "all matching" is slightly better since it reduces the
guessability (1/256 for 8 images), but still not very good. A traditional
captcha using only A-Z is 1/308915776. To do as well with image picking,
you'd need to ask the user to choose the matches from a set of about 28.
Adding in numbers 2-9 is 1/1544804416, needing a set of about 31 images.
The set of possible images also needs to be very large and the
categorization private.
https://www.mediawiki.org/wiki/Talk:Requests_for_comment/CAPTCHA#Issue:_ima…
into much more detail on this issue.
Then there's the issue of different interpretation. Take for example
https://www.mediawiki.org/wiki/File:Find-all-captcha-idea.png. Is the
second image wearing glasses? Or is that a lorgnette or something like
opera glasses, both of which are held in front of the eyes rather than worn?
https://www.mediawiki.org/wiki/File:Find-the-different-captcha-idea.png has
a similar problem. The first image is the only one with a cigarette, and
the only one with non-realistic coloring. The second is the only bald one,
and the only one with something resembling a lorgnette, and the only one
not looking in the general direction of the camera, and the only one with a
book. The fourth is the only child. The sixth is the only obvious female
(I'm not sure about the cat). The eighth is the only one smiling, and the
only one with visible teeth.
Also a picture with a part chipped in could be shown and chipped pictures
could be given as options like find the missing part
from a jigsaw puzzle.
The image which would be shown is
http://imgur.com/uefeb08
http://imgur.com/KEJqCg3 is the picture which would be the correct option.
The other options could be rotated versions of this , which would not be
so easy for the bot to match. (unless it somehow worked some digital
processing algorithm and matched the color gradients or something like
that).
That seems very simple for a computer to solve. Just find the option with
minimal difference along the join edges, which is probably easier than what
they already do for OCRing text captchas.
As far as captchas, I still think
https://xkcd.com/810/ is the way to go.
--
Brad Jorsch (Anomie)
Software Engineer
Wikimedia Foundation