At 2:00 AM -0500 4/1/07, Richard Lynch wrote:
On Fri, March 30, 2007 2:29 pm, tg-php@xxxxxxxxxxxxxxxxxxxxxx wrote:
For those of you not following, what tedd's done with is blue dot is
created a larger image that contains a blue dot somewhere. You can
click in a fairly wide area, but only clicking on the blue dot yields
a success.
But it's probably not any harder to "crack" than the traditional
WORD_AS_IMAGE captchas -- One just has to write an algorithm to
edge-detect the blue circle instead of edge-detecting the letters and
converting them to ASCII.
Actually, edge-detecting the blue circle is probably easier, when you
get right down to it.
It's the same kind of basic "robot vision" that's covered in CS
Algorithms 101 class.
That's true to a certain extent, but robot vision has it's limits and cost.
The alpha channel image allows text and the image to merge as a
screen reader would view it -- as such the text on the page become
noise for edge detection routines.
Plus, I can certainly change the blue-dot image to anything I want
such as a red pig, yellow star, green leaf, white-dot (i.e., the
dimming or absence of text), even different text, or anything that
could be easily recognized by a sighted human but be a nightmare for
edge detection and subsequent object identification.
http://sperling.com/examples/p-captcha
I can also add logic to the presentation such as "click the largest
blue dot" and present three of different sizes -- or present "click
the tree" in a presentation of several different objects.
The number of different types of presentations and how easily the
presentations can be changed would be an effective method for
stopping bots. I can't envision how a bot could accurately solve such
problems.
As I see it, the only means a bot would have at successful navigation
would be a random "guess" at what the coordinates might be and then
the success of that is a function of how large the background is as
compared to the image (background pixels/image pixels) and how many
times they are allowed to guess.
Plus, what's to stop one from using multiple layer dot-capthca's?
Present two test and the odds against random success go up
tremendously.
Considering that most graphic CAPTCHA's are broken with respect to
blocking bots and are clearly approaching the point where even able
visioned users have problems with them, I see this technique as a
very viable alternative. Would you rather try to decipher twisted
letters in an image OR click the dot twice?
The biggest problem here, of course, is we are considering ONLY able
vision users. Those with impaired vision or limited cognitive or
physical skills would be also blocked and therein lies the rub. Thus,
my continued work to find a solution as with my recent work on my
audio CAPTCHA, regardless of speech recognition techniques.
I just need a way for vision impaired users to find the dot -- any suggestions?
Cheers,
tedd
--
-------
http://sperling.com http://ancientstones.com http://earthstones.com
--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php
