Of all the forms of CAPTCHA available, which one is the \"least crackable\" while remaining fairly human readable?
I wonder if a CAPTCHA mechanism that uses collage made of pictures and asks human to type what he sees in the collage image will be much more crack-proof than the text and number image one. Imagine that the mechanism stitches pictures of cat, cup and car into a collage image and expects human visitor to tick (checkboxes) cat, cup, and car. How long do you think will hackers and crackers will come up with an algorithm to crack the mechanism (i.e. extract image elements from the collage and recognize the object depicted by each picture) ...