These slides illustrate an idea of mine that I developed further with Kathrin during a course.
It outlines how I would approach the CAPTCHA problem with a collaborative approach (I call it FUNTCHA). Humans provide riddles for other humans, containing of a picture and a question. As in CAPTCHAS, the answer is one word, but the search space to crack FUNTCHAs can be astonishingly high, as it requires general problem solving ability (as humans possess it) to come up with an anwser (if the FUNTCHA is made good).
The problem is less if this would work, but if a large enough community of riddle-loving people can be developed.
Here is the original blog post where I came up with the idea: http://www.nicolashoening.de/?blog&nr=63
7. All work and no play...
• today’s approaches make users repeat
themselves
• approx 150,000 man hours a day
• arms race (e.g. cruder letters in Captchas)
leads to bad accessibility
7
8. Captchas will be solved
• A lot are already solved
• http://sam.zoy.org/pwntcha
• Image recognition makes great advances
8
10. Funtcha
• FUNny Turing test to tell Computers and
Humans Apart
• the idea: let a community of humans create
riddles - let the people define how they
want to be tested
• 1. humans like surprises
• 2. machines do not
10
13. Deliver fitting Funtchas
• they should fit the user
• they should have worked before
• this should cost the user minimal effort
(remember: the goal is to have fun)
13
14. Deliver fitting Funtchas
• We learn how solvable a Funtcha is
• Users can demand a different Funtcha
(express dislike)
• We use tagging to match users of websites
and Funtchas
14
16. Funtcha Selection
• select candidates based on tag similarity
• sort list according to solved/disliked ratio.
give new Funtchas a chance to be highly
ranked
• choose one randomly out of the best n
Funtchas
16
19. Open Issues
• Accessibility: Image not a must, also Audio
• Security: SSL, changing filenames, etc.
• Copyrights
• Human Labour
• Personalisation: Creators, Websites
• Numbers
19