Mobile internet users are prompted to solve the same text-based CAPTCHAs used on desktop computers, but it has been found that mobile devices are poorly suited for solving these text-based CAPTCHA tests. These tests are more difficult on mobile devices because text is time-consuming to enter and the size of mobile displays can make the CAPTCHA hard to see and interpret. We introduce a new CAPTCHA mechanism called Picturesque specifically designed for mobile phone devices, based on image recognition techniques.
3. Design Problem
Mobile internet users are
prompted to solve the same
text-based CAPTCHAs used
on desktop computers, but it
has been found that mobile
devices are poorly suited for
solving these text-based
CAPTCHA tests
4. Solution
An image based CAPTCHA that does
not require keyboard based input
7. Goals
Typing on mobile
devices is hard, so no
more keyboard based
input
Design
Goals
8. Goals
Typing on mobile
devices is hard, so no
more keyboard based
input
Design
Goals
Should take less than a minute to
solve
9. Goals
Typing on mobile
devices is hard, so no
more keyboard based
input
Design Must be highly secure
Goals to avoid any SPAM
attack
Should take less than a minute to
solve
10. Contextual Inquiry & Sketches
We chose few representative tasks in which CAPTCHAs are
generally encountered (for example user registration) and
asked our test users to complete those tasks
11. Contextual Inquiry & Sketches
We chose few representative tasks in which CAPTCHAs are
generally encountered (for example user registration) and
asked our test users to complete those tasks
As a result of contextual inquiry, we conducted a brainstorming session to
sketch three concepts for our design
12. Paper Prototypes
Evaluated all the three sketches rigorously keeping in mind the
design goals & user needs and as a result of evaluation one of
the three concepts was chosen for paper prototypes
Design for our paper prototype
consisted of two steps, first
users had to identify images
from specific category and
then they had trace the
outline of identified image
13. Design Iteration
Based on the results from user studies, we realized image
tracing is neither fast nor obvious
Choosing images is much more easier from user standpoint
15. Interactive Prototype: Design
User is asked to select
images from a specific
category, such as Eiffel
Two grids of 9 images in
Tower, Lions, or Shoes
3-by-3 format
16. Interactive Prototype: Design
User is asked to select
images from a specific
category, such as Eiffel
Two grids of 9 images in
Tower, Lions, or Shoes
3-by-3 format
User has to select images
that are related to
chosen category to pass
the test
17. User Studies
Extensive study with 61
subjects using Amazon
An in-person study with 6 Mechanical Turk
users
Image: http://www.flickr.com/photos/leeander/4132537169/
18. User Study Results
Both studies compared our Picturesque
technique to reCAPTCHA, a common text-
based CAPTCHA
Based on the results, we observed that time
completion rate for a task of Picturesque was
better than reCAPTCHA task in most of the
cases
19. Picturesque: Final Design
# of images that needs to
be selected in a grid is
randomly chosen to be
either 4 or 5
20. Picturesque: Final Design
# of images that needs to
be selected in a grid is
randomly chosen to be
either 4 or 5
Placement of the images
is random
21. Picturesque: Final Design
# of images that needs to
be selected in a grid is
randomly chosen to be
either 4 or 5
Placement of the images
is random
Two grids of nine images are necessary from security perspective