1

Suppose a user has just created a new account on some site. You ask him to submit an image of a para of neat handwritten text, along with a typed version of the same text. Then this is kept secret.

If some time in the future, there is a doubt as to whether a computer is accessing his account, you show him the handwritten text (one or two words) that some other user uploaded. Suppose he types correctly. He is now asked to type out one or two words from the text which he himself submitted. Only if both tests are cleared, the user is identified as human. He will have 4-5 attempts to clear test 1, but only 2 attempts to clear test 2, failing which, his account will be locked.

Is this system a fail-safe verification of human users? And does anybody use it? Why/ why not?

3
  • I get the feeling that when you wrote fail-safe, what you really meant was reliable in the meaning of can be trusted to produce accurate results most of the time. Also, what's the threat model here? Seems to me you are overcomplicating things quite a bit.
    – user
    Nov 27, 2015 at 18:35
  • Are you just proposing a crowd sourced CAPCHA? How is this better than existing systems? Nov 27, 2015 at 21:14
  • It is also worth noting that a large proportion of users are going to balk at getting out a scanner or camera to digitize the text, let alone writing it out in the first place. Also, you want the user to define the text? Mar 6, 2016 at 17:48

1 Answer 1

4

The second test is easily defeated because the user knows the original image. A bot could easily compare the selected words against the original image and determine which is the best match. In fact, they could submit an image which is intentionally easy for a computer to detect, such as installing a digital watermark or other stenographic content into the image.

The first test is interesting but has some oddities to it:

  • You presume everyone can read eachother's handwriting. Good luck reading mine!
  • You presume it is safe to send user generated content from one person to another (there are some images that many users would like to not see)

In all, the first part reminds me a lot of the CAPTCHA which had users writing phrases from books out of Google Books archives, which doubled as a service for digitizing otherwise hard-to-transcribe books.

In all, the only fail-safe way to determine if someone is a human is face to face contact, and even that is drawn into question as computers get more powerful.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .