[OSM-talk] ReMAPTCHA Demo BETA 0.2 online! (Was: Hate captchas!!!!)

Stefan Keller sfkeller at gmail.com
Mon Mar 31 13:02:40 UTC 2014


Hi moltonel

2014-03-31 13:21 GMT+02:00 moltonel 3x Combo <moltonel at gmail.com>:

> > I'm not an OCR expert, but your scrambling looks less intense than
> most CAPTCHAs out there, so I assume that it isn't much of a challenge
> for current bots and consider it "solved".

The scrambling of the Control Word" is as intense as other Captchas but
it's only 5-6 chars (instead of 10 or more).
We can afford this because an OCR needs to find first the boundaries of the
word - and that's more difficult with labels on a map.

You have to realize that the other word has to be written just to indicate
if there is a path - else you can ommit it. The fact that there is a path
is unknown to our system.
So in your estimation, humans always succeded when only typing the "Control
Word".
That (human) trick and not knowing the correct answer for the "Control
Word" applies to all reCAPTCHAs.
Bots need first to find 1. which one is the Control Word (including
boundary) and then 2. to try OCR.

> Please drop the "scrambled text" idea altogether. And make solving a
> CAPTCHA a fun activity in the process.

Feedback so far was, that it's at least more fun than typing 15 characters
and helping G* instead of OSM.

> ...                        A "click features on the
> satellite imagery" task is one way to do it, but I'm sure there are
> others.

This seems like a good idea and I'm open to collect those.
Unfortunately nobody came up until now with one, which fulfilled the
properties of a reCAPTCHA, i.e. fast and easy to understand challenge by
humans.

--S.



2014-03-31 13:21 GMT+02:00 moltonel 3x Combo <moltonel at gmail.com>:

> On 29/03/2014, Stefan Keller <sfkeller at gmail.com> wrote:
> > Hi moltonel
> >
> > You wrote:
> >> I'm worried about bots still having a very high chance of sucess. With
> >> two fairly-legible words in the image and a chalenge asking me to
> >> write either one of the words or both, a bot still has 33% chance of
> >> success if answering randomly
> >
> > I can't follow what 33% means and why you are more worried than about the
> > usual CAPTCHAs: The known word (the one which is more scrambled) in our
> > ReMAPTCHA currently consists of 5 characters, and the word is placed
> > anywhere in the image - slightly tilted. This makes it at least as
> > difficult as usual CAPTCHAs.
>
> I'm not an OCR expert, but your scrambling looks less intense than
> most CAPTCHAs out there, so I assume that it isn't much of a challenge
> for current bots and consider it "solved". From that point onward, the
> challenge asks me to look at the map to decide wether to type word A
> or words A and B. If a bot can't read the map (that's what we hope),
> it'll just try random combinations: either word A, or word B, or words
> A and B. That's 33% of success.
>
> Again, you could scramble more heavily or add more words to choose
> from. But if you do that, ReMAPTCHA will quickly become as annoying
> (if not more) as traditional CAPTCHAs.
>
> The original idea of CAPTCHAs is that humans are better at reading
> scrambled characters than computers. But computers got better at that
> task, to the point that we have to make the task diffucult for humans
> in order to make it difficult from computers.
>
> Today we have a new type of task that humans are supposedly better at:
> interpreting satellite imagery. Even better: we're asking the computer
> to reproduce a brain process (armchair mapping) rather than a computer
> process (reversing the scrambling algorythm).
>
> Please drop the "scrambled text" idea altogether. And make solving a
> CAPTCHA a fun activity in the process. A "click features on the
> satellite imagery" task is one way to do it, but I'm sure there are
> others.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20140331/eb495c74/attachment.html>


More information about the talk mailing list