I thought reCaptcha was meant to be OCR-safe



  •  Registering here, got a captcha which looked a bit easy. Got my phone out, and Bing recognised the word correctly. 

     



  • "meant" != "is"



  •  With reCaptcha, only one item is actually a Captcha, the other is a word from a digitised book that wasn't OCRed properly and is crowd-sourced through reCaptcha for verification.



  • @rpjs said:

     With reCaptcha, only one item is actually a Captcha, the other is a word from a digitised book that wasn't OCRed properly and is crowd-sourced through reCaptcha for verification.

    All right, everyone... from now on, enter "FUCKASS" for one of the words whenever you fill out a captcha! >:-D



  • @ekolis said:

    @rpjs said:

     With reCaptcha, only one item is actually a Captcha, the other is a word from a digitised book that wasn't OCRed properly and is crowd-sourced through reCaptcha for verification.

    All right, everyone... from now on, enter "FUCKASS" for one of the words whenever you fill out a captcha! >:-D




     



  •  Here's an interesting talk from one of the guys who developed reCaptchas and their new project:

    [url]http://www.ted.com/talks/luis_von_ahn_massive_scale_online_collaboration.html[/url]



  • @joshbuckley said:

    Bing recognised the word correctly. 
     

    It recognized the perfectly legible part of the captcha.

    Call me when it reads the googly bit.



  • I got one ouf these today:




  • @dhromed said:

    @joshbuckley said:

    Bing recognised the word correctly. 
     

    It recognized the perfectly legible part of the captcha.

    Call me when it reads the googly bit.

    Read rpjs's post. The "googly" bit is the created one, the other is the one it failed to ocr correctly and wants you to solve.

    OTOH, I guess most people have seen far worse captchas.

     

     



  • Edit timed out, sorry for double posting:

    The "googly" bit would still be pretty easy to solve for a
    solver dedicated to this kind of captchas. (The reCaptchas all have this
    kind of transformation added, so creating one could make sense if you
    want to defeat it).

    It's just a simple text of random letters, put
    on a swirly base line with a deformed "circle" XORed over it. Detecting
    the circle boundary should be doable for a reasonable percentage of
    captchas and straightening the text is a solved problem, too. Just takes
    way too much effort for your usual spammer.

     



  • @topspin said:

    Read rpjs's post.
     

    My post still stands.



  • I thought you're the local image wiz?

    Take a look at the gradient of the image. Although it's an actual picture of the screen, it's not too hard to read.
    IMO, the hardest thing might actually be the low resolution / low quality / smudged characters, not the distortion.

    Gradient



  • So long as Google is continuing to make money from their OCR solving contracts with such clients as the New York Times, the consensus will be that reCaptcha works, regardless of evidence to the contrary. I got a kick out of the fact that journalists unquestioningly accepted a report from researchers at Stanford U that implied that reCaptcha is 100% resistant to computed attacks. Nobody wondered if a university department that is heavily subsidized by Google might be a little bit tender of producing results that hurt their products. By consensus, most journalists ignored a report published at the same time by Newcastle University that produced startlingly different results.

    The man who created reCaptcha, Luis Von Ahn, has said himself that he didn't expect his technology to remain viable form more than five years. This is why I, along with some other small companies have been researching alternatives to glyph-based CAPTCHAs. These solutions include Confident Technologies' solution, (http://www.confident.com), the new upstart Are You Human, (http://areyouahuman.com), and my own invention: VouchSafe (http://www.vouchsafe.com).

    Naturally, I think my idea is the best. ;) However I think that considering the fact we're all up against a monolithic titan, you're not going to hear much about alternatives until reCaptcha breaks down completely.


Log in to reply
 

Looks like your connection to What the Daily WTF? was lost, please wait while we try to reconnect.