October 25, 2013

reCAPTCHA just got easier (but only if you’re human)



For over a decade, CAPTCHAs have used visual puzzles to help webmasters keep automated software from engaging in abusive activities on their sites. However, over the last few years advances in artificial intelligence have reduced the gap between human and machine capabilities in deciphering distorted text. Today, a successful CAPTCHA solution needs to go beyond just relying on text distortions to separate man from machine.

The reCAPTCHA team has been performing extensive research and making steady improvements to learn how to better protect users from attackers. As a result, reCAPTCHA is now more adaptive and better-equipped to distinguish legitimate users from automated software.

The updated system uses advanced risk analysis techniques, actively considering the user’s entire engagement with the CAPTCHA—before, during and after they interact with it. That means that today the distorted letters serve less as a test of humanity and more as a medium of engagement to elicit a broad range of cues that characterize humans and bots.

As part of this, we’ve recently released an update that creates different classes of CAPTCHAs for different kinds of users. This multi-faceted approach allows us to determine whether a potential user is actually a human or not, and serve our legitimate users CAPTCHAs that most of them will find easy to solve. Bots, on the other hand, will see CAPTCHAs that are considerably more difficult and designed to stop them from getting through.

A new and easier numeric CAPTCHA

Humans find numeric CAPTCHAs significantly easier to solve than those containing arbitrary text and achieve nearly perfect pass rates on them. So with our new system, you’ll encounter CAPTCHAs that are a breeze to solve. Bots, however, won’t even see them. While we’ve already made significant advancements to reCAPTCHA technology, we’ll have even more to report on in the next few months, so stay tuned.

34 comments:

  1. Although slightly better, still extremely difficult for visually impaired. Sadly the audio version is impossible to understand in most cases (example the audio and captcha for this comment is blank and when I downloaded the mp3 it was too distorted to hear) Accessibility should be of concern

    ReplyDelete
  2. Need to improve. some bot still can read it. maybe better if we can have some variation in number position. color and other custom things. so bot can not detect all. maybe they can detect the default, but with custom setting it will make more difficult

    ReplyDelete
  3. Interesting, but if the system has already determined that someone is human then why even present the easy captcha?

    ReplyDelete
  4. The above is a little confusing:

    "serve ... legitimate users CAPTCHAs that [are] easy to solve. Bots ... will see CAPTCHAs that are ... difficult ...

    you’ll encounter CAPTCHAs that are a breeze ... Bots ... won’t ... see them."

    If you can determine in advance who is a bot and who a human, why do you need a CAPTCHA at all?

    ReplyDelete
  5. What's the point of a captcha? To determine if you're a human.

    So... if you already know that I'm a human, why show the captcha in the first place?
    Unless the recaptcha system is really in place to just decipher the streetview photos of house numbers

    btw, to leave this comment, I didn't get a numerical captcha. Am I a machine?

    ReplyDelete
  6. What about the text digitization function that made reCAPTCHA so valuable in the first place?

    ReplyDelete
  7. If the software is capable of differentiating between bots and humans before presenting the captcha, then what is the point of the captcha?

    ReplyDelete
  8. Do I need to make any changes on my web site that uses reCAPTCHA in order to take advantage of these improvements, or are the changes automatic?

    ReplyDelete
  9. How can I get in contact with the team behind SafeBrowsing?

    I have a massive list of domains I think SafeBrowsing should include.

    ReplyDelete
  10. I am human but things have not become easier for me :-(

    Where do I complain?

    http://imgur.com/InqcLyx

    ReplyDelete
  11. How do I enable numeric only recaptcha? It's still showing as letters to me.

    ReplyDelete
  12. This particular user, however, is happy to continue to help decipher scanned texts via reCAPTCHA, even if the computer thinks I'm human.

    ReplyDelete
  13. And obviously, the non-generated numbers are house/mailbox numbers from the Streetview database that Google wants to make semantic.

    [EDIT: Solving a ReCAPTCHA to comment on an article about ReCAPTCHAs seems a bit too much of a good thing. :P
    1642 utofio. Wavy words are not quite done with, it seems.]

    ReplyDelete
  14. The megatypers Captcha works awesome and pays well so do register and use the invitation code which is mandatory to register.
    Invitation CodeS : 7VF1
    7VF2

    ReplyDelete
  15. Hi
    How can i make it only numeric
    we have used it in one site but the alphabets captacha is most of the time unreadable for many users
    can you tell the settings how to make it only numeric for making our life easy

    Thanks and regards
    Parvez

    ReplyDelete
  16. How can I configure the reCaptcha control to use this update? I found the current version is still showing very difficult words to read.

    ReplyDelete
  17. Actually, the captchas are only easier if you have any cookies from Google.
    So its actually Google's way of punishing those that don't use their services.

    Open up a private browser window and try a captcha, you'll know that they're significantly harder, not even difficult, they just automatically fail.

    ReplyDelete
  18. Actually, the captchas are only easier if you have any cookies from Google. Open up a private browser window, you'll notice that the captchas have black blobs in them and automatically fail.

    If you're being tracked by Google's cookies, then the captchas are easier. If you're not being tracked, then its harder.

    ReplyDelete
  19. The example shown shows up on my home Mac using Safari. However, on my Windows PC using Internet Explorer at work I am now seeing words that are even more indecipherable than in the past (two words with a black ink blot over them). This is at the "Public PAIR" page of the USPTO.

    ReplyDelete
  20. so how do we get the new AP?
    Woooow you are not even using it for this form, will this might be the answer to my q.

    ReplyDelete
  21. I personally think this is too bad. The point of reCAPTCHA, beside detecting bots, is to improve OCR. Improving OCR for numbers is not that helpful (if that is what it is even doing, but I assume it does not) as improving OCR for letters and other characters.
    It was a very nice and helpful idea and this essentially ruins it. :(

    ReplyDelete
  22. OCR has become very advanced, to the point that it reads the distorted letters with better accuracy than some humans.

    I'm pretty sure there's still a benefit (Google Street View) from getting the numbers right. As well as preventing bots abusing the system.

    So, no, it is not ruined.

    ReplyDelete
  23. I am confused - the text implies that form of captcha (numbers vs. twisted letters) depends on whether or not the system believes I am a human. But if the system already knows if I am a human, BEFORE I am presented a captcha, what is the point of showing it.
    Do I understand the term CAPTCHA ("Completely Automated Public Turing test to tell Computers and Humans Apart") in a wrong way?

    ReplyDelete
  24. To me it looks like the (second) numbers are house numbers. So perhaps it will help improve Google Maps?

    ReplyDelete
  25. How to use this new captcha? Like the one being asked in this comment box ?

    ReplyDelete
  26. Hi,

    my website always serves the very hard recaptchas. But I'm human!

    Why am I always detected as a bot?

    Best regards,
    --Martin

    ReplyDelete
  27. I run a website for older people, and need to more closely hug the line in allowing bots through versus the clients. Through testing, it seems that if the system ever triggered to think the client is a bot then they would essentially become locked out of the system as the words would be too difficult for them.

    ReplyDelete
  28. reCAPTCHA is not showing good please check this page http://www.topseoforum.com/ucp.php?mode=register

    ReplyDelete
  29. This comment has been removed by the author.

    ReplyDelete
  30. This comment has been removed by the author.

    ReplyDelete
  31. If I'm understanding this correctly, in an attempt to block spam and bots, the captcha will also get increasingly difficult over multiple attempts while being testing by web designers tweaking their code eventually bringing testing to a halt!

    ReplyDelete
  32. Alan, Check out http://techcrunch.com/2012/03/29/google-now-using-recaptcha-to-decode-street-view-addresses/


    At first reCaptcha was used to help digitize the 30% of words current OCR tech couldn't solve.

    Now that google owns recaptcha they are using it to help decipher google street maps house numbers, street signs and business names.

    I don't think it's ruined.

    ReplyDelete
  33. Once reCAPTCHA (wrongly) determines you're a bot, you start being served challenges with two almost impossible to decipher random letter sequences (both of them "control words").

    http://i.imgur.com/01F2eES.png

    Maybe you think you can read this, but believe me, you can't. I tried "ryntrInt llyHFrv" on this particular one, and it was wrong. And they're all this bad. It takes me an average 10-15 tries to get one of them right, which of course means that I'll never get out of the "hole" of reCAPTCHA thinking that I'm a bot, because I fail so often.

    I appreciated getting the easier, Street View captchas while I could still get them. But this adaptivity approach fails badly when going into the other direction (making challenges harder than classical reCAPTCHAs, rather than easier).

    From my very limited dataset of just my own encounters, reCAPTCHAs human/bot pre-distinction is atrociously bad. And from my own horrible experiences with it, assuming that many of my own legitimate users will have to deal with the same, I've had no choice but to ditch reCAPTCHA entirely. I cannot afford to subject my users to this torture and scare them away forever. It's gone too far. Especially since I hear that there are bots out there with automatic captcha solvers that have a much better hit rate than myself at this point.

    ReplyDelete
  34. Any way to fix the issue whereby the big reCAPTCHA default logo is displayed, which makes users click on it because it includes a huge circle with arrows that makes people think that is the place to click on to reload a new set?

    That logo has always been a big problem in terms of usability because it makes people click on it, rather than on the nearby very similar (but much smaller) reload symbol. I am surprised the issue is still there, as illustrated on top of this post.

    ReplyDelete

You are welcome to contribute comments, but they should be relevant to the conversation. We reserve the right to remove off-topic remarks in the interest of keeping the conversation focused and engaging. Shameless self-promotion is well, shameless, and will get canned.

Note: Only a member of this blog may post a comment.