Security By Sexiness

Via Valleywag comes word of an interesting technology implementation called the “Hot Captcha”. A “captcha” is essentially a test you perform in a web browser to prove you’re human. Its purpose is to keep automated bots from creating registrations and wreaking havoc on websites.

Typical captchas ask you to type a series of letters from a distorted image, answer a simple question, or perform mouse actions on objects. Users generally don’t like dealing with captchas because they require brainpower and sometimes trial-and-error, and there is no immediate pleasure gained from using them.

Enter the Hot Captcha from a fellow calling himself “frozenbear”. The Hot Captcha pulls photos of 9 women or 9 men using the open API from HotOrNot.com and asks you to select the three “hot” people from the group. The chances of picking the correct answer randomly are only 1 in 84 and you could bump that to 1 in 220 if you added a row, so the false positive rate would be extremely low.

What’s interesting to me is how effective this captcha test is, at least from my 20 or so tests. With these things, it’s not only important to make sure humans can easily pass the test but that computers cannot, and I honestly can’t imagine how a computer could programmatically determine the “hotness” of a photo. And what’s extra nice is that since you’re clicking on photos of attractive women or men, you could argue that this captcha may be the first ever that could be considered “fun” to use.

Interestingly, when I switched over to the “men” version, my accuracy went down from 100% to about 25%. I bet this effect would happen to a lot of men but not a lot of women. In other words, it seems to me that both women and men could easily identify “hot” women, while only men would struggle to identify “hot” men. Perhaps it’s because the physical traits that make men “hot” seem less obvious than the corresponding traits in women.

I wonder if this captcha could also be used as a reliable, surreptitious gender test for incoming visitors. As the operator of a social site which gives users lots of control, I’ve often thought to myself “If I could just tell for sure you were a female when you signed up, I’d probably give you ‘trusted’ status right away”. I say this not as a “ladies night” type of thing, but because the majority of evil-doers on the internet (e.g. spammers, network marketers, general assholes) seem to be men.

Though probably meant more as a joke than anything else, the Hot Captcha shows yet another function that the human brain is surprisingly efficient and consistent with that computers may never be able to match.

29 comments on “Security By Sexiness”. Leave your own?
  1. Andy Hume says:

    Amazing… I’ve not got a single hot man one correct. I clearly have shocking taste.

    Clever though, although I don’t see why it has to be such a subjective test. Are there many bots out there that can accurately recognise simple objects within photos, eg. chair, car, newspaper, etc..?

    And wouldn’t something as simple as “What’s this a picture of?” be effective in most situations?

  2. Julz says:

    Great idea, but I’m afraid I’ll have trouble implementing that at work ;-).

    However this idea could be “adjusted”, so that the test was for something else that was equally subjective, like which one is the kitten? (from a series of cats)

  3. Jehiah says:

    Another dynamic is how racial groups play into things in addition to gender groups. People are rather bad at facial recognition when it comes to racial groups they don’t have much contact with; for instance i can’t tell many Asian faces apart, they just seem so similar. I assume the same problem applies here and someone from say Japan might not be able to pass the Hot Captcha test even though they fit the right gender group

    … but then again, there is always a way to use that extra bit of info to your advantage as well

  4. Phil Scott says:

    I was unable to get past this. I generally only could find one or two attractive girls per post, and then I just picked the one with the biggest boobs figuring that’s how most Internet voters would vote. Maybe my preferenc towards athletic or geek girls makes my taste in girls all mesesd up

  5. Tigerblade says:

    I’d hate to be one of the people that gets pulled for the “not” status. See, this is why I don’t post to sites like HotOrNot… I could end up being used as a captcha.

    I can’t imagine how I’d do on the guys… not well I imagine. I’d prefer not to test my luck.

    Very creative though.

  6. Phil Scott says:

    I take that back. I just went on an absolute tear getting 10 right in a row.

    The real problem, of course, would come when someone happens upon themselves and has to pick someone else as being hot in order to login.

  7. Jeff Croft says:

    Freaking brilliant.

  8. David says:

    Haha, or they could just have you pick out the bad guys from a group of good guys. Anyone in the Bush administration would work just fine for this role.

  9. Jeff Wheeler says:

    I wasn’t able to pass on either the male or female tests once. Uh oh. :/

  10. Jack says:

    That’s a brilliant idea that’s probably been implemented in a slightly different way somewhere else. You could just as easily have photos of household objects and ask someone to click on 3 fruit or 3 kitchen utensils or 3 potted plants.

    Also, I feel sort of guilty for supporting an objective system for ranking attractiveness. It’s one thing to vote from 1 to 10 but it’s another thing to think that this person is objectively unattractive and everyone else on earth will agree to that.

  11. Fitz says:

    As a gay man it’s odd I couldn’t get thru the male test even once but had no problem with the females. I think they need to choose their male pix a bit more carefully.

  12. Dan A says:

    You know, I was going to ask about how well the gay male contingent would fare, but it seems Fitz has already answered. Great idea though, and I vaguely remember seeing one of these in place somewhere.

  13. Chris says:

    While this is doomed to be unreliable because of the subject matter, and the subjectiveness of ‘hot,’ I can definitely see how it could be adapted for other uses, such as choosing a particular set of subject matter from a larger set of random images.

  14. Rafael says:

    The site is down. Amazing what word of mouth on the internet can do.

    I’m going to register fluffycaptcha.com and make people choose the three fluffiest kittens.

  15. Tom Dolan says:

    Same as Mike, I pick hot girls like a mindreader but I blow it (maybe a bad choice of phrase there) on trying to pick the third hot guy. Maybe there’s a more universal set of characteristics for ‘hotness’ for girls that has as much to do with the style components than the actual hotness — i.e., photographed in lacey underwhere, cleavage visible, etc. The hot guys don’t seem to be exhibiting any characteristics like this consistently that would help one ‘cheat’ — interesting tho.

  16. Patrick Collins says:

    I’m thinking that this is an adaption of this MS Research project which I came across the other day. They are using cats and dogs and partnered with Petfinder.org to use their database of pets.

    http://research.microsoft.com/asirra/

  17. sam hotchkiss says:

    so… i’ve tried 4 times and can’t get it to think that i’m human.

    maybe they’re looking for “who’s skinny and slutty” not “who’s hot”

  18. Bruce Boughton says:

    Patrick, I think you’ve got your dependies the wrong way round. The Asirra site clearly cites HotCaptcha as an inspiration.

    I couldn’t switch to men to test out my skills on them, but I fared well on women.

    I don’t think this is really a viable method for big sites such as MSN, BBC, etc. as it’s not very PC. The cat/dog thing though…

  19. Ash Haque says:

    Considering using that type of captcha for your own site? :-P

  20. Simon Pioli says:

    Nice twist on a feature I’m not keen on at all.
    Captcha tests are difference between an accessible website and one that can only be visited by sighted, non-colour-blind users who happen to have images enabled…

    That isn’t to say I don’t see the benefits in them… Spam is immediately cut from forums, phoney accounts are reduced from web-based email systems etc etc…

    So, how do you maximise your potential visitors while keeping out the unwanted bots out?

  21. Alan H. says:

    lordy. here we go swinging-into the accessibility forest again! i was actually amazed it took this long.

  22. Simon Pioli says:

    lordy. here we go swinging-into the accessibility forest again! i was actually amazed it took this long.

    Well… Someone had to bring it up, eh? :p

  23. Ben Buchanan says:

    It’s an interesting exercise, but the implementation has some problems:
    1) “hot” is too subjective. what if a person simply can’t agree with the test? should they be blocked from the system because they don’t share the admin’s taste? as you’ve already noted, working out a random combination is going to take a looooong time.
    2) internationalisation is a problem – how do you account for cultural/racial differences?
    3) no way in hell this is going to get used in any corporate environment, where it would instantly result in equity complaints.
    4) what’s the accessibility fallback – a recording of “this photo is hot” or “this photo is not”? :) i wonder if voice recognition would be able to game that system?

    when captchas are inaccessible i think it’s generally because they’re created by coders with an overly-visual mindset – it’s not that captchas can’t be accessible, people just don’t think about it. it’s not good enough to dismiss the issue with a “here we go again”, it’s something that has to be addessed when developing new technology.

    using kittens might be better, but then “cute” remains relative – and what about dog lovers who don’t like cats? ;)

    perhaps the test could be done with a set of photos of dogs and cats – pick the cats to get in. the surrounding text can be translated to the localised language, the pictures themselves don’t need translation and aren’t subjective – a cat’s a cat, a dog’s a dog.

  24. Alan H says:

    I’ve seen basic math problems. Lullabot.com has one on contact forms, so I assume there’s a Drupal solution. I like the game idea. What about a quick game like tic-tac-toe… guess a number… which hand is it in? ;-)

  25. Ian Lloyd says:

    Fun, for sure, but not something you could rely on for accurate results. After all, one man’s ‘beautiful big woman’ is another man’s ‘chubby munter’. I’m so glad I was able to add to the richness of this discussion. :-)

    On a serious note, though, this would highly likely fail in many corporate environments:

    1) because if people in an office discovered that the site might randomly display NSFW images for something innocent, they won’t risk going there again and
    2) these kinds of things are usually blocked by firewalls, and hence will be impossible to get past (that’s already the case for me here at work)

  26. I just tested this by putting a female friend of mine in front of it. 10 or so failures on the hot man version later I think I’m pretty sure that the data on men on hotornot is seriously broken or that it varies too much between women to be useful. Personally I think it’s the former and that this is based on men going to hotornot and rating down hot guys out of jealousy or some other similarly lame emotion :P

  27. Hilary says:

    As a female I can honestly say that the hot males side is not valid at all. I failed 9 times in a row! However, I got nearly perfect results on the female side judging mainly by who was the biggest slut. I suppose humans vary too much for it to be an effectively used. Or maybe I’m just a messed up girl. =P

  28. Kara says:

    I’m with Hilary – I failed the male side 8 or 9 times in a row, but passed the female side every time. Apparently I have greater insight into what men perceive as being hot than women. Last I knew, I was in fact a member of the lady side of the species – I wonder what this says about me as a woman?

Leave a Reply

Your email address will not be published. Required fields are marked *

Subscribe by Email

... or use RSS