You're marketing. Not the time for balanced language.
I know that in the past some captchas were broken at rates of 30-40% (Hotmail, I think). Is this really still the case?
But more deserves to be said. A site like Hotmail is a big, focused target for spammers. They face pain if a bot can beat their CAPTCHA even 10% of the time. Smaller sites get spam too, but they generally get it from dumber bots looking for unprotected sites. A 2 panel peoplesign CAPTCHA is probably all they need.
With 2 panels at 6 labels each, a random guess has a 1/36 chance to pass. Fortunately, a large number of incorrect guesses is behavior that can be detected and blacklisted.
Futhermore, peoplesign can be customized to offer more security. Number of panels (and soon number of menu labels) and other characteristics can be modified. Check out a customization demo at http://peoplesign.com/main/pickTheLabelDemo.html
Has this stood up to any bot attacks yet? What's the significance of the image discoloration?
But don't throw the proverbial baby out with the metaphorical bath water! The literacy requirement may be an effective defense against the armies of human sweatshop CAPTCHA solvers that we hear about in the news.
Regarding attacks, would be attackers have more important targets for now. peoplesign isn't yet protecting any major sites.
peoplesign images have been recolored using a "secret sauce" algorithm. Recoloring one of two similar pictures reduces effectiveness of color histogram matching. It even reduces effectiveness of more sophisticated color profile matching techniques. In short, the recoloration is defense against a particular type of attack.
Also, what turned out to be a hydrant looked to me like the leg of a moon lander. (I stared at that thing for a really long time, but I still saw the leg of a moon lander.)
Also, you correctly identified hydrant because "moon lander" fortunately was not one of the choices:)
If this spreads, it will be worth making a bot for, so that you can spam any of the sites that use it -- right now that bot just needs a few hits on average, and a slightly smarter bot can space them out, so that they are conceivably like commenters.
