Bots and animals

Even though it is available since 2007, I haven't heard of Asirra until today, when I read someone's blogpost on spam.

Asirra is a database of pictures of cats and dogs. Each picture comes labeled with the type of animal it contains. You can't see the database directly; but you can access it through their webservice. The idea is to help fight bot spam by requiring the user to flag 12 images as either cats or dogs. I encourage you to test it out here.

I tried their demo and it is kind of cool. I didn't make any unintentional mistake in about ten tries and it is really quite a joy to look at the pets. Their academic paper is available here and it is a light read if you are interested in the details. In my opinion, Microsoft's researchers did quite a good job of analyzing the technical advantages and disadvantages of their scheme.

However, I believe Asirra will only remain safe if no high-profile website starts using it. Three million pictures is really not a lot, if you have the financial incentive to label them all. Furthermore, the database servers themselves could be compromised. Also, it is a poor business decision to rely on a single service that, at any time, may close shop or decide to charge a possibly absurd amount of money for their service.

All in one, I believe Asirra is a good alternative to CAPTCHAs for smaller sites or amateur projects and I hope the state of the art in the field advances well beyond the annoying text recognition systems we have today.