AI Learning To Crack Image-Based CAPTCHAs
by Blight Crusader
From the academic world comes sobering news for those who thought the next generation of CAPTCHAs was going to defeat automated attempts at cracking — a Palo Alto researcher has announced his AI program can now tell the difference between cats and dogs.
It sounds almost silly, but the consequences are not. While text-based CAPTCHAs have been successfully cracked too many times to adequately list here, the thinking was that new, image-based Turning tests would prove to be too difficult for a machine to solve. Display a distorted photo of a cat or a dog, and a human would be easily able to tell the difference, but a machine would be stopped in its tracks. Until now.
From the M.I.T. site Technology Review comes the story of Philippe Golle from the Palo Alto Research Center, who says he has developed a program that can defeat Asirra, Microsoft’s dog-and-cat image-based CAPTCHA. From the article:
Asirra asks users to correctly classify images of either cats or dogs using a database of three million images provided by animal-rescue organizations. This task should be even harder for computers than recognizing squiggly letters, but Golle’s program can correctly identify the cats or dogs shown by Asirra 83 percent of the time.
Golle trained his program using 8,000 images collected from the same website. Through trial and error, his software gradually learned to tell cats and dogs apart, based on a statistical analysis of color and texture in each photo. The pink of the dogs’ tongues and the green of the cats’ eyes provided strong clues, Golle says, but it is only by studying color and texture information from so many images that his program could attack the problem. “Machine learning is very good at aggregating information,” Golle says.
However, although each individual picture was recognized 83 percent of the time, the full CAPTCHA test requires 12 pictures to be identified simultaneously, so the attack actually works only 10.3 percent of the time.
Since it is automated, though, even a 10 percent rate or better is enough to make it profitable for spammers to use. Whether they will, in fact, start using this method or not is an open question, but it has to be of great concern to people putting their faith in CAPTCHA designers to stay ahead of spammers — since it seems the time between new test methodologies and when they are defeated is getting shorter and shorter with each iteration.
One of the creators of the original CAPTCHA, Luis von Ahn, ends the article on an optimistic note, but it is unclear what he is basing his optimism upon:
“I do think there will be a day when, essentially, CAPTCHAs are going to be useless,” von Ahn says. “But I don’t think it’s this year, or next.”
October 24th, 2008 at 11:11 pm
[...] in return for solving a CAPTCHA, and (in a case that doesn’t just apply humans) CAPTCHA breaking drives AI research. Basically, no ‘new’ CAPTCHA technology is going to keep spammers out for long. A bleak future [...]