Stanford Boffins on the Brink of Breaking Captcha Codes

Illustration for article titled Stanford Boffins on the Brink of Breaking Captcha Codes

Captcha systems, those psychedelic-font phrases designed to weed out bots from users, are a staple website security. And, thanks to Stanford Researchers, they may be quickly becoming completely useless.

Advertisement

The researchers employed machine vision algorithms to successfully crack 66 percent of Visa's Captchas, 70 percent of Blizzard's, and 25 percent of Wikipedia's. For reference, a one percent successful cracking rate is regarded as grounds for the Captcha's immediate discontinuation. Only one company's system, Google's ReCaptcha, was able to confound the researchers' robotic eyes—presumably because the vision algorithms still aren't advanced enough to handle ReCaptcha's blurring effects but how long will that advantage last?

So, since it's only a matter of time before A) researchers figure out how to break all of these systems, including Google's, consistently and B) the technology hits the Internet and we're inundated with bot advertising, does anybody have ideas for the next generation of bot filtration technology? Is the time for widespread biometric-scanner use finally upon us? [Maximum PC]

Advertisement

You can keep up with Andrew Tarantola, the author of this post, on Twitter, Facebook, or Google+.

Share This Story

Get our newsletter

DISCUSSION

We're getting pretty close the point where humans can't read these things, so if computers catch up again, there's really nowhere else for them to go. But just because character-based captchas are dying, doesn't mean we have to abandon the idea entirely. As we get better software for cracking captchas, we should simultaneously be getting better character recognition for other OCR purposes too. Perhaps we should make newer captchas audio-based, and we could get better speech recognition—or we could identify objects in videos or any number of other tasks that computers currently suck at and have software improve in those areas.

Each time computers reach human parity in one area, we can move on to another—with the world a little better behind us.