Google has built an algorithm that can solve the firm’s CAPTCHAs with 99.8% success.
While developing technology to improve address accuracy in Street View, the panoramic pedestrian’s view in the company’s mapping application, the firm found it was even more effective at deciphering the anti-spam mechanism common to millions of websites.
Vinay Shet, product manager of Google’s reCAPTCHA, wrote on a company blog: "These findings have surprising implications for spam and abuse protection on the Internet as well.
"Thanks to this research, we know that relying on distorted text alone isn’t enough."
CAPTCHAs are anti-spam and security tools that commonly take the form of distorted images of text designed to be legible to humans but illegible to machines, with noisy audio recordings available for those with poor vision.
The advances have been detailed in a scientific paper submitted to the International Conference on Learning Representations (ICLR), a machine learning conference held in Canada this week.
In the paper Google claims that it can recognise 96% of Street View house numbers, and has collected 100 million of them through the proposed technology.
"It’s important to note that simply identifying the text in CAPTCHA puzzles correctly doesn’t mean that reCAPTCHA itself is broken or ineffective," Shet said. "On the contrary, these findings have helped us build additional safeguards against bad actors in reCAPTCHA."
Last October Google updated its CAPTCHA system, saying it now relied less on the actual recognition of text and more on the behaviour of the user. Humans using reCAPTCHA are presented with numbers, while bots are shown more complicated puzzles to solve.
