Google has discovered a method to solve CAPTCHA puzzles with high levels of accuracy, a recent paper by the company has revealed. When specifically tackling reCAPTCHAs, a popular implementation owned by Google, the algorithm boasts accuracy of 99.8%, posing questions about the usefulness of the puzzles.
CAPTCHA, which stands for Completely Automated Public Turing test to tell Computers and Humans Apart, is a test most often used on forms to ensure submissions are made only by humans. If Google's sophisticated algorithm can make a bot act like a human in these tests, it's only a matter of time before we'll need better CAPTCHA implementations.
Google's algorithm is good enough to solve these tricky reCAPTCHAs
The algorithm used to solve CAPTCHAs was primarily developed to read signs and house numbers in Street View imagery, allowing each image to be mapped to a location with greater accuracy. Surprisingly, the algorithm is more effective at solving CAPTCHAs than its primary task: when reading whole street numbers, it achieves 96% accuracy; it's 97.84% accurate at discerning individual numbers; and just 90% accurate at reading text.
Luckily, Google claims that they've found new ways to keep CAPTCHAs from being solved by bots. Vinay Shet, product manager for reCAPTCHA, says that "relying on distorted text alone isn’t enough", with the company instead looking to analyze the interaction between the user and the puzzle to determine what is a human and what is a bot.