Get the latest tech news How to check Is Temu legit? How to delete trackers
Personal computers

Are you smarter than a robot? Study finds bots better than us at passing CAPTCHA tests

Those tedious 'Are you a robot?' image verification tests are crucial to keeping bots off of websites. Or are they?

Eric Lagatta
USA TODAY
A visitor touches robotic fingers Thursday during the annual World Robot Conference at the Etrong International Exhibition and Convention Center on the outskirts of Beijing. A new study shows robots may be better than humans at solving those CAPTCHA tests intended to keep bots at bay.

We've all been there: You click on a website and are immediately directed to respond to a series of puzzles requiring you to identify images of buses, bicycles and traffic lights before you can go any further.

For more than two decades, these so-called CAPTCHA tests have been deployed as a security mechanism, faithfully guarding the gates to many websites. The acronym − it stands for Completely Automated Public Turing test to tell Computers and Humans Apart − started out as a distorted series of letters and numbers that users had to transcribe to prove they were human.

But throughout the years, evolving techniques to bypass the tests have required that CAPTCHAs themselves become more sophisticated to keep out potentially harmful bots that could scrape website content, create accounts and post fake comments or reviews.

First day of school:Think twice about that first-day-of-school photo: Tips for keeping kids safe online this school year

A CAPTCHA based on a classic Computer Vision problem of image labeling.

Now perhaps more common are those pesky image verification puzzles. You know, the ones that prompt you to click on all the images that include things like bridges and trucks?

It's a tedious process, but one crucial for websites to keep out bots and the hackers who want to bypass those protections. Or is it?

Study finds bots more adept than humans at solving CAPTCHA

AI, Machine learning, Hands of robot and human touching on big data network connection background, Science and artificial intelligence technology, innovation and futuristic.

A recent study found that not only are bots more accurate than humans in solving those infamous CAPTCHA tests designed to keep them out of websites, but they're faster, too. The findings call into question whether CAPTCHA security measures are even worth the frustration they cause website users forced to crack the puzzles every day.

Researchers at the University of California, Irvine recruited 1,400 people to take 10 CAPTCHA tests each on websites that use the puzzles, which they said account for 120 of the world’s 200 most popular websites.

The subjects were tested on how quickly and accurately they could solve various forms of the tests, such as image recognition, puzzle sliders and distorted text. Researchers then compared their successes with those of a number of bots coded with the purpose of beating CAPTCHA tests.

The study was published last month on arxiv, a free distribution service and repository of scholarly articles owned by Cornell University that have not yet been peer-reviewed.

"Automated bots pose a significant challenge for, and danger to, many website operators and providers," the researchers wrote in the paper. "Given this long-standing and still-ongoing arms race, it is critical to investigate how long it takes legitimate users to solve modern CAPTCHAs, and how they are perceived by those users."

Findings: Bots solved tests nearly every time

Google says, “reCAPTCHA is a free service that protects your site from spam and abuse.”

According to the study's findings, researchers found bots solved distorted-text CAPTCHA tests correctly just barely shy of 100% of the time. For comparison, we lowly humans achieved 50% to 84% accuracy.

Moreover, humans required up to 15 seconds to solve the challenges, while our robot overlords decoded the problems in less than a second.

The only exception was for Google's image-based reCAPTCHA, where the average 18 seconds it took humans to bypass the test was just slightly longer than the bots’ time of 17.5 seconds. But bots could still solve them with 85% accuracy.

The conclusions, according to researchers, reflect the advances in computer vision and machine learning among artificial intelligence, as well as the proliferation of "sweatshop-like operations where humans are paid to solve CAPTCHA," they wrote.

iPhone settlement:Apple agrees to pay up to $500 million in settlement over slowed-down iPhones: What to know

Because CAPTCHA tests appear to be falling short of their goal of repelling bots, researchers are calling for innovative approaches to protect websites.

"We do know for sure that they are very much unloved. We didn't have to do a study to come to that conclusion," team lead Gene Tsudik of the University of California, Irvine told New Scientist. "But people don't know whether that effort, that colossal global effort that is invested into solving CAPTCHAs every day, every year, every month, whether that effort is actually worthwhile."

Eric Lagatta covers breaking and trending news for USA TODAY. Reach him at elagatta@gannett.com.

Featured Weekly Ad