Logic-based CAPTCHAs

Will AI bots force CAPTCHAs to become so clever that many humans can’t solve them?


It’s Not You. Those ‘I Am Not a Robot’ Tests Are Getting Harder.

Captchas that aim to distinguish humans from nefarious bots are demanding more brain power; ‘things are going to get even stranger’

By Katie Deighton, The Wall Street Journal, April 22, 2024

For years, people trying to shop online or log into social-media accounts might be pressed to complete bothersome but largely simple tasks—deciphering words in distorted type, clicking on pictures of buses, adding up numbers. Now those tasks are getting odder and require a few notches more brain power…

The companies and cybersecurity experts who design Captchas have been doing all they can to stay one step ahead of the bad actors figuring out how to crack them. A cottage industry of third-party Captcha-solving firms—essentially, humans hired to solve the puzzles all day—has emerged. More alarmingly, so has technology that can automatically solve the more rudimentary tests, such as identifying photos of motorcycles and reading distorted text. …

“Software has gotten really good at labeling photos,” said Kevin Gosschalk, the founder and CEO of Arkose Labs, which designs what it calls “fraud and abuse prevention solutions,” including Captchas. “So now enters a new era of Captcha—logic based.”

That shift explains why Captchas have started to both annoy and perplex. Users no longer have to simply identify things. They need to identify things and do something with that information—move a puzzle piece, rotate an object, find the specter of a number hidden in a roomscape. … [end quote]

On the one hand, I’m all for better security. On the other hand, I think that many people may find this frustrating.