Developed over 20 years ago to prevent hackers from stealing content, inserting malicious posts, engaging in fraudulent transactions, or slowing down websites to the point where they become completely inoperable, the acronym for this ubiquitous line of defense makes its mission clear: Define it clearly. Fully Automated Public Turing Test Computers and Humans Are Apart – CAPTCHA.
For almost 20 years, CAPTCHA has been widely used as a protection against bots. As their use spread over time, techniques to bypass or spoof CAPTCHA continued to improve. However, CAPTCHAs have also advanced in complexity and variety, making them increasingly difficult for both bots (machines) and humans to solve. Given this long-term and still-relevant technology race, it is very important to examine how much time legitimate users need to solve modern CAPTCHAs and how these CAPTCHAs are perceived by these users.
Today, CAPTCHA still remains one of the main concerns of users.
But researchers at the University of California, Irvine have concluded that robots seem to be better at solving problems than humans.
Scientists study CAPTCHA under natural conditions to evaluate users’ problem-solving performance and perception of currently used CAPTCHA. They collect this data by manually checking popular websites and user studies where 1,400 participants collectively solved 14,000 CAPTCHAs. The results show significant differences between the most widely used CAPTCHA types. Surprisingly, the time it takes to solve does not always correlate with user perception. A comparative study was conducted to investigate the impact of the experimental context, particularly the difference between solving CAPTCHA directly and solving it as part of a more natural task, such as account creation. Despite several potentially confounding factors, the results indicate that the experimental context can influence this task and should be considered in future CAPTCHA studies. Researchers also investigate users’ refusal to complete tasks due to CAPTCHA and analyze participants who start tasks and do not complete them.
They found that the bot was not only better at solving various forms of CAPTCHAs, including image recognition, slider puzzles, and distorted text, but also faster.
As CAPTCHAs evolve in complexity and variety, they become increasingly difficult for both bots (machines) and humans to solve. However, advances in computer vision and machine learning have greatly improved the ability of bots to recognize distorted text, reaching accuracies of over 99%. Bots can successfully overcome CAPTCHAs containing garbled text in almost 100% of cases. Human accuracy when solving CAPTCHA varies from 50% to 84%. Moreover, while humans take up to 15 seconds to solve these tasks, bots can complete them in less than a second.
Based on this study, researchers reached the following clear conclusions: This means that there is no longer a simple way to reliably distinguish between humans and bots based on small images or other features. Instead, we recommend leveraging advances in artificial intelligence to develop ‘intelligent algorithms’ that can more effectively distinguish between bot behavior and human behavior.
You can read the full research paper at the following link: