'CAPTCHA' which is made to understand the character reading and image quizzes is the fate that difficulty will rise as the bot evolves

by Duncan Rawlinson - Duncan.co - @ thelastminute

In order to prove that a guest user is a human rather than a bot when logging in, it may be required to clear the test called CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) . If you make a mistake it will be over and over again, and there are many people who have been irritated. The problem of such CAPTCHA is "The clear difficulty gradually becomes high as the bot evolves," The Verge argues.

Why CAPTCHAs have gotten so difficult - The Verge
https://www.theverge.com/2019/2/1/18205610/google-captcha-ai-robot-human-difficult-artificial-intelligence

CAPTCHA was developed by AltaVista in 1997 to eliminate bots trying to add spam URLs to search engines. It is said that introduction of CAPTCHA was able to prevent more than 90% of URL addition by bot. The results were evaluated and CAPTCHA was quickly adopted for various services as an essential part of the authentication system.

In order to break through CAPTCHA, it is necessary to look at the image of "distorted character" and decode it and input it in the specified form. Using this mechanism, we display " reCAPTCHA " which characters that could not be read by image recognition software are displayed as "subjects" and used for data collection to improve human image recognition software instead of being solved by human beings instead of Carnegie Melon University developed in 2007. This tried successfully, and in 2009 Google acquired this technology.

by Alan Levine

However, along with the penetration of CAPTCHA, the bot evolved, it became possible to read the image of the distorted character with considerable precision, and the original function of CAPTCHA "to eliminate the bot" was not being fulfilled. For example, in 2012, a script that breaks reCAPTCHA by over 99% accuracy has been announced.

A script that automatically breaks over Google's 'reCAPTCHA' used to avoid spam has appeared - GIGAZINE

Therefore, Google released a new reCAPTCHA API called " No CAPTCHA reCAPTCHA " in 2014.

No CAPTCHA reCAPTCHA requires no character input, the user can simply check the checkbox "I am not a robot" OK. Then, No CAPTCHA reCAPTCHA judges whether it is a human being or a bot from the behavior of the user and the movement of the mouse when clicking. Even if it is judged to be a bot despite being a human being, a problem of selecting a specific image is issued, and if it is cleared it will be authenticated as a human.

According to Google's lecture in 2016, the percentage of users who answered image selection quizzes was 99.8% for bots and 33% for humans. Although about one in three people have solved image selection quizzes, No CAPTCHA reCAPTCHA has a high degree of accuracy in terms of "to identify and properly eliminate bots from among users" .

However, Google's efforts are also vain and the accuracy of the bot is further rising. Jason Polakis, Professor of Computer Science at the University of Illinois, in order to clear Google's CAPTCHA with an accuracy of 70%, we have published a paper (PDF file) paper (PDF file) of experiments using commercially available image recognition tools including Google's image search It announces in the year. In addition, it is also verified to convert the image into sound using Google 's Speech - to - Text API to clear CAPTCHA.

Researchers announced that they can break through reCAPTCHA using Google's speech recognition - GIGAZINE

Polakis says, "CAPTCHA requires not only physical abilities but also issues beyond culture and languages - it is a simple matter for average human beings and at the same time a difficult problem for computers What people can do is quite limited, it should be something that human beings can quickly solve and still not irritate. "

Aaron Malenfant, the engineering leader of Google's CAPTCHA team, said, "Withdrawing from the Turing test means avoiding the competition that humans keep losing, as more investments will be poured into machine learning , The task of CAPTCHA will become increasingly more difficult for human beings. "After 5 to 10 years, like reCAPTCHA v3 which appeared in 2018, he said that it would not be necessary to check it" It is no longer a robot " I'm predicting. However, as reCAPTCHA evolves, monitoring of user behavior by Google's CAPTCHA will be strengthened, so it is subtle whether users are delighted with joy.

Recently, "CAPTCHA" has been developed, "Explain the problem abstractly by using symbols rather than sentences, rotate the figure to a specific angle to the user, solve the puzzle using the figure". This is based on the fact that the program requires concrete and clear instructions and issues such problems that can not be solved unless it is a human who can read and judge the context and situation. However, of course, when a human solves it requires some troublesome operation and thought time. In addition, CAPTCHA technology that forces users to prove that they have "human body" not using BOT with augmented reality (AR) has also been announced, clearing CAPTCHA not only for BOT but also for humans You can easily imagine the future that will become difficult with.

by Linking Paths

Meanwhile, Amazon devised "Turing test via failure" as a new CAPTCHA system. In 2017 we obtained a patent . In the summary of the patent, "As artificial intelligence improves, bots can clear CAPTCHA like human by using optical character recognition (OCR) , so that human users mistakenly answered the problem It is said that it is a CAPTCHA technology that gives challenges that are difficult for human beings to make use of optical illusions and logic puzzles.

The Verge stated that "CAPTCHA will continue to exist in this world only after all" in the end only "It is only human beings to make mistakes", after all CAPTCHA is said to be only the magazines of humans and bots.