Future of verification code: What about reCAPTCHA?

Future of verification code: What about reCAPTCHA?

The birth and significance of reCAPTCHA

CMU (Carnegie Mellon University) designed a powerful system called reCAPTCHA, allowing computers to seek help from humans. The specific method is to send unrecognized text scanning images of OCR (Optical Automatic Recognition) software to major websites around the world to replace the original verification code images; after the website users correctly recognize the text, their answers will be sent back to CMU.

ReCAPTCHA uses the CAPTCHA principle (CAPTCHA's Chinese name is a completely automated Turing test to distinguish computers from humans) and uses the human brain's ability to identify hard-to-recognize characters, this section describes how to identify characters that are hard to be recognized by OCR in old books. That is to say, reCAPTCHA can not only reverse spam (spam), but also help with the digitization of Ancient Books (also known as manual OCR ).

Each reCAPTCHA will display two words for recognition. One is a difficult word to be recognized by the user, and the other is a real CAPTCHA word known to the answer. The software regards users who can correctly recognize CAPTCHA words as humans. When CAPTCHA words are correctly identified, the program records users' answers to unreadable words and adds them to its database. This completes a manual OCR recognition. Process 1.

(Figure 1)

To improve the accuracy of the software, reCAPTCHA sends the most difficult words to multiple users and selects those with the same answer as the correct answer. It is said that the accuracy can reach 99%. Every time a user uses this program, it is actually helping to reproduce a page of the New York Times in 1908 or a page of other ancient books, which is of great significance to archaeology.

The following is an instance of a website that uses reCAPTCHA for registration verification (Figure 2 ):

 

 

Figure 2)

ReCAPTCHA acquired by Google

ReCAPTCHA was acquired by Google in 2009. After that, the reCAPTCHA Verification Code content of some websites in other countries has changed. The content displayed is not only ancient books, but also photos. The side of the photo shows the distorted text that everyone is familiar, the other side is a vague number, which is undoubtedly a street address. A Google spokesman said the system is not limited to street addresses, street names and even traffic signs are included. Example 3.

(Figure 3)

Google asked reCAPTCHA to display pictures of Google Street View. In this way, data such as street names and traffic signs are often extracted from street views, and useful information such as shop addresses and locations is added to Google Maps.

New reCAPTCHA-noCAPTCHA

 

Google published an article "Are you a robot? Introducing "No CAPTCHA reCAPTCHA"

The article began to talk about the traditional verification code method, which makes "Real Humans" a headache, and research shows that the current AI technology has been able to solve 99.8% of the verification code, therefore, distorted text verification methods may not be reliable.

The new reCAPTCHA is called a verification code without a verification code ("No CAPTCHA reCAPTCHA") by Google. It allows users to simply check and confirm that you are a real user rather than a malicious robot, the operation is very simple.

NoCAPTCHA only provides a check box that says "I am not a robot ". After you hook up, Google can use the risk analysis engine to perform a series of seamless checks to determine whether you are a real person.

(Figure 4)

If noCAPTCHA thinks you are a real person, you don't have to do anything anymore. This is really easy. If noCAPTCHA thinks you are not a real person, it will ask you to fill in a traditional CAPTCHA string or a more advanced string, such as selecting the correct image from a group of images.

 

 

(Figure 5)

Vinay Shet, product manager of noCAPTCHA, discussed the technology behind the improved version of the service. He said the check box sounds simple, but it is indeed complicated. The new risk analysis engine focuses on user engagement before, during, and after users click the check box to determine whether you are a real person.

Xue te said that the risk analysis engine uses machine learning technology, which uses a lot of strategies to know what normal users look like, so that they will know what to check next time.

Currently, not all users can use the new noCAPTCHA version. Only a small part of users can use it, developers who want to use most of the new noCAPTCHA experience must use new APIs to use it in their products. At present, Snapchat, WordPress, Humble Bundle and cloud9 are also testing the new system noCAPTCHA.

The following figure shows how to use noCAPTCHA when registering cloud9:

(Figure 6)

Summary

ReCAPTCHA is not only a verification code service, but also an important cultural project. After being acquired by Google, reCAPTCHA has more content. Currently, noCAPTCHA seems to free humans from the era of "Ten evil verification codes", pay more attention to user experience, make real human use more convenient, and make machine registration more difficult.