Driving your verification code crazy, why is it getting more and more complicated

  On this day, you have been preparing for the postgraduate entrance examination for a long time, and you opened the postgraduate entrance examination registration website with pride.
  However, before you have time to show your talents, you are stuck on the verification code for website login. The overlapping fonts and the mosaic background make you lose the verification code several times without losing it correctly.
  You have been defeated and fought again and again, but the “chattering” that appears on the verification code instantly makes you doubt your level of knowledge.
  Once again, you motivate yourself and decide that you must lose right this time, but you see two words written on the refreshed verification code: Don’t take the test.
  Driving your captcha crazy, why is it getting more and more complicated?
Tormented captcha

  Is this an o or a 0? Is it 1 or l?
  Every time you enter a verification code, you can’t help but ask this question. Its twisted shapes, glued-together characters, and snowflake-like background often make you wonder if you’re blind.
  However, this kind of character verification code is only a daily basic question, a difficult question, which often appears in special moments like the Spring Festival.
  The image verification code to choose when logging in to 12306 was once the nightmare of everyone who grabbed a train ticket during the Spring Festival.
  You have to pick out all kinds of objects that are “half-covered” in a pile of high-definition pictures. Sometimes you are unlucky. When you encounter things beyond common sense, such as exhaust fans and generators, you can only be trapped in the wrong choice again and again, and watch the tickets become less and less.
  Netizens also ridiculed that to do such a picture verification code question correctly is like calling out each member of the Korean group IZ*ONE one by one.
  Some verification codes are not enough to only test your eyesight, but also test your IQ. When grabbing concert tickets on a ticketing website, you must not only understand some literary allusions, but also read four famous novels. At the end, there is also a math problem for you to calculate.
  Every verification code on the screen seems to remind you: If you don’t have any culture, don’t learn to chase stars with others.
  Even some verification codes have become a testing ground for friendship between you and your friends. When you log in to facebook for verification, the system will randomly pick out a few photos posted by your friends, allowing you to identify who the person in the photo is. You, who are already face-blind, encounter the p-picture technology that your friend has practiced for many years, and the difficulty factor of the verification code will double as a result.
  Research from Stanford University in the United States pointed out that verification codes are already too difficult.
  The researchers collected 8,500 different forms of verification codes for the subjects to identify, and found that it took them an average of 9.8 seconds to judge the verification code, and at the same time let three people identify the same verification code. 71%. If you listen to the verification code of voice recognition, the accuracy rate will be lower, only 31.2%.
  The captcha rolls in like crazy, and you just want to smash the mouse.
The evolution of the verification code is also forced to help

  In fact, the verification code is not so complicated from the beginning.
  Before CAPTCHA was invented, many free email websites were often attacked by malicious plug-in software. For example, in 2000, there were a large number of vest accounts registered with machines in Yahoo Mail, and they would send a lot of spam and fraudulent emails to users.
  To solve this problem, program developers try to find a way to log in and verify whether it is a human or a machine that is operating.
  The engineers at Alta Vista found an efficient way. They found that at the time, even computers equipped with the most advanced optical character recognition systems had difficulty correctly recognizing words printed on paper, especially when the characters were stretched and distorted.
  However, this is easy for humans. We have been trained to read since childhood, even if the words on the paper are crooked and partially occluded, it does not affect our interpretation of the whole. The ghosts on your workbook can be carefully identified by teachers; the medical records written by doctors can still be deciphered.
  Twisted characters have become the most commonly used login verification method, and have the official name of “Captcha”. Programmers will pre-set a character library, the characters in the library will be arranged and combined into strings, and then distorted, smeared, and positioned on a random background according to random variables, thus becoming the verification code we have seen.
  These character puzzles are indeed effective in reducing spam on the website, and by 2001, spam in the “ADD-URL” mailbox was reduced by 95%.
  However, with the continuous development of computer optical character recognition technology, this form of verification code has become more and more difficult to keep the plug-in machine out of the door. In tests conducted by Google engineers in 2014, the computer recognized even the most difficult distorted character type captcha with 99.8 percent accuracy.
  In order to block more and more smart machines, the verification code is forced to evolve, and there are graphic verification codes, logical verification codes, voice verification codes, and even the ultimate form of mixing several.
  However, every time you suffer, it is not in vain. The brain cells you reduce in ten seconds are actually transferred to an AI.
  Some of the CAPTCHA characters we recognize come from old newspapers and classics. These handwritten and stained handwriting are difficult to be recognized by the computer. Programmers add these characters to the CAPTCHA.
  Such a captcha usually consists of two segments, the first segment is a word that has been successfully digitized, and the other segment is a scripture character for which the correct answer is not known. As long as the user answers the first paragraph correctly, the interpretation results of the second paragraph will also be recorded, and then the identification results of multiple users will be compared to draw a conclusion.
  Under the interpretation of people all over the world, more than 200 million words are deciphered and transcribed every day. The New York Times has achieved digitization in this way. The New York Times you see a hundred years ago may be the result of generating electricity with love.
  Not only character verification codes, but Google allows users to identify “zebra crossings” and “traffic lights” in the verification code pictures. It is also strengthening the robot’s image recognition ability and applying machine learning results to autonomous driving.
  We are caught in such a vicious cycle: the more verification codes are entered, the stronger the AI’s learning ability, the advancement of technology allows more verification codes to be cracked, and we are forced to design more complex verification codes again…
In the future, CAPTCHA may “disappear”

  The verification code seems to have deviated from its original intention, and it has become a display that stuns humans but cannot help machines.
  More and more technology companies have also realized this problem and began to try to simplify the complicated verification code.
  For example, when you log in to 12306 now, you don’t need to find Wang Luodan in Bai Baihe, but just drag the slider.
  Such verification codes are not only judging the results of your operations, but also recording and analyzing the behavior of your operations. The computer can recognize you’re human by watching how fast you move the slider.
  Your twitchy mouse, the slowness of aligning the slider, and maybe even a few repetitions of wiggling it from side to side to align—all human clumsiness.
  Another implicit verification code is also based on this principle. When you check “I’m not a robot,” the program can tell if you’re lying.
  It’s not that it really has mind-reading skills, but after you check it, it will automatically analyze your behavior when browsing the web page, including the mouse movement track, browsing speed, etc., to analyze whether these behaviors belong to normal human operation.
  So far, in this race between humans and machines, we have stopped trying to prove ourselves smarter, but have turned to another strategy: that inconsistent behavioral trajectory, inaccurate answers, and clumsy operations are what make us live. And human characteristics.
  However, such an implicit verification code is not a panacea. Shortly after it appeared, academics at the University of Toronto in Canada published a paper claiming that their robot could imitate human behavior and successfully pass the implicit captcha test.
  In 1950, the famous computer scientist Turing proposed a thought experiment: a human asks a series of questions, the machine gives the answer, and then the human judges whether the other party is a machine.
  A machine is considered to have human intelligence if more than 30% of people misjudge it as a human.
  But over the past 70 years, humans have fallen into the exact opposite situation – one captcha, the exact question the machine throws to humans: “Are you human?”
  This time, the judgment is in the hands of the machine, and the human being stumped by the verification code is still shouting hard: I am not a robot.
  Shen Bo excerpted from the WeChat public account “The answer is as follows”

error: Content is protected !!