What stops the robot from clicking the I’m not a robot button?
In: Technology
The thing is, it can.
The other thing is, Google engineers are smart.
Have you ever noticed that the I am not a robot button is almost never on the screen initially? You actually have to scroll down to see it. A robot wouldn’t have to scroll down.
Also, a human had to manually move a mouse across the screen in an imperfect manner which a robot may or may not do.
Now you’re likely signed in on your Google account. Well Google has a pretty good profile on your browsing habits to check against.
All of these things go into deciding whether a human or a robot clicked that button. And if these things Fail,b well then you now have to identify pictures of trees etc etc.
Programming a robot to type things, such as trying a bunch of passwords, followed by the Enter key, over and over again, is really easy.
Programming a robot to perform image recognition on a display, see where the button is, recognize it for what it is, and provide a simulated mouse-click or touch input to that location is hard. Some robots can do it, but it’s a lot harder to make that robot.
Those really annoying robot checks where you have to squint to see what letters the malformed colored squiggles inside of the rectangle are supposed to be and type them, or click every picture on the grid that contains crosswalks or stop signs, are nearly impossible to program a robot to do without advanced image recognition and machine learning to the point where it’s not worth making the robot.
These things are called CAPTCHA which stands for Completely Automated Public Turing test to tell Computers and Humans Apart (Computer Scientists aren’t good with the concept of Acronyms).
The one’s with a check mark actually work by sending some data to Google (who owns reCaptcha), stuff like your mouse movement, your connection details and your cookie history. Google, understandably, doesn’t actually publish exactly what data is sent. They then use that to determine if you’re likely to be a bot or a human ( for example a bit may move the mouse almost instantly in perfectly straight lines, whilst a real humans mouse movement would be slow and jerky). If they then determine that you’re possibly/likely a bot you’ll get one of the secondary checks, such as selecting all pictures with stop signs. This is something that’s relatively easy to do for humans, but almost impossible for a conventional computer programme (it’s doable, and getting easier, with advances in machine learning, but still takes a lot of effort and time to get a program to any level of reliability).
Nothing. Bots can easily click it it. The thing is, while it looks like a simple button, it’s actually tracking a number of things about you from the moment you load the page. Your mouse movements, your typing, and even things like your browser version and your IP address. It tallies up all of those pieces of data and uses that to decide if you’re a bot or not.
If it thinks you look human enough, it lets you through without any fuss. If, however, you look too bot-like it will make you answer those image challenges where you have to click the pictures of cars or stoplights or whatever. (All of which are extremely difficult for AI’s to reliably answer, precisely because Google uses images that its own experimental AIs have trouble with)
As a practical example, the biggest downside of using a VPN to protect my privacy is that I *always* get those image challenges. Since my VPN gets used by a lot of people (including plenty of bots, no doubt), the system always flags me as “might be a bot”.
Latest Answers