Can Geetest verification be cracked?

My small station has been using the pole test for about a year now, and I have done a lot of homework on the pole verification, let me first talk about my understanding of the pole verification function mode (the following is limited to my professional knowledge, if there is something inappropriate, do not spray): it mainly uses behavior as the criterion for judging the terminal interaction object, but the whole relies on a small part of the ** recognition verification technology. So if you want to crack it, there are three technical thresholds, the first is image recognition, the second is the behavior simulation of dragging the slider, and the last is data encoding and decoding.

So the main steps to crack the Geetest verification are as follows:

2. Recognize the image and get the location of the answer.

3. Simulate a dragging trajectory according to the position of the answer.

4. Encode the behavior and answer position data and submit it for validation.

5. Get the result returned by the Geetest server and submit it to the client server.

6 Complete a verification and produce a desired action or a vandalism on the client server.

There is only hope of cracking if you complete the above steps, and each of these steps is quite difficult, making cracking software difficult to write (at least not now, and in the future). As a**Lord, it's hard to find a verification that takes into account both anti-swiping and user experience.,I used to use a more difficult-to-recognize**CAPTCHA.,Every time I'm swiped, I continue to change the verification that is more difficult to identify.,In fact, it's a vicious circle.。 Later, the pole verification of the integration on discuz, the main reason for integrating it at that time was because its painting style was more harmonious with my small and fresh station.

Those who have built their own stations should know the consequences of being crawled by reptiles, and Nima leaves advertisements everywhere. Here I worship the great gods who do reptiles for three minutes, and ask for this technology to do something serious)...Later, I made up a lot of homework on its functional mode, and I basically tried to crack software with an academic attitude, and tried to attack my own station with the cracked version circulating on the Internet, and finally it was forbidden.

In the security industry, there will always be a mutual technical and technological competition between hackers and white customers. There is no forever impermeable wall in the world, hackers continue to progress, so white customers must continue to upgrade the effect of defense means and increase defense solutions to improve their security technology level, from the perspective of a ** master, I still hope that the future verification can be better and better, and constantly carry out dynamic updates of characteristics to protect our small station.

Let's briefly talk about my thoughts, mouse sliding data can be obtained through some trajectory recording software, the collection of data is my own row, if successful, record this set of data, and then sort out the data, get the time and displacement data of the mouse, just find a few sets of data to make a line chart, you can find the general change in movement speed. If you have a machine learning foundation, you can regress linearly according to these data, and if you don't have a machine learning foundation, you can also pass the collision estimation based on the trajectory. That's probably the idea.

success rate. At the beginning, it was very high, about nine times out of ten times it could pass, and now it may be possible to pass five or six times out of ten times after upgrading the algorithm, but I find that there are many times when I slide it manually.