Answer the question
In order to leave comments, you need to log in
How do you rate the hack resistance of this captcha?
Each animation frame is a separate captcha.
(there are about 150 captchas, but many are similar)
The size of the captcha can be smaller, you need to select it.
I wonder how to break this? It is relatively easy to generate, and the reverse algorithm is not invented.
Answer the question
In order to leave comments, you need to log in
If you hack purposefully, then you can simply create a font with letters from captcha and stupidly look for the most similar letter.
Some will be fairly easy to break.
Head-on solution: cut into letters, take an array of white dots of each letter and compare with an array of white dots of the reference (pure white letter)
A lot of frames only make the task of the robot easier. He will take it apart frame by frame and recognize the letters in the most successful cases.
The idea is cool, but when the frames are combined, the letters become quite tangible. It will be possible to train the tesseract in this font. Protection is simple - do not put letters in a line and rotate them at different angles. Although, while maintaining the same font, it will be possible to do a complete enumeration of the picture and select the most similar option. In other words, also use a different font for each letter.
It is quite easy to write an analytical algorithm for such a captcha with a guessing of about 80%, i.e. captcha is ineffective. Reasons: the symbols are located at a uniform distance from each other, they are quite contrasting.
If you want to make a good captcha, then look at how KCaptcha is organized . Please note that there the characters are distorted and overlap each other. The coefficient of analytical guessing of such captcha is less than 1%.
If you wish, it is quite possible to hack this - only first you will need to collect a database of examples. But there is nothing fundamentally complicated about it. Although the added noise greatly complicates the task, the fixed position of the letters relative to each other and the absence of rotations make the task much easier.
Captcha of medium difficulty. Here is the one with only a part of the contours, probably more complicated, but the font is still the same and not distorted, so it’s not good. Look like Yandex and Google.
Here are the layers. Still play with offset and you can read. Very easy.
Why bother? If only for self-study :)
There is here www.google.com/recaptcha if you need to quickly deploy the solution in practice.
Are you sure you need a captcha? There are also many other ways. For example, an array of questions with answers, of which 1 is displayed on the page and a bunch of input forms, of which only one is visible to the user, the rest are behind the screen, under the layer or somewhere else. The bot must fill in a certain field with the correct answer and not touch the rest, which is already unrealistic, unless they are digging specifically for the resource. And if on purpose, then there is little that will help.
With such distortions, the rotation is no longer needed, IMHO. This is hard to parse.
But now, indeed, human readability has plummeted.
By the way, most of these points in the picture are simply removed by a simple filter (if there are > 6 points of the same color next to the point, assign the same value to this point).
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question