Уникайте витоків CDP, замінюючи доріжки миші з нуля

[MiddleSchoolStudent]

https://chromedevtools.github.io/devtools-protocol/tot/Input/#method-dispatchMouseEvent

Всі основні антиботи визначають траєкторію руху миші, потрібно замінити це і інтегрувати деяку логіку в C++.

Необов'язково:
https://github.com/riflosnake/HumanCursor
https://github.com/Xetera/ghost-cursor

[Eth3rnit3]

Do you mean that the displacements don't have to be linear, otherwise they can be detected?

[MiddleSchoolStudent]

Do you mean that the displacements don't have to be linear, otherwise they can be detected?

Check this out: https://blog.castle.io/bot-or-not-can-you-spot-the-automated-mouse-movements/

[Eth3rnit3]

Do you mean that the displacements don't have to be linear, otherwise they can be detected?

Check this out: https://blog.castle.io/bot-or-not-can-you-spot-the-automated-mouse-movements/

Yes indeed, the article is interesting. So the bezier curve isn't enough, you need speed variations and possibly unnecessary scroll or hover movements on various elements before clicking on the target. As long as it doesn't look like a big mess, I think that drowning out the movements should make it possible to fool detection.

But the article is nearly 5 years old, so it's still not new in theory. If it's still not really applied, it must be difficult to differentiate.

[MiddleSchoolStudent]

Do you mean that the displacements don't have to be linear, otherwise they can be detected?

Check this out: https://blog.castle.io/bot-or-not-can-you-spot-the-automated-mouse-movements/

Yes indeed, the article is interesting. So the bezier curve isn't enough, you need speed variations and possibly unnecessary scroll or hover movements on various elements before clicking on the target. As long as it doesn't look like a big mess, I think that drowning out the movements should make it possible to fool detection.

But the article is nearly 5 years old, so it's still not new in theory. If it's still not really applied, it must be difficult to differentiate.

That true isn't very new, the major antibots have been collecting mouse movements to analyze trajectories for a long time ---- when reverse-engineering their source code and finding out that they're collecting these trajectories, it's probably enough to guess what they're sneaking things in.
Then again, a lot of Bots are still relying on puppeteer / playwright built-in functions to move the mouse, and these linear methods are definitely easy to detect.
So I'd like to implement this logic in botbrowser's c++ code, such that when Input.dispatchMouseEvent gets invocked, I'll refer to ghostcursor to add some intermediate points to avoid detection.

[Eth3rnit3]

Very well, then yes, that would probably be enough for now. I've set up a bezier curve applied during each move https://github.com/Eth3rnit3/chromate/blob/main/lib%2Fchromate%2Fhardwares%2Fmouse_controller.rb#L85 I'd be curious to see what the combination of the two would give. Probably even more credibility to the movement 🙂