Do you know that most of these ppl don't use the aimbot for educational reasons but for ruining the game experience to legit players?

[scerelli]

or also to earn money by boosting.

[PlutoNameless]

We empathize deeply with your concerns, which is why we have decided to abandon the development of this particular aspect of the research. It currently serves as a mere exploration in this field, with its practical value far outweighed by its research value. If you have any suggestions on how to limit the scope of safe usage while maintaining the research-oriented nature of the project, we would be open to discussing them here.

[scerelli]

I have to spend sometime thinking about it, what I would do, for sure, is research in the opposite way. So train something that can predict if someone is using an aimbot or not.

[PlutoNameless]

I have to spend sometime thinking about it, what I would do, for sure, is research in the opposite way. So train something that can predict if someone is using an aimbot or not.

If you are a game developer, particularly of independent games, consider the following issue we have previously thought about regarding identification:

Current projects or similar ones often use Python's MSS library or the Microsoft DXGI interface for acquiring images. They continuously capture screenshots and feed them to the YOLO model for target recognition. Although I haven't delved deeply into this, it should be possible to use a Hook to detect if a process has been performing that operation.

Another approach to consider is whether the program is providing mouse input through Ghub or an external hardware simulation of the mouse. The input characteristics during the enabled period will be significantly different from the user's input characteristics during the non-enabled period. A new model can be built to compare these two types of input. This method can potentially address the issue of mouse macros as well.

Lastly, you may want to utilize statistics. If you can access and read the player's game data, you can compare the player's multi-dimensional data within their segment to that of other players with the same rank. However, this method may have a certain misclassification rate and may only be able to label the player as suspicious. Nevertheless, you can use this approach to adjust the player's rank, similar to how the recent DOTA2 update introduced the confidence level for a player's rank. The server can then re-assign players in this category to compete together.

[scerelli]

Thanks for all the detailed info. This guys seems to have developed a sort of behavioural real time analysis to understand if it's a human aiming or not.

anybrain.gg

but of course it's not open source.