An experiment of auto-aim mechanic for all FPS games using neural network and AI humanoid posture recognition. Experiment and discussion of anti-cheat only. Contact me if you want to discuss.
First Note:
-
I personally hate hackers in games and will seriously fight against all kinds of cheats in games.
-
This is only an exploration of how AI could ruin modern FPS games and the methods to fight against it.
-
Therefore, only part of the code was released. Main.py is made private.
-
You will only get the basic functionalities such as mouse input, screen capture and AI posture estimation.
-
I'm not responsible for any consequences cause by modifying or rewriting the code.
-
All Experiments are done with bots. It was never used in official matches.
Insights:
-
As neural network and AI is evolving, the technology of face recognition becomes better over time, which also provides new methods to cheat in FPS games.
-
Different from traditional cheats which fetch the information from servers, this AI-based cheat does not use any of the internal game data.
-
Similar to how the player plays the game, it just looks at the computer screen and clicks on the detected head position, which makes it almost impossible to be detected.
-
Currently, it seems that encrypting or hiding the mouse data could be one way to fight against this cheat.
Results:
-
The results are terrifying. The aimbot can handle all kinds of humanoid bodies in FPS games such as CSGO, Back For Blood and Remnant: From The Ashes.
-
The accuracy is relatively high. Although sometimes it will miss a few shots, but this is inevitably a concerning problem if the model is well-trained for a specific game on the market.
