GabeFollower discusses how effective Valve's AI anti-cheat: "Theoretically, VACnet is a great idea, and it has a future"
2024-09-30 20:46:00Dataminer Maksim "Gabe Follower" Poletayev talked about Valve's CS2 anti-cheat in an interview with Cyber.Sports.ru. He speculated that the progress is slow because it's hard to identify non-obvious cheaters in the game. However, he believes in future, as the developers have an advantage in terms of data.
VACnet was somewhat effective in CS:GO, but just like every AI or neural network, it had false positives. They are hard to avoid. It's tough to separate a cheater with sophisticated software from a regular player. VACnet only helps with blatant cheating. <...>
Valve is the only company that can put infinite resource in such a controversial project [VACnet]. Also, they have a lot of data to train [AI anti-cheats]. Theoretically, this is a great idea, and it has a future. The VACnet and Trust Factor patents state that Valve potentially knows about all user actions on the platform: about all the accounts connected to hardware and everything happening inside Steam. If you study all this information, in theory, you can make quite a reliable conclusions on a suspected cheater.
There will be AI cheats, and it will be a game of cat and mouse, but hypothetically, the side with the most data wins. Which is the developers of the anti-cheat and the game.
The mentioned "sophisticated" cheating software can imitate human behavior. An AI trying to analyze a suspect might not spot the difference in such a case. For example, users might add a delay to the reaction time of an aimbot or a triggerbot. The headshot accuracy can also vary this way. Wallhacks are nearly impossible to detect via AI observations.
In August, Valve announced tests of VACnet 3.0. The new anti-cheat can track suspicious behavior live and cancel games as soon as possible. The developers asked for feedback in false-positive cases. There has been no public update from Valve since then. Based on the evidence on the web, some games were indeed cancelled, but the real scale and success rates are a mystery.