A common cheat in FPS games watches the rendered frame and sends mouse input to aim+fire. The server being authoritative doesn’t help in this situation.

These sort of cheats are often entirely out of band, running on a mirror of the video stream on a second machine, and transmitting user input back via USB. Clientside snooping won’t catch it either.

Yep.. "DMA" direct memory access.. reading the HDMI output, simulating controllers.. it all exists.

The google term is "hardware cheats"

That seems like it ought to be detectable statistically, unless the cheater adds jitter to the inputs to mimic human hands on a controller. And if not, it seems literally impossible to prevent so you may as well give up.

(As an aside, hearing about this kind of setup really makes me wonder what the point is of cheating. Prestige? Among what set of peers?)

> And if not, it seems literally impossible to prevent so you may as well give up.

But if you give up, and your game's competitors don't, players will switch games. The way I see it, a game developer anti-cheat goals isn't necessarily to have no cheaters, but definitely to have less cheaters than your competitor.

Determinism as a networking model is rare now a days(Still used though!) client/server is a better solution most of the time. And having determinism means the cheater could see everything. But having a deterministic sim means you must insert the cheating in the distributed to everyone inputs... So -if- you could detect cheated input reliably this could be useful.. This route probably isn't worth pursing for just cheating deterrence alone though.