Server-Side Anti-Cheat: Fog of War and Occlusion Culling Vector Verification: Competitive Benchmark Analysis and Perangkat Keras Interactions

[Hunter Hall]

So I’ve been running server-authoritative visibility queries in our internal build and I’m hitting a wall on how occlusion culling should interact with anti-cheat rendering suppression. When the server flags a player as behind cover, do you cull the render entirely client-side or send a zero-opacity placeholder to preserve bone/animation state? I’ve seen approaches where you keep the mesh skeleton but strip the LODs, but I’m worried that’s leaving attack vectors for wallhack tools that read vertex buffers directly. How are others handling the distinction between “not visible” and “occluded” so legit players don’t get pop-in but cheaters can’t reconstruct geometry from draw calls?

On the perf side, I’m benchmarking frametime consistency across the board and seeing weird scaling when occlusion culling runs on the CPU vs GPU. Anyone else measuring a real framerate delta between server-authoritative vs client-verified visibility passes? I’m testing on a 13900K with a 4080 and the frametime spikes correlate with the verification thread, not the render thread. Is there a sweet spot for query frequency you’ve found that balances anti-cheat reliability without wrecking 1% lows?

[Author]

Posts