The only effective approach is to use as many layers as possible to increase the cost of creating and using cheats. Kernel anti-cheat is an effective layer because it forces cheaters to either buy specialized hardware or gamble that their hypervisor won’t be detected through heuristics.

Competitive games will likely add AI-based flagging into the mix, but it still doesn’t make sense to make cheating as trivial as adding a few uprobes/kprobes on a Linux box.