The answers and broader philosophy are all over this comments section so I'm not sure there's much point to my answering you but why not.

The tl;dr is that you largely don't. It's a fools errand.

First you need to recognize the general behavioral pattern and motivations behind it. Once you do that you'll realize that the same people snooping on RAM right now are going to turn to ML botting if they can't do that. Those bots are usually already superhuman, are only going to get better, and the hardware to run them is quite cheap (a video camera and a fake keyboard and mouse).

It's the same problem online chess has. Snooping RAM doesn't help you in online chess but that doesn't mean there aren't cheaters.

Remember catbot? That certainly managed to ruin people's day.

A workable solution has to be end-to-end. Identification based on behavior, competent moderation to review those cases, a model that ensures the moderation efforts are sustainable in the long term, and some way to make sure that bans are sufficiently sticky so that there's actually a point to the whole thing.

Score based methods like Valve employs are a reasonable alternate approach. If cheaters all get thrown in the same pool then who cares if they cheat? Let them have their fun!

Community servers instead of centralized matchmaking are another option. Those once again group the cheaters together, shift moderation costs away from the publisher, and give you a stronger community.