Well, statistics would be the only mechanism. If a player was on average playing at level X in one setting, but at a lower level Y in a setting where it was considered impossible to cheat, that's about as good as you can do.

But it's pretty impossible to point to a single move and say "that's definitely a cheat move".

You can look at moves as a series of probabilities. For each move, classify if it's more a blunder or inspired move and then look at people's games and see if they consistently have 1-2 moves that are much much better than their typical.