Sometimes user can partially see them, game client would need that position. Then user can make a mod that flashes silhouette that just appeared behind a wall for a moment.

You can do it to a degree (basic room detection), but it'll never be 100% accurate because of latency and compute cost, you have to give leeway.

not a trivial solution.

you run into some really difficult to solve issues when it comes to the gameplay loop, the graphics engine, and network latency trying to solve that unless you are playing some sort of turn based game where all data can be resolved before the next action