I imagine this approach would be insecure because an attacker could easily change the distribution from miles away. And I don't think it solves the problem of r

I imagine this approach would be insecure because an attacker could easily change the distribution from miles away. And I don't think it solves the problem of radio noise being time-dependent random variables.