Would it be possible to spread so much noise that data like this becomes useless? Could an LLM be used to help here?