The more I'm reading this article the less I understand their point. Is there an actual paper that describes how their "random data" is generated?

I can also generate random data that looks like any distribution by carefully choosing the random distribution. What's their point?