Ross Anderson's Security Engineering Chapter 9 is all about this https://www.cl.cam.ac.uk/~rja14/Papers/SEv2-c09.pdf (Anderson releases the textbook free online)
This is a fairly well developed field (i.e., in medicine there highly screened "Medical Statisticians" who give data to research groups only after it's properly anonymized. They know beforehand what the researchers are interested in, and introduce noise that effects all the measures _except_ the ones the researchers said they were interested in looking at beforehand.
> The idea behind the differential privacy strategy, as I understand it, is that you can basically introduce statistical noise into the block-level results in a way that will make it impossible to reverse-engineer the demographic profile of people who live at specific addresses.
>This has, I think, roughly zero real-world privacy value
This was painful to read...