Is differential privacy just a fucking meme? If you add too much noise to make a database properly secure, a researcher is then working with inaccurate or bad data. If you add too little noise so that the database is still giving useful insight to the researchers, you get data leaks. There doesn't seem like there is a good middle ground whatsoever mathematically. There also doesn't seem to be any useful information measurement for the value of specific database identifiers.