I mean, you can always anonymise data by removing all the useful data.
And that is the problem here: how do you remove enough data to make it NOT personally identifiable (or close too) AND not remove so much data that the whole thing is pretty useless.
No one has really managed that yet. People who have not tried assume it is possible. But it probably isn't except maybe in very specific cases where you only need very limited data and don't care about correlations with other factors...
There are some quite high profile examples of orgs releasing anonymised data and people linking it back to the individuals:
Interestingly the UK (I am a limey brit) actually has some really good experience with this, both from NHS medical records and public studies on Civil Servants...
And that is the problem here: how do you remove enough data to make it NOT personally identifiable (or close too) AND not remove so much data that the whole thing is pretty useless.
No one has really managed that yet. People who have not tried assume it is possible. But it probably isn't except maybe in very specific cases where you only need very limited data and don't care about correlations with other factors...
There are some quite high profile examples of orgs releasing anonymised data and people linking it back to the individuals:
https://www.theguardian.com/technology/2019/jul/23/anonymise...
Interestingly the UK (I am a limey brit) actually has some really good experience with this, both from NHS medical records and public studies on Civil Servants...
https://en.wikipedia.org/wiki/Whitehall_Study