We've known about COTRAVELER[1] for several years.
The basic idea of correlating arbitrary data (with timestamps) to estimate associations can easily be extended to other types of data. This is why almost any kind of data can potentially be "personally identifying". On it's own a random piece of analytics or "anonymized" data doesn't reveal much, but when you correlate many of those data points together you get a very nice fingerprint.
And that's just a a few SQL JOINs. I'm sure a lot more is possible with modern analysis techniques.
We've known about COTRAVELER[1] for several years.
The basic idea of correlating arbitrary data (with timestamps) to estimate associations can easily be extended to other types of data. This is why almost any kind of data can potentially be "personally identifying". On it's own a random piece of analytics or "anonymized" data doesn't reveal much, but when you correlate many of those data points together you get a very nice fingerprint.
And that's just a a few SQL JOINs. I'm sure a lot more is possible with modern analysis techniques.
[1] https://www.washingtonpost.com/apps/g/page/world/how-the-nsa...