I'm not really sure if that's a scary facebook thing or merely just an inconvenient reality of the world we live in now. There's quite a few papers out on fingerprinting/identifying cameras from images taken with them using intrinsics (dust, scratches, slight offset of image sensor with respect to lens).
The more ways we have to implicitly correlate/extract data using these techniques, the more accurate model of the world we can build.
It reminds me of this pretty cool research about extracting audio from silent videos:
http://news.mit.edu/2014/algorithm-recovers-speech-from-vibr...