You learn some great things on HN!

2006: https://www.researchgate.net/publication/3455253_Digital_Cam...

Notes (by El Gamal of FPGA fame, not cryptography fame): http://isl.stanford.edu/~abbas/ee392b/lect07.pdf

Forensics: https://www.mobiledit.com/camera-ballistics

Anti-forensics (counter-measures/imitation): https://arxiv.org/abs/1707.07795v1 https://www.researchgate.net/publication/305181626_Removing_...

In that case just put on a few more layers. Shot noise is fairly random.

True. But if you share enough images from the same chip, eventually the fixed pattern noise can be computed to a decent enough accuracy to use it as a fingerprint. Now, you could measure the fixed pattern noise and subtract it from subsequent images. This is a widespread practice in scientific imaging. I'm not devious enough to think of how this could be defeated by someone who really wants to use your camera chip as a fingerprint, but it might be enough to defeat an amateur.

Dark-Signal Non Uniformity (DSNU):

https://harvestimaging.com/blog/?p=814

So likely to provide a few bits of entropy.