In the security newsletter #5, Frédéric Lefebvre presented the research works of Jessica Fridrich. Through analyzing the noise of pictures, she attempts to uniquely fingerprints a camera. Each CCD generates a unique template of noise. Thus, it should be possible to detect if pictures were taken by a given camera.

It seems that this work has been spotted by the community and raised some fears. The site instructables proposes a process “anonymizing” the pictures. Obviously, the author has no serious knowledge of signal processing theory. Some of the tricks are more than questionable. Nevertheless, he is serious. he did not forget the most obvious steps 1 and 6. In step 1, he removes the metadata attached to a picture (How many people ignore or forget that Microsoft documents embed identification metadata?. In step 6, he suggests to use TOR to anonymize the Internet postings.

The lesson is that the community check the latest works of the academic world. Although, they do not necessarily understand the scientific details (thus they may have a wrong estimation of the maturity), they clearly understand the potential consequences and outcomes.

An occasion to read the latest results from Jessica Fridrich? :Wink: Thanks Bertrand