Statistical analysis of digital image fingerprinting based on random projections

Digital fingerprints have recently received a lot of interest in multimedia applications due to their low dimensionality and security/privacy preserving capabilities. In digital fingerprint extraction, random projections play an important role due to their approximate distance preserving property. This paper is dedicated to the analysis of the statistical properties of digital fingerprints obtained based on random projections. In particular, we were able to demonstrate that this sort of mapping guarantees approximate decorrelation of its output with high probability.