AI Authentication Versus the Physical World

Artificial intelligence is becoming an important component of banknote and identity document authentication, particularly where smartphones or compact readers analyse fine print structures. Research presented in 2025 at Optical & Digital Document Security (ODDS), and follow-on work proposed for ODDS 2027, shows why conventional classification accuracy is not enough to establish that these systems are secure.

The work by Julian Knaup, Julian Bültemeier, Christoph-Alexander Holst and Volker Lohweg of the Institute Industrial IT in Germany focuses on machine-readable intaglio. Genuine banknotes and some identity documents use intaglio printing, which produces characteristic fine-line structures, relief and micro-textures. Counterfeiters are more commonly limited to offset or digital reproduction, creating different spatial-frequency characteristics that can be analysed using wavelet transforms and machine-learning models.