FakeNets project:

morphological analysis for fake generation and similarity metrics for fake detection by neural networks
January 2022 - December 2023

Description

The motivation of this project is to explore the potential offered by deep learning in the context of copy-sensitive graphical codes (such as Copy Detection Patterns) in secure printing, from the attacker's point of view and from the verifier's point of view.

Two main scientific challenges we address in the FakeNets project are:

• reconstruction of CDP which will be approached from the angle of morphological neural networks in order to produce an estimated CDP.
• increase the fake detection performances by learning similarity metrics.

Involved reserchers

• Iuliia Tkachenko - Principal investigator in LIRIS, Lyon
• Thierry Fournel - Principal investigator in Laboratoire Hubert Curien, Saint-Etienne
• Alain Trémeau - Collaborator in Laboratoire Hubert Curien, Saint-Etienne
• Master internship on the topic "Segmentation de codes graphiques par réseaux de neurones morphologiques" in Laboratoire Hubert Curien (01/02/2022 - 22/07/2022)
• Master internship on the topic "Metric learning for forgery detection" in LIRIS (01/04/2023 - 22/07/2023)
• Tetiana Yemelianenko - post-doc resercher on the topic "Similarity learning: state of the art and applications" in LIRIS, funded by PAUSE Ukraine (08/06/2022 - 08/09/2022)

Link via FIL projects

Publications related to the project

1. H. Zeghidi, C. Crispim-Junior, I. Tkachenko, “CDP-Sim: Similarity metric learning to identify the fake Copy Detection Patterns”, IEEE WIFS 2023, December 2023, Nuremberg, Germany.[pdf]
2. T. Yemelianenko, A. Trémeau, I. Tkachenko, “Printed packaging authentication: similarity metric learning for rotogravure manufacture process identification”, VISAPP 2023, February 2023, Lisbon, Portugal.[pdf][dataset]