EINS Research and Development Projects:

Which technology a document was printed with (source printer) or whether all pages were produced by the same printer and on the same paper (substrate) is valuable information when detecting document forgery, even if the primary security features were successfully copied. 
This is where the "MLForPrint" project comes in: However, while manual forensic document examination can take hours and requires the examiner's many years of experience and is therefore comparatively rarely used, the project uses automated procedures based on machine learning for this purpose. 
The goal is that a software-based and automated examination of printed products and substrates can bring about a reduction in the examination effort with comparable accuracy. For this purpose, the project uses neural networks (CNNs), with which it has already been possible to demonstrate in prototype form that it is possible to classify documents efficiently with regard to properties such as printing technology (such as offset, dry/wet toner or ink jet). 
The goals of the "MLforPrint" project are, on the one hand, to research and improve the robustness of the CNN against disturbances that could be specifically used by counterfeiters and, on the other hand, to classify substrates for which a software solution is to be demonstrated for the first time that can derive paper types, aging states and condition predictions from scans. The challenge of a learned, i.e., data-oriented approach is to be able to react quickly to unknown documents and textures. For use in digital forensics, it is further important to improve the explainability of the deployed CNN in order to better understand its decisions and optimize parameters of the network for deployment purposes.

This joint project was funded until 02/2024 by the BMBF. Funding measure KMU-innovativ: IKT.