SOTAMD: State of the art of Morphing Detection

Funding Agency: European Union’s Internal Security Fund – Borders and Visa.

Lead organization: National Office for Identity Data (NOI)

Team members: National Office for Identity Data (NOI), Bundeskriminalamt (BKA), Alma Mater Studiorum – Università di Bologna, Norwegian University of Science and Technology (NTNU), Hochschule Darmstadt, Universiteit Twente

Performance period: February 2019 – January 2020

Project description:

With the widespread deployment of biometric recognition systems, in particular the EU Smart Borders initiative, the interest in attacking these systems is increasing. In the recent past and as one of the main results of the EU funded research project FIDELITY, the vulnerabilities of biometric systems to so-called morphed face image attacks in the application process of identity documents and in automated border control applications have been unveiled to some governments within the EU. In such attacks, biometric samples of multiple subjects are merged, in order to allow an unlawful but successful verification of all contributing subjects against the created artificial identity. This is a rather new area of research and only a few European research institutions have developed detection mechanisms for such attacks. These labs are co-applicants of this project.

The objective of the SOTAMD project is to identify the state-of-the-art of morphed face image detection mechanisms by collecting in a distributed effort a dataset of morphed face images, for which image quality according to ICAO and EU Regulation 2252/2004 is ensured. Furthermore the project will evaluate the existing morphed face image detection mechanisms based on a common established benchmark protocol and morphed face image detection metrics. The project will focus on the scenario of differential morphing detection, where a bona fide facial image (captured in a trusted environment) is compared against the target of investigation (e.g. the printed image in the passport application process) in a signal to signal analysis mode. The project will mimic the automated border control gate scenario, where the live face image of the gate is compared against the image in the chip of the passport.