November 23, 2023

New milestone for the detection of morphing attacks: Fraunhofer HHI researcher wins first place in NIST Challenge

November 23, 2023

Scientist Clemens Seibold has acheived several accomplishments in the Face Analysis Technology Evaluation (FATE) MORPH challenge. His work on the detection of face morph attacks was ranked first worldwide in several categories in the FATE Challenge organized by the US National Institute of Standards and Technology (NIST). At Fraunhofer Heinrich-Hertz-Institut (HHI), Clemens Seibold is investigating the topic of face morphing in the Vision and Imaging Technologies (VIT) department.

State-of-the-art identification processes usually scan the biometric information on ID cards, passports and visas automatically using special facial recognition systems. However, criminals can bypass conclusive identification through so-called "morphing attacks": By digitally merging the faces of two people they create an artificial but realistic new face that contains features of both faces. Using such an image for an official travel document, for example, two people could share the same fake travel document without being stopped by the facial recognition system.

Clemens Seibold developed his detection algorithms while researching his dissertation and contributing to two completed Fraunhofer HHI research projects in particular: ANANAS (Anomaly detection to prevent attacks on facial recognition systems), which was funded by the Federal Ministry of Education and Research as well as the D4Fly project (Detecting Document frauD and iDentity on the fly). 

ANANAS aimed to analyze the risk of morphing attacks and developed methods to detect them. The research on morphing attack detection methods in ANANAS was highly diversified, focusing on semantic image content, camera characteristics and biometrics. Fraunhofer HHI researchers contributed significantly to the consortium project with their investigation of detection methods.

The goal of the D4FLY project was to improve and speed up processes at European border crossing points. The researchers developed new methods to enhance the current capabilities and capacities of border authorities at manual controls and highly automated border checkpoints. Among these were the use of new biometric methods to identify individuals and novel methods to detect various types of identity and document fraud.

Clemens Seibold joined Fraunhofer HHI immediately after completing his computer science degree at Humboldt University in Berlin. He has worked in the "Computer Vision and Graphics" group in the VIT department since December 2015.

The FATE challenge has been ongoing for several years. Researchers around the world can upload their own methods to test them on a non-public data set and have them evaluated by FATE. New algorithms are measured against all previously uploaded methods and ranked accordingly. Their report is updated whenever new algorithms are evaluated, new datasets are added, and new analyses are included. The FATE MORPH assessment will remain open for submission of new algorithms until further notice.