Morphing attack detection by secunet
Update on FRVT-MORPH test of NIST from November 16, 2022
These types of fraud include so-called morphing attacks, in which fraudsters use image editing software to merge their biometric passport photos into a single image that then looks like a composite of two people. When successfully applying this photo for a passport, both fraudsters are then recognizable as the person in the picture and can thus both use the same identity document. If the morphing is done well, neither the facial recognition software nor the officer at the border control counter will notice the difference between the person and the morphed image.
Software algorithms that recognize facial morphs during automated border control can increase border security substantially. Therefore, work on morphing attack detection (MAD) has increasingly been going on for several years by now.
secunet has recently submitted its algorithm for the recognition of morphed facial images to the independent and internationally recognized National Institute of Standards and Technology (NIST) Face Recognition Vendor Test (FRVT)-MORPH test. secunet’s algorithm implements a differential morphing attack detection, which checks a potentially morphed facial image against a second, usually live captured and thus trusted image. The result from the NIST FRVT-MORPH test shows that the recognition of morphed images has now reached a level of performance that allows operational use in the context of border control. Interested readers can view the full report here.
Especially in the category of "High Quality Morphs", e.g., for manually generated morphs, good results were achieved: If set to a false positive rate of approximately 4% (meaning four out of every hundred facial images are misclassified as morphs and have to be manually verified), secunet’s algorithm spotted 84% of all morphed images. This presents a huge step in comparison to both human test subjects and previous algorithms.
The new algorithm is integrated throughout secunet’s entire border control portfolio and complements other security measures that are already available. Among others, this includes the automated border control system secunet easygate as well as the self-service system secunet easykiosk for pre-enrolment of passenger data. The new algorithm will also be included in the border control application secunet bocoa in combination with secunet easytower as facial image camera. All products and solutions from secunet provide sufficiently high-quality live images for the use as trusted comparison images for differential MAD.
As a next step, secunet has already identified areas of improvement to further optimize the performance of the algorithm.