University of Twente Student Theses


Detecting GAN-generated face morphs using human iris characteristics

Hesson, Alexandru (2023) Detecting GAN-generated face morphs using human iris characteristics.

[img] PDF
Abstract:Face morphing is a computer graphic technique that merges the features of two individuals into a single composite image and it finds applications in diverse fields such as entertainment and cosmetic surgery simulation. However, it also poses a potential threat in the form of identity document forgery. This research focuses on the detection of morphed facial images generated by Generative Adversarial Networks (GANs) by analysing the characteristics of the iris. In genuine human irises, the shape is predominantly circular, whereas morphed images may exhibit extreme deformations or deviate from the expected roundness. To address this issue, a program capable of scanning facial images and extracting the iris contour for radius samples was developed to be used in evaluating the iris roundness. A dataset comprising 44 authentic irises and 44 morphed irises was compiled for experimenting, ensuring diversity by including individuals from various ethnic backgrounds. The findings from this study showed that using iris detection alone is not accurate enough to reliably detect morphs, but it can be added to a multi-facet approach to improve the accuracy of already existing morphing detection systems.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science BSc (56964)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page