University of Twente Student Theses
GAN Generated Morphing Attack Detection by Analyzing the Geometry of Pupils
Jerkovic, Christian (2023) GAN Generated Morphing Attack Detection by Analyzing the Geometry of Pupils.
|
PDF
1MB |
| Abstract: | Facial morphing is the process in which 2 images of 2 distinct faces are merged in order to create a new image that is a blend of both. Recent innovations in the field have resulted in facial morphs that are difficult for humans to identify visually. The type of morphs that will be evaluated in this research are GAN (Generative Adversarial Network) morphs. In order to combat the unreliability of humans’ ability to distinguish between authentic images and generated ones, numerous methods for morphing attack detection have been developed and are being improved. Face morphing attacks have proven to be a challenge that severely hampers border control agencies’ capabilities in detecting and preventing passport forgery. Due to the morphing procedure, however, the eyes of the resulting morphed image tend to have artefacts. These artefacts in many cases can distort the shape of the pupils and cause them to deviate in shape from a circle. Firstly, the eyes will be extracted from the source image in order to remove eyelashes and any other elements that could impede the segmentation of the pupil. After adjusting the image to improve the definition of the pupil, a contour will be drawn around it. The roundness of the pupil can then be used to evaluate whether or not a pupil is the result of a GAN Morph. In addition to the roundness of the original pupil contour, the roundness of the convex hull of the pupil was utilized for the detection. Finally, the ratio between the areas of the original pupil contour and its convex haul was used as a feature. The classification was done using a decision tree classifier model. This method correctly classified 66.6% of eyes as either morphed or bonafide. |
| 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: | https://purl.utwente.nl/essays/96011 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page