TABLE 6.
Classification accuracy of the proposed and existing algorithm for video based morphing attack detection on the FaceForensics database.
| Algorithm/Network | Accuracy % |
|---|---|
| Steganalysis Features + SVM Fridrich and Kodovsky, (2012) | 99.40 |
| Cozzolino et al. (2017) | 99.60 |
| Bayar and Stamm (2016) | 99.53 |
| Rahmouni et al. (2017) | 98.60 |
| Raghavendra et al. (2017b) | 97.70 |
| Zhou et al. (2017) | 99.93 |
| XceptionNet Chollet (2017) | 99.93 |
| MesoNet Afchar et al. (2018) | 96.80 |
| VGG-16 Simonyan and Zisserman (2015) | 99.50 |
| ResNet-50 He et al. (2016) | 99.93 |
| ResNet-152 He et al. (2016) | 99.89 |
| Multi-patch ResNet-18 Kumar et al. (2020) | 99.96 |
| Proposed (MagNet) | 100.00 |
Result of the best performing algorithm is highlighted in bold.