Table 1.
Deepfake detection method proposed in the past three years.
| Methods | Key Features | Architecture | Published |
|---|---|---|---|
| Microscopic analyses [3] |
Mesoscopic properties of images | MesoNet (based on CNN) |
2018 |
| Temporal inconsistencies [4] |
Frame level temporal features | CNN + LSTM | 2018 |
| Eye blinking [5] | Temporal patterns of eye blinking | CNN + LSTM | 2018 |
| Face warping [6] | Inconsistencies in warped face and surrounding area |
VGG16, ResNet50 (based on CNN) |
2019 |
| Discrepancy [7] | Temporal discrepancies across frames | CNN + RNN | 2019 |
| Spoken phoneme mismatches [8] |
Mismatches between the dynamics of the mouth shape |
CNN | 2020 |