Deepfake Detection on ViT Adversarial
92.85AccuracyOurs (DCT-based fourth-order moment pooling)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Ours (DCT-based fourth-order moment pooling)Attack=Sec5.2 AdvImages w/ Surrogate Models2026.06 | 92.85 | 96.35 | |
| D3Attack=Sec5.2 AdvImages w/ Surrogate Models2026.06 | 81.3 | 90.2 |