Speaker Verification on VoxCeleb1
3.35PGD EERNoise-0.01
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Noise-0.01Attack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 3.35 | — | — | — | |
| PnP-Diff-2Attack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 3.4 | — | — | — | |
| AudioPureAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 3.55 | — | — | — | |
| PnP-DiffAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 3.55 | — | — | — | |
| PnP-Diff + AudioPureAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 3.9 | — | — | — | |
| PnP-Diff + DiffWavePnPAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 4.8 | — | — | — | |
| PnP-GaussianAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 6.15 | — | — | — | |
| SpeechTokenizerAttack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 8.65 | — | — | — | |
| MDD-10%*Attack type=PGD-l2, Attack steps=50, Base ASV Model=ECAPA-TDNN2026.07 | 18 | — | — | — | |
| ChaRVoCSpeaker Extractor=RawNet22026.05 | — | 2.879 | 0.995 | 0.751 | |
| IoM2026.05 | — | 8.26 | 0.856 | 0.631 | |
| RoE2026.05 | — | 3.213 | 0.959 | 0.713 | |
| WTA2026.05 | — | 9.341 | 0.798 | 0.576 |