Speech Deepfake Detection on ASVspoof 5 (evaluation)
4.04EERSZU-AFS
Evaluation results
| Method | Links | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SZU-AFSTrain=5, Aug=Yes, Fusion=Fusion2026.06 | 4.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MFPT (Ours)Train=5, Aug=No, Fusion=Single2026.06 | 4.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SLIM (Reality Defender)Train=5, Aug=Yes, Fusion=Single2026.06 | 5.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoLExTrain=5, Aug=No, Fusion=Single2026.06 | 5.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QAF-StaticCodec features=codecT, SSL features=sslF, Modeling=Method 2, SSL status=frozen2026.03 | 5.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 13.9 | |
| SSL-IVSPTTrain=5, Aug=Yes, Fusion=Single2026.06 | 5.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QAF-StaticCodec features=codecF, SSL features=sslF, Modeling=Method 1, SSL status=frozen2026.03 | 6.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 8.9 | |
| QAF-StaticCodec features=codecF, SSL features=sslF, Modeling=Method 2, SSL status=frozen2026.03 | 6.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 8.5 | |
| ATTM + LSTM (baseline)SSL status=fine-tuned2026.03 | 6.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| WavLM-ResNet18-SA fusionTrain=5, Aug=Yes, Fusion=Fusion2026.06 | 7.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Best challenge submissiontype=Public baseline2026.03 | 8.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BUT system (fused)Train=5, Aug=No, Fusion=Fusion2026.06 | 9.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSL + SCDFusion System=Concatenation-based fusion (d), Acoustic Feature=Spectral Correlation Density (SCD)2026.03 | 14.8 | 3.78 | 7.73 | 4.68 | 5.61 | 9.02 | 10.54 | 21.43 | 60.01 | 3.93 | 4.79 | 6.95 | 31.46 | 5.16 | 14.57 | 32.26 | 5.67 | 34.23 | — | |
| SSL-onlyFusion System=SSL-only (c)2026.03 | 15.81 | 5.42 | 11.6 | 8.51 | 9.82 | 8.99 | 10.54 | 19.61 | 27.56 | 5.17 | 7.13 | 15.44 | 25 | 5.49 | 24.46 | 37.38 | 12.33 | 36.31 | — | |
| WavLM (FT-DA)type=Public baseline2026.03 | 17.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Wav2Vec2-AASIST-KANtype=Public baseline2026.03 | 22.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SEMAA-1type=Public baseline2026.03 | 23.63 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AASIST-CAM++ fusedtype=Public baseline2026.03 | 25.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSL + CQTFusion System=Concatenation-based fusion (d), Acoustic Feature=CQT2026.03 | 27.54 | 5.93 | 29.67 | 20.88 | 29.27 | 4.78 | 15.25 | 31.92 | 57.76 | 14.46 | 19.77 | 41.16 | 21.01 | 3.03 | 42.41 | 38.32 | 33.45 | 69.15 | — | |
| AASIST (B02)System type=ASVspoof 5 Challenge Baseline2026.03 | 29.12 | 16.44 | 50.02 | 59.99 | 33.7 | 17.05 | 17.63 | 32.78 | 13.35 | 21.01 | 31.35 | 27.6 | 32.1 | 8.93 | 42.85 | 27.48 | 19.5 | 71.06 | — | |
| RawNet2 (B01)System type=ASVspoof 5 Challenge Baseline2026.03 | 36.04 | 22.58 | 57.64 | 63.75 | 46.95 | 25.67 | 24.5 | 30.35 | 23.61 | 29.78 | 41.95 | 38.85 | 39.43 | 17.54 | 42.26 | 33.22 | 29.78 | 82.66 | — | |
| SSL + STFT Spect.Fusion System=Concatenation-based fusion (d), Acoustic Feature=STFT Spectrogram2026.03 | 39.56 | 33.1 | 36.39 | 46.22 | 50.82 | 33.96 | 36.11 | 36.24 | 38.74 | 37.84 | 33.74 | 47.6 | 35.91 | 32.24 | 47.49 | 47.44 | 50.95 | 93.13 | — |