Hugging Face's logo

SpeechAntiSpoofingBenchmarks
/
ProSDD-ASV19

PyTorch
audio
anti-spoofing
deepfake-detection
speech-anti-spoofing-arena
Model card Files
xet
 Community
ProSDD-ASV19 / README.md
korallll's picture
korallll
Model card with Arena badges; final meta.yaml
d4060a0 verified
preview code
|
Raw History Blame Contribute Delete
7.97 kB
---
license: other
tags:
- audio
- anti-spoofing
- deepfake-detection
- speech-anti-spoofing-arena
library_name: pytorch
---
# ProSDD-ASV19: Speech Anti-Spoofing Arena wrapper
Packaged for the [Speech Anti-Spoofing Arena](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena) from the authors' [code](https://github.com/ProSDD/codes/tree/bdb80440cd946368e77f4b51f6ddf26c7443dc33) and [paper](https://arxiv.org/abs/2604.13229).
ProSDD (Mahapatra et al., 2026): a fully fine-tuned wav2vec 2.0 XLS-R 300M encoder, first trained (Stage I, bona fide LibriSpeech train-clean-100 + dev only) to predict masked frame-level speaker + prosody targets, then (Stage II) jointly fine-tuned with that objective and a binary real/fake head on ASVspoof 2019 LA train/dev (RawBoost series 3). Inference uses only the clean path: XLS-R CNN front-end + feature projection, latent sequence fixed at 200 frames (zero-padded), 24-layer transformer encoder, mean pool, Linear(1024,512)-ReLU-Linear(512,2). Input as the authors' data_utils_eval.py: 16 kHz, no amplitude normalisation, fixed 4 s (64000 samples) centre crop, shorter clips centre zero-padded. FP32; score = logits[:,1] (bona fide class). Authors' released checkpoint; no upstream license, so the weights are linked from the authors' Google Drive, not re-hosted.
- Weights: **not re-hosted** (the authors grant no license for them). Get the authors' [checkpoint](https://drive.google.com/file/d/1RHY-VELyi4XH-L1egeRcgWjHdkdWV2bO/view); its sha256 is in the `meta.yaml` notes.
- Params: 315.963522 M. Trained on LibriSpeech (stage 1) and ASVspoof 2019 LA train/dev (stage 2); the Arena's 19LA set is eval.
- Score: higher = bona fide. J-SPAW_LA was scored with the PyTorch wrapper; the other datasets with the
TensorRT drop-in class (FP32 on the RTX 4070 Ti SUPER; FP32 on the RTX 5060 Ti), which matched PyTorch on the full J-SPAW_LA set.
## Arena results
[![arena tier](https://img.shields.io/endpoint?url=https://speechantispoofingbenchmarks-speechantispoofingarena.hf.space/badge/prosdd-asv19/tier.json)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![arena rank](https://img.shields.io/endpoint?url=https://speechantispoofingbenchmarks-speechantispoofingarena.hf.space/badge/prosdd-asv19/rank.json)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
EER% (lower is better); EmoSpoofTTS and LRLspoof are spoof-only and use 1-SRR% at the model's
DeepVoice EER threshold. Per-dataset scores are under `.eval_results/`.
[![EER% 0.42 on ASVspoof2019_LA](https://img.shields.io/badge/EER%25%20on%20ASVspoof2019__LA-0.42%25-brightgreen)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 8.29 on ASVspoof2021_LA](https://img.shields.io/badge/EER%25%20on%20ASVspoof2021__LA-8.29%25-yellow)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 8.61 on ASVspoof2021_DF](https://img.shields.io/badge/EER%25%20on%20ASVspoof2021__DF-8.61%25-yellow)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 16.14 on ASVspoof5](https://img.shields.io/badge/EER%25%20on%20ASVspoof5-16.14%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 15.16 on InTheWild](https://img.shields.io/badge/EER%25%20on%20InTheWild-15.16%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 20.6 on ADD22_eval_31](https://img.shields.io/badge/EER%25%20on%20ADD22__eval__31-20.6%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 34.17 on ADD2023_track12_test_r1](https://img.shields.io/badge/EER%25%20on%20ADD2023__track12__test__r1-34.17%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 21.1 on CFAD](https://img.shields.io/badge/EER%25%20on%20CFAD-21.1%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 34.04 on CD-ADD](https://img.shields.io/badge/EER%25%20on%20CD--ADD-34.04%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 16.37 on CVoiceFake_small](https://img.shields.io/badge/EER%25%20on%20CVoiceFake__small-16.37%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 46.89 on DeepVoice](https://img.shields.io/badge/EER%25%20on%20DeepVoice-46.89%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 16.87 on LibriSeVoc](https://img.shields.io/badge/EER%25%20on%20LibriSeVoc-16.87%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 33.55 on SONAR](https://img.shields.io/badge/EER%25%20on%20SONAR-33.55%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 35.16 on ArAD](https://img.shields.io/badge/EER%25%20on%20ArAD-35.16%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 26.14 on DECRO](https://img.shields.io/badge/EER%25%20on%20DECRO-26.14%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 24.73 on J-SPAW_LA](https://img.shields.io/badge/EER%25%20on%20J--SPAW__LA-24.73%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 42.36 on ODSS](https://img.shields.io/badge/EER%25%20on%20ODSS-42.36%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 6.97 on HABLA](https://img.shields.io/badge/EER%25%20on%20HABLA-6.97%25-yellow)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 17.88 on DFADD](https://img.shields.io/badge/EER%25%20on%20DFADD-17.88%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 16.67 on PyAra](https://img.shields.io/badge/EER%25%20on%20PyAra-16.67%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 34.12 on XMAD](https://img.shields.io/badge/EER%25%20on%20XMAD-34.12%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![EER% 3.7 on EmoFake_test](https://img.shields.io/badge/EER%25%20on%20EmoFake__test-3.7%25-green)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![SRR COMPLEMENT 24.87 on EmoSpoofTTS](https://img.shields.io/badge/SRR%20COMPLEMENT%20on%20EmoSpoofTTS-24.87%25-blue)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
[![SRR COMPLEMENT 33.58 on LRLspoof](https://img.shields.io/badge/SRR%20COMPLEMENT%20on%20LRLspoof-33.58%25-blue)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=prosdd-asv19)
## Citation
```bibtex
@article{mahapatra2026prosdd,
title = {ProSDD: Learning Prosodic Representations for Speech Deepfake Detection against Expressive and Emotional Attacks},
author = {Mahapatra, Aurosweta and Ulgen, Ismail Rasim and Lee, Kong Aik and Andrews, Nicholas and Sisman, Berrak},
journal = {arXiv preprint arXiv:2604.13229},
year = {2026}
}
```