Hugging Face's logo

Datasets:
MTUCI
/
RuASD

Tasks:
Audio Classification
Modalities:
Audio
Text
Formats:
webdataset
Languages:
Russian
Size:
100K - 1M
Tags:
audio
speech
anti-spoofing
audio-deepfake-detection
tts
Libraries:
Datasets
WebDataset
Dataset card Data Studio Files
xet
 Community
1
RuASD / README.md
HapBuXBOPOCT's picture
HapBuXBOPOCT
Create README.md
3b5d4e7
preview code
|
raw
history blame contribute delete
10.8 kB
metadata 
language:
  - ru
tags:
  - audio
  - speech
  - anti-spoofing
  - audio-deepfake-detection
  - tts
task_categories:
  - audio-classification
pretty_name: RuASD
size_categories:
  - 100K<n<1M

RuASD: Russian Anti-Spoofing Dataset

RuASD is a public Russian-language speech anti-spoofing dataset designed for developing and benchmarking audio deepfake detection systems. It combines spoofed utterances generated by 37 Russian-capable speech synthesis systems with bona fide recordings curated from multiple heterogeneous Russian speech corpora. In addition to clean audio, the dataset supports robustness-oriented evaluation through reproducible perturbations such as reverberation, additive noise, and codec-based channel degradation.

Models: ESpeech, F5-TTS, VITS, Piper, TeraTTS, MMS TTS, VITS2, GPT-SoVITS, CoquiTTS, XTSS, Fastpitch, RussianFastSpeech, Bark, GradTTS, FishTTS, Pyttsx3, RHVoice, Silero, Fairseq Transformer, SpeechT5, Vosk-TTS, EdgeTTS, VK Cloud, SaluteSpeech, ElevenLabs

Overview

  • Purpose: Benchmark and develop Russian-language anti-spoofing and audio deepfake detection systems, with a focus on robustness to realistic channel and post-processing distortions.
  • Content: Bona fide speech from multiple open Russian speech corpora and synthetic speech generated by 37 Russian-capable TTS and voice-cloning systems.
  • Structure:
    • Audio: .wav files
    • Metadata: JSON with the fields sample_id, label, group, subset, augmentation, filename, audio_relpath, source_audio, metadata_source, source_type, mos_pred, noi_pred, dis_pred, col_pred, loud_pred, cer, duration, speakers, model, transcribe, true_lines, transcription, ground_truth, and ops.

Statistics

  • Number of TTS systems: 37
  • Total spoof hours: 691.68
  • Total bona-fide hours: 234.07

Table 4. Antispoofing models on clean data.

Model Acc Pr Rec F1 RAUC EER t-DCF
AASIST3 0.769±0.0006 0.683±0.001 0.769±0.0006 0.724±0.001 0.841±0.0006 0.231±0.0006 0.702±0.002
Arena-1B 0.812±0.001 0.736±0.001 0.812±0.001 0.772±0.001 0.887±0.0005 0.188±0.001 0.385±0.001
Arena-500M 0.801±0.001 0.722±0.001 0.801±0.001 0.760±0.001 0.864±0.0005 0.199±0.001 0.655±0.002
Nes2Net 0.689±0.0007 0.589±0.001 0.689±0.0007 0.634±0.0008 0.779±0.0007 0.311±0.0007 0.696±0.001
Res2TCNGaurd 0.627±0.001 0.520±0.001 0.627±0.001 0.569±0.001 0.691±0.001 0.373±0.001 0.918±0.001
ResCapsGuard 0.677±0.001 0.575±0.001 0.677±0.001 0.622±0.001 0.718±0.001 0.323±0.001 0.896±0.001
SLS with XLS-R 0.779±0.001 0.700±0.001 0.779±0.001 0.737±0.001 0.859±0.001 0.221±0.001 0.650±0.001
Wav2Vec 2.0 0.772±0.0006 0.687±0.001 0.772±0.0006 0.727±0.001 0.850±0.0006 0.228±0.0006 0.558±0.002
TCM-ADD 0.857±0.001 0.797±0.001 0.859±0.001 0.827±0.001 0.914±0.0004 0.143±0.001 0.424±0.001

Table 5. Antispoofing models on augmented data (EER). Aug. denotes the applied degradation: R — RIR reverberation, N — MUSAN additive noise, and suffixes (alaw, amr, g722, mp3, mlw, op16, op8, spx8) indicate encode-decode transcoding with the corresponding codec. Combined labels such as RNmp3 apply R + N followed by codec transcoding.

Aug. AAS3 AR1B AR5M N2NT R2NT RCPS XSLS W2AS TCM
Codec only
alaw 0.468 0.331 0.237 0.435 0.331 0.332 0.485 0.242 0.270
amr 0.373 0.133 0.147 0.378 0.271 0.272 0.478 0.212 0.372
g722 0.279 0.264 0.245 0.353 0.323 0.323 0.473 0.275 0.190
mp3 0.239 0.191 0.171 0.340 0.322 0.318 0.475 0.270 0.187
mlw 0.463 0.330 0.238 0.449 0.333 0.325 0.478 0.234 0.261
op16 0.264 0.216 0.176 0.383 0.308 0.303 0.481 0.278 0.297
op8 0.341 0.208 0.205 0.418 0.293 0.297 0.481 0.297 0.404
spx8 0.372 0.141 0.147 0.370 0.305 0.302 0.475 0.273 0.420
Noise: N and N+Codec
N 0.458 0.446 0.360 0.440 0.330 0.310 0.481 0.292 0.292
Nalaw 0.503 0.430 0.321 0.483 0.320 0.318 0.498 0.291 0.380
Namr 0.441 0.264 0.235 0.448 0.286 0.269 0.481 0.270 0.476
Ng722 0.448 0.428 0.357 0.434 0.323 0.305 0.479 0.296 0.296
Nmp3 0.448 0.372 0.291 0.437 0.327 0.312 0.476 0.296 0.296
Nmlw 0.502 0.402 0.304 0.480 0.319 0.316 0.497 0.282 0.373
Nop16 0.420 0.384 0.305 0.447 0.321 0.289 0.481 0.319 0.414
Nop8 0.437 0.336 0.319 0.473 0.309 0.301 0.475 0.348 0.512
Nspx8 0.430 0.234 0.217 0.428 0.288 0.273 0.479 0.312 0.490
Reverberation: R and R+Codec
R 0.351 0.482 0.319 0.499 0.332 0.336 0.483 0.331 0.456
Ralaw 0.472 0.488 0.373 0.564 0.313 0.342 0.457 0.347 0.420
Ramr 0.444 0.404 0.288 0.515 0.334 0.346 0.479 0.339 0.396
Rg722 0.397 0.491 0.305 0.500 0.337 0.348 0.489 0.334 0.476
Rmp3 0.394 0.444 0.243 0.515 0.326 0.336 0.491 0.336 0.454
Rmlw 0.471 0.488 0.365 0.565 0.318 0.346 0.468 0.341 0.406
Rop16 0.388 0.444 0.295 0.489 0.328 0.329 0.491 0.366 0.494
Rop8 0.410 0.421 0.308 0.509 0.321 0.337 0.499 0.380 0.405
Rspx8 0.454 0.400 0.292 0.504 0.316 0.341 0.490 0.361 0.413
Combined: RN and RN+Codec
RN 0.503 0.486 0.408 0.527 0.324 0.379 0.504 0.365 0.511
RNalaw 0.493 0.493 0.401 0.551 0.316 0.381 0.528 0.372 0.422
RNamr 0.473 0.478 0.349 0.515 0.319 0.377 0.512 0.352 0.385
RNg722 0.477 0.498 0.401 0.529 0.322 0.384 0.504 0.366 0.473
RNmp3 0.469 0.487 0.350 0.530 0.332 0.371 0.505 0.366 0.510
RNmlw 0.503 0.492 0.397 0.544 0.310 0.377 0.510 0.379 0.409
RNop16 0.477 0.498 0.403 0.519 0.329 0.385 0.506 0.400 0.427
RNop8 0.493 0.496 0.405 0.510 0.324 0.384 0.488 0.390 0.379
RNspx8 0.477 0.449 0.333 0.507 0.318 0.387 0.497 0.379 0.391

Download

Using Datasets 

from datasets import load_dataset

ds = load_dataset("MTUCI/RuASD")
print(ds)

Git clone 

git lfs install
git clone https://huggingface.co/datasets/MTUCI/RuASD

Contact