Pith. sign in

REVIEW 4 major objections 5 minor 52 references

Generalizable Detection of Audio Deepfakes

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that a single audio-deepfake detector can beat the best reported single system of the ASVspoof 5 challenge, reaching a 4.48% equal-error rate on its test set while also generalizing across many other spoof benchmarks.

desk verdict Solid ablation study with an interesting teacher-based data-mixing trick, but the SOTA claim is compromised by test-set-driven configuration selection and an unfair comparison. read the letter →

arxiv 2507.01750 v1 pith:NPWEVJQT submitted 2025-07-02 eess.AS cs.SD

Jose A. Lopez , Georg Stemmer , Héctor Cordourier Maruri This is my paper
classification eess.AScs.SD
keywords audiodeepfakedetectionspoofingcountermeasuresgeneralizationself-supervisedspeechmodelsdataaugmentationfocallosscenterASVspoof
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that a single audio-deepfake detector can generalize beyond the dataset it was trained on. It reports that one model, built from a pre-trained Wav2Vec2 XLS-R 300M backbone, a simple classifier head, a focal-plus-hinged-center loss, and a heavy augmentation recipe, reaches a 4.48% equal-error rate on the ASVspoof 5 test set, better than the 5.56% of the best reported single system in that challenge. The key move is to feed ASVspoof 5 information through a teacher model instead of mixing that data directly into training, because direct training improved some benchmarks but degraded others such as In-The-Wild. If true, this gives a practical single-model recipe for deepfake detection that does not require an ensemble, and it isolates loss and augmentation choices that transfer across years and data sources.

What carries the argument

The central mechanism is configuration R's full pipeline: a Wav2Vec2 XLS-R 300M self-supervised backbone preceded by bandpass filtering (0.3-3.4 kHz) and power normalization, with temporal average pooling and a three-layer fully connected head trained using focal loss ($\gamma=2$) plus a hinged center loss written as $\max(0, L_{\mathrm{center}} - 1)$, under augmentation with additive white Gaussian noise, room impulse responses, and RawBoost. A second mechanism is teacher-based knowledge transfer: a teacher model trained only on ASVspoof 5 provides soft supervision so the student never sees ASVspoof 5 training data directly, which the paper shows avoids cross-dataset degradation. The augmentation ablations show RawBoost carries the largest single gain, and the bandpass filter removes out-of-band spectral content that earlier first-generation models tended to overfit.

What would settle it

Rerun the J-through-R configurations while choosing the final model only on a validation split that contains none of the test sets in Table IX, then evaluate the chosen model on the full ASVspoof 5 test set; if its EER is not below the 5.56% single-system baseline, the reported gain depends on test-set selection rather than on the method itself.

Watch

Extended reading notes

Core claim

Stated on the paper's own terms, the discovery is that generalization in audio deepfake detection is driven more by the training recipe than by the architecture. A self-supervised backbone (Wav2Vec2 XLS-R 300M), preceded by bandpass filtering between 0.3 kHz and 3.4 kHz, followed by average pooling and a three-layer fully connected head, reaches low equal-error rates on every benchmark the authors evaluated: ASVspoof 2015, 2019, 2021 (logical access and deepfake), In-The-Wild, M-AILABS/MLAAD, FakeAVCeleb, and ASVspoof 5. On the full ASVspoof 5 test set the single model scores 4.48% EER, below the 5.56% of the challenge's best reported single system and competitive with top ensembles. The authors identify the decisive ingredients as focal loss with $\gamma=2$, a hinged center loss that stops the compactness term from fighting the classification loss, augmentation with additive noise, room impulse responses, RawBoost, and vocoded speech from 28 public vocoders, and indirect transfer of ASVspoof 5 knowledge through a teacher model.

Load-bearing premise

The reported generalization scores assume the model configuration was not selected by peeking at the benchmark test sets; configuration R was chosen in Section VIII because it had the lowest average EER across the same test sets that Table IX later reports, and if that selection materially inflated the numbers, the 4.48% versus 5.56% comparison is optimistic.

Editorial extensions

If this is right

  • Because the winning configuration is a single model, the result makes edge deployment more plausible: no ensemble of large language models is needed at inference time.
  • The same model reports low equal-error rates across the evaluated benchmark families, suggesting the recipe captures a generalizable 'fake audio' cue rather than dataset-specific artifacts.
  • Direct inclusion of ASVspoof 5 training data hurt some out-of-distribution sets, while teacher distillation improved them; this makes teacher-based transfer a reusable tool for incorporating new attack corpora.
  • Focal loss plus hinged center loss improves over cross-entropy and one-class softmax without adding inference cost, so the loss change is nearly free at deployment.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to apply the teacher-distillation trick to other large spoof corpora: any dataset that is too costly, too license-restricted, or too domain-shifted to train on directly could be injected through a teacher, and the paper's ASVspoof 5 result predicts this should improve rather than hurt generalization.
  • Because the reported EER rises sharply on short and noisy speech, a production detector built from this recipe could route such inputs to a separate 'low confidence' channel instead of forcing a real/fake decision.
  • The paper's gains on M-AILABS/MLAAD data hint that English-trained cues may transfer to other languages, but the paper does not test this directly; scoring configuration R on non-English subsets of MLAAD or the ADD challenge data would separate language-agnostic cues from English-specific artifacts.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper presents an empirical study of audio deepfake detection using self-supervised pretrained backbones (Wav2Vec2, WavLM, Whisper), various loss functions (focal loss, hinged center loss, one-class softmax), data augmentation strategies (AWGN, RawBoost, RIR, resampling, proprietary vocoded data), and a teacher-student setup in which ASVspoof5 information is provided through a teacher model rather than direct training data. The headline claim is that a single model, configuration R, achieves an EER of 4.48% on the ASVspoof5 test set, surpassing the best reported single system in the ASVspoof5 challenge (5.56%), and obtains a 2.62% average EER across a wide range of benchmarks including ASVspoof 2015/2019/2021, In-The-Wild, M-AILABS/MLAAD, and DFDC. The paper also reports analyses of duration and speech-quality effects and a fairness check on the FB ASR Fairness dataset.

Significance. If the reported protocol were prespecified, the result would be a meaningful empirical contribution: a non-ensemble detector with strong cross-dataset generalization, documented configuration choices, and per-dataset EERs across many benchmarks. The paper is transparent about its training data, augmentation choices, and the teacher-student design, and Table VIII provides a useful comparison of eleven configurations. However, the headline claim is weakened by the fact that configuration R was selected by evaluating configurations J-R on the same test sets later reported, and the teacher model in R was itself selected on ASVspoof5 test EER. All metrics come from single runs without confidence intervals, and several ASVspoof5 evaluations use a 100k-file sample rather than the full test set. These issues do not disprove the empirical findings, but they mean the central comparison to the ASVspoof5 challenge is not an unbiased estimate of a prespecified system's performance.

major comments (4)
  1. [Section VIII, Tables VII-IX] The central claim that configuration R surpasses the best single ASVspoof5 system is compromised by selection on the test sets that are later reported. Section VIII states that configurations J-R were compared and that the best performance was achieved with configuration R, and Table IX is explicitly labeled 'PERFORMANCE OF MODEL WITH LOWEST AVERAGE EER.' Thus the reported 4.48% ASVspoof5 EER and 2.62% average EER are minima over at least ten configurations evaluated on the same benchmarks reported in Tables VIII and IX, not unbiased estimates of a fixed model's generalization performance. Comparing this minimum to the 5.56% best single challenge system in Table IX therefore overstates the improvement unless the authors can show that selection bias is negligible. Please either select the configuration on a held-out validation split, report all configurations with a selection-bias-aware analysis, or reframe the claim as an exploratory best-of-N result rather than a state-of-the-art comparison.
  2. [Section VII-A, Table VI, and Section VIII] The teacher model used in configuration R inherits additional test-set information. Configuration H, which is the teacher in configurations Q and R (Table VII), was chosen in Table VI as the configuration with the lowest ASVspoof5 test EER among F-I (3.57% on the full ASVspoof5 test set). The same ASVspoof5 test set is then the headline benchmark for the final model. Consequently, the reported 4.48% EER reflects not only the final configuration selection but also the teacher selection, compounding the selection bias described above. An independent validation protocol that does not use the ASVspoof5 test set for either teacher or student selection is needed to support the claimed comparison.
  3. [Section IV and Tables VIII-IX] All reported EERs are single-run results with no seeds, variance estimates, or confidence intervals. In Table VIII, several configurations are evaluated on a 100k-file ASVspoof5 sample (indicated by the dagger), while the final headline result uses the full test set; without uncertainty quantification, differences of a few tenths of a percent between configurations, such as the 3.60% vs. 4.34% difference between configurations Q and R on the ASVspoof5 test sample, are within plausible sampling or training noise. At minimum, the ASVspoof5 comparison that supports the main claim should include multiple seeds or bootstrap confidence intervals for the final configuration.
  4. [Section IX and Figure 3] The claim that the FB ASR Fairness evaluation 'did not indicate any kind of bias' is stronger than the presented evidence supports. The text reports only that the model correctly identified samples as authentic across categories; no per-category EERs, confidence intervals, or statistical tests are provided, and the dataset appears to contain only authentic speech, so the analysis cannot assess bias in fake-detection behavior across groups. The authors should either report per-group error rates with uncertainty or soften the claim accordingly.
minor comments (5)
  1. [Section IV] The evaluation protocol is described inconsistently: Section V-A states that Table II used full test files, while subsequent experiments used 3.5-second windows with a 0.5-second step, but the exact procedure used for Tables VIII and IX is not stated. Please specify the windowing and scoring procedure for each main results table.
  2. [Section VII-A] There is a typo in 'In hingsight' near the end of Section VII-A; it should read 'In hindsight.'
  3. [Table VIII] The dagger notation for ASVspoof5 test subsets is not fully explained in the table footnote: for some configurations only the 100k sample is reported, while for others the full test is reported. Please add a clear footnote stating which configurations use the sample and which use the full test set.
  4. [Section VI] The claim that focal loss and hinged center loss are 'previously unexplored in the deepfake detection literature' should be supported by a brief related-work search; center loss was already used in [39], and focal loss is a standard method, so the novelty likely lies in the specific combination and the hinge modification rather than in the individual loss functions.
  5. [Table IX] The DFDC row reports only accuracy (92.40%) without the corresponding threshold-dependent definition used for other rows; since the text later says a threshold of 0.5 and uncalibrated predictions were used, please state this explicitly in the table caption or a footnote.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper reports direct empirical measurements on external benchmarks; configuration selection on test sets is a validity concern, not a derivation that folds back into its inputs.

full rationale

The paper is an empirical study, not a derivation chain. Reported EERs are direct measurements on external ASVspoof, In-The-Wild, M-AILABS/MLAAD, FakeAVCeleb, and DFDC benchmarks, from models trained on ASVspoof2019 LA, ASVspoof5 training data, and a proprietary vocoded collection. The teacher model (config H) uses ASVspoof5 training data, not test data, and the comparison baseline of 5.56% EER is the external ASVspoof5 challenge result. Loss functions (focal, hinged center) and augmentations are standard, cited, or explicitly described; none is defined in terms of the target EER. The only notable protocol issue is that configuration R was selected after comparing configurations J-R on the same benchmarks later reported in Section VIII and Tables VIII-IX, so the headline numbers are affected by selection-on-test-set. That is a validity/optimism concern about unbiased generalization estimates, not a logical circularity, because the model weights are not fit to the test labels and no reported EER is an identity with a training objective or with the selection criterion itself. No load-bearing self-citation, uniqueness-imported-from-authors, or ansatz-smuggled-via-citation step occurs. Under the stated hard rules, the derivation is self-contained and the score is 0.

Assumptions & free parameters 11 free parameters · 5 assumptions · 0 invented entities

The paper is an empirical study, so the main load-bearing inputs are training data, preprocessing choices, and hyperparameters rather than derived axioms. No new physical entities are introduced. The most fragile modeling assumption is that vocoder-generated speech can stand in for realistic deepfakes, and the most fragile evaluation premise is that test-set-based configuration selection leaves the reported generalization numbers unbiased.

free parameters (11)
  • focal loss focusing parameter gamma = 2.0
    Fixed in Section VI for most experiments; controls emphasis on hard examples.
  • hinged center loss margin and beta = margin=1.0, beta=20.0
    Introduced in Equations 3-4 and fixed by hand.
  • backbone learning rate = 1e-6
    Section IV optimizer setting for the pre-trained backbone.
  • classifier head learning rate = 1e-3
    Section IV optimizer setting for the classifier head.
  • weight decay = 0 (backbone), 0.1 (head)
    Section IV setting for the two parameter groups.
  • training and evaluation segment duration = 3.5 seconds
    Used for training and windowed evaluation; affects EER.
  • bandpass filter band = 0.3 to 3.4 kHz
    Preprocessing from Tomilov et al. [39]; applied in all experiments.
  • random power scaling range = 1e-5 to 1.2
    Training-time augmentation in Section III.
  • AWGN settings = SNR 5-30 dB, applied 50% of the time
    Section VII data augmentation configuration.
  • RawBoost application probability = 75%
    Section VII data augmentation configuration.
  • Platt calibration coefficients a0, a1 = fit on calibration data
    Equation 6; used where it improved threshold-dependent metrics.
assumptions (5)
  • domain assumption Vocoder-generated speech is equivalent to an ideal TTS engine or voice converter for creating fake training audio.
    Stated in Section II-A: 'The underlying assumption is that using vocoders is equivalent to using an ideal text-to-speech engine or voice converter.'
  • domain assumption Bandpass filtering to 0.3-3.4 kHz preserves the information needed for deepfake detection after Wav2Vec2 feature extraction.
    Applied in all experiments following [39]; no dedicated validation for SSL backbones is reported.
  • domain assumption The benchmark test sets used for evaluation are representative of real-world deepfake audio and are not themselves used for model selection.
    The generalization claim in Sections VIII and IX depends on this; Section VIII shows test-set-based configuration selection, so the second half of the assumption is violated.
  • domain assumption Pre-trained self-supervised speech representations transfer to the spoofing-detection task when fine-tuned with a small head.
    Chosen backbone design built on Wang and Yamagishi [20] and validated only through the reported benchmark scores.
  • standard math EER and AUC are threshold-independent and directly comparable across systems reported in different papers.
    Used in Table IX to compare with literature references; comparability depends on identical evaluation protocols.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generalizable Detection of Audio Deepfakes." pith.science (2026). https://pith.science/paper/NPWEVJQT

@misc{pith2026250701750,
  author       = {Pith},
  title        = {Pith review of: Generalizable Detection of Audio Deepfakes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NPWEVJQT}},
  note         = {Machine review of arXiv:2507.01750}
}
read the original abstract

In this paper, we present our comprehensive study aimed at enhancing the generalization capabilities of audio deepfake detection models. We investigate the performance of various pre-trained backbones, including Wav2Vec2, WavLM, and Whisper, across a diverse set of datasets, including those from the ASVspoof challenges and additional sources. Our experiments focus on the effects of different data augmentation strategies and loss functions on model performance. The results of our research demonstrate substantial enhancements in the generalization capabilities of audio deepfake detection models, surpassing the performance of the top-ranked single system in the ASVspoof 5 Challenge. This study contributes valuable insights into the optimization of audio models for more robust deepfake detection and facilitates future research in this critical area.

Figures

Figures reproduced from arXiv: 2507.01750 by the authors.

Figure 1
Figure 1. Model Architectures [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. EER Distribution Across SI-SDR Bins X. CONCLUSIONS Our research has confirmed the value of data augmentation in training deepfake detection systems, and we have introduced novel loss functions that were previously unexplored in the deepfake detection literature. Through our exploration, we have developed a state-of-the-art approach that significantly [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 46 canonical work pages

  1. [1]

    Fbi warns of increasing threat of cyber criminals utilizing artificial intelligence,

    “Fbi warns of increasing threat of cyber criminals utilizing artificial intelligence,” https://www.fbi.gov/contact-us/field- offices/sanfrancisco/news/fbi-warns-of-increasing-threat-of-cyber- criminals-utilizing-artificial-intelligence, accessed: 2025-03-13

    work page 2025

  2. [2]

    Pay-to-play: Access to justice in the era of ai and deep- fakes,

    R. Delfino, “Pay-to-play: Access to justice in the era of ai and deep- fakes,” Seton Hall Law Review , vol. 55, pp. 789–845, 01 2025

    work page 2025

  3. [3]

    Asvspoof 2015: the first automatic speaker verification spoofing and countermeasures challenge,

    Z. Wu, T. Kinnunen, N. Evans, J. Yamagishi, C. Hanilc ¸i, M. Sahidullah, and A. Sizov, “Asvspoof 2015: the first automatic speaker verification spoofing and countermeasures challenge,” in Interspeech 2015 , 2015, pp. 2037–2041

    work page 2015

  4. [4]

    Automatic speaker verification spoofing and countermeasures challenge (asvspoof 2015) database,

    Z. Wu, T. Kinnunen, N. Evans, and J. Yamagishi, “Automatic speaker verification spoofing and countermeasures challenge (asvspoof 2015) database,” 2015. [Online]. Available: https://datashare.ed.ac.uk/handle/10283/853

    work page 2015

  5. [5]

    The asvspoof 2017 challenge: Assessing the limits of replay spoofing attack detection,

    T. Kinnunen, M. Sahidullah, H. Delgado, M. Todisco, N. Evans, J. Ya- magishi, and K. A. Lee, “The asvspoof 2017 challenge: Assessing the limits of replay spoofing attack detection,” in Interspeech 2017, 2017, pp. 2–6

    work page 2017

  6. [6]

    The 2nd automatic speaker verification spoofing and countermeasures challenge (asvspoof 2017) database, version 2,

    T. Kinnunen, Sahidullah, Md, H. Delgado, M. Todisco, N. Evans, J. Yamagishi, and K. A. Lee, “The 2nd automatic speaker verification spoofing and countermeasures challenge (asvspoof 2017) database, version 2,” 2018. [Online]. Available: https://datashare.ed.ac.uk/handle/10283/3055

    work page 2017

  7. [7]

    ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection,

    M. Todisco, X. Wang, M. Sahidullah, H. Delgado, A. Nautsch, J. Yam- agishi, N. Evans, T. Kinnunen, and K. A. Lee, “ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection,” in Proc. of Interspeech 2019, 2019

    work page 2019

  8. [8]

    Asvspoof 2019: The 3rd automatic speaker verification spoofing and countermeasures challenge database,

    J. Yamagishi, M. Todisco, Sahidullah, Md, H. Delgado, X. Wang, N. Evans, T. Kinnunen, K. A. Lee, V . Vestman, and A. Nautsch, “Asvspoof 2019: The 3rd automatic speaker verification spoofing and countermeasures challenge database,” 2019. [Online]. Available: https://datashare.ed.ac.uk/handle/10283/3336

    work page 2019

Show all 52 references
  1. [9]

    ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild,

    X. Liu, X. Wang, M. Sahidullah, J. Patino, H. Delgado, T. Kinnunen, M. Todisco, J. Yamagishi, N. Evans, A. Nautsch, and K. A. Lee, “ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild,” 2022. [Online]. Available: https://arxiv.org/abs/2210.02437

    2021 arXiv

  2. [10]

    Asvspoof 2021 challenge - logical access database,

    H. Delgado, N. Evans, T. Kinnunen, K. A. Lee, X. Liu, A. Nautsch, J. Patino, M. Sahidullah, M. Todisco, X. Wang, and J. Yamagishi, “Asvspoof 2021 challenge - logical access database,” 2021. [Online]. Available: https://zenodo.org/record/4837263

    2021

  3. [11]

    Asvspoof 2021 challenge - physical access database,

    ——, “Asvspoof 2021 challenge - physical access database,” 2021. [Online]. Available: https://zenodo.org/record/4834716

    2021

  4. [12]

    Asvspoof 2021 challenge - speech deepfake database,

    ——, “Asvspoof 2021 challenge - speech deepfake database,” 2021. [Online]. Available: https://zenodo.org/record/4835108

    2021

  5. [13]

    Asvspoof 5: Crowdsourced speech data, deepfakes, and adversarial attacks at scale,

    X. Wang, H. Delgado, H. Tak, J. weon Jung, H. jin Shim, M. Todisco, I. Kukanov, X. Liu, M. Sahidullah, T. Kinnunen, N. Evans, K. A. Lee, and J. Yamagishi, “Asvspoof 5: Crowdsourced speech data, deepfakes, and adversarial attacks at scale,” 2024. [Online]. Available: https://ar...

    2024 arXiv

  6. [14]

    Asvspoof 5: Design, collection and validation of resources for spoofing, deepfake, and adversarial attack detection using crowdsourced speech,

    X. Wang, H. Delgado, H. Tak, J.-w. Jung, H. Shim, M. Todisco, I. Kukanov, X. Liu, M. Sahidullah, T. Kinnunen, N. Evans, K. A. Lee, J. Yamagishi, M. Jeong, G. Zhu, Y . Zang, N. Zhang, S. Maiti, F. Lux, N. Muller, W. Zhang, C. Sun, S. Hou, S. Lyu, S. Le Maguer, C. Gong, H. Guo, ...

    2024

  7. [15]

    Add 2022: the first audio deep synthesis detection challenge,

    J. Yi, R. Fu, J. Tao, S. Nie, H. Ma, C. Wang, T. Wang, Z. Tian, X. Zhang, Y . Bai, C. Fan, S. Liang, S. Wang, S. Zhang, X. Yan, L. Xu, Z. Wen, H. Li, Z. Lian, and B. Liu, “Add 2022: the first audio deep synthesis detection challenge,” 2024

    2022

  8. [16]

    Does audio deepfake detection generalize?

    N. M. M ¨uller, P. Czempin, F. Dieckmann, A. Froghyar, and K. B¨ottinger, “Does audio deepfake detection generalize?” 2022

    2022

  9. [17]

    Dataset artefacts in anti- spoofing systems: A case study on the asvspoof 2017 benchmark,

    B. Chettri, E. Benetos, and B. L. T. Sturm, “Dataset artefacts in anti- spoofing systems: A case study on the asvspoof 2017 benchmark,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 28, pp. 3018–3028, 2020

    2017

  10. [18]

    The effect of silence and dual-band fusion in anti-spoofing system,

    Y . Zhang, W. Wang, and P. Zhang, “The effect of silence and dual-band fusion in anti-spoofing system,” in Interspeech 2021, 2021, pp. 4279– 4283

    2021

  11. [19]

    Speech is silver, silence is golden: What do asvspoof- trained models really learn?

    N. M. M ¨uller, F. Dieckmann, P. Czempin, R. Canals, K. B ¨ottinger, and J. Williams, “Speech is silver, silence is golden: What do asvspoof- trained models really learn?” 2021

    2021

  12. [20]

    Investigating self-supervised front ends for speech spoofing countermeasures,

    X. Wang and J. Yamagishi, “Investigating self-supervised front ends for speech spoofing countermeasures,” 2022

    2022

  13. [21]

    Does audio deepfake detection generalize?

    N. M. M ¨uller, P. Czempin, F. Dieckmann, A. Froghyar, and K. B¨ottinger, “Does audio deepfake detection generalize?” Interspeech, 2022

    2022

  14. [22]

    The m-ailabs speech dataset,

    “The m-ailabs speech dataset,” github.com/imdatceleste/m-ailabs- dataset, accessed: 2024-10-11

    2024

  15. [23]

    Mlaad: The multi- language audio anti-spoofing dataset,

    N. M. M ¨uller, P. Kawa, W. H. Choong, E. Casanova, E. G ¨olge, T. M ¨uller, P. Syga, P. Sperl, and K. B ¨ottinger, “Mlaad: The multi- language audio anti-spoofing dataset,” International Joint Conference on Neural Networks (IJCNN) , 2024

    2024

  16. [24]

    Deepfake de- tection challenge,

    benpflaum, B. G, djdj, I. Kofman, J. Tester, JLElliott, J. Metherd, J. Elliott, Mozaic, P. Culliton, S. Dane, and W. Kim, “Deepfake de- tection challenge,” https://kaggle.com/competitions/deepfake-detection- challenge, 2019, kaggle

    2019

  17. [25]

    FakeA VCeleb: A novel audio-video multimodal deepfake dataset,

    H. Khalid, S. Tariq, M. Kim, and S. S. Woo, “FakeA VCeleb: A novel audio-video multimodal deepfake dataset,” in Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2) , 2021. [Online]. Available: https://openreview.net/forum?id...

    2021

  18. [26]

    Dfdc video audio labels,

    “Dfdc video audio labels,” www.kaggle.com/datasets/basharallabadi/dfdc- video-audio-labels, accessed: 2024-06

    2024

  19. [27]

    SPEECON – speech databases for consumer devices: Database specification and validation,

    D. Iskra, B. Grosskopf, K. Marasek, H. van den Heuvel, F. Diehl, and A. Kiessling, “SPEECON – speech databases for consumer devices: Database specification and validation,” in Proceedings of the Third International Conference on Language Resources and Evaluation (LREC‘02). Las...

    2002

  20. [28]

    Silero vad: pre-trained enterprise-grade voice activity detector (vad), number detector and language classifier,

    S. Team, “Silero vad: pre-trained enterprise-grade voice activity detector (vad), number detector and language classifier,” https://github.com/snakers4/silero-vad, 2024

    2024

  21. [29]

    Spoofed training data for speech spoofing countermeasure can be efficiently created using neural vocoders,

    X. Wang and J. Yamagishi, “Spoofed training data for speech spoofing countermeasure can be efficiently created using neural vocoders,” 2023

    2023

  22. [30]

    Hugging face,

    “Hugging face,” https://huggingface.co, accessed: 2024-10

    2024

  23. [31]

    Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis,

    J. Kong, J. Kim, and J. Bae, “Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis,” 2020

    2020

  24. [32]

    Melgan: Generative adversarial networks for conditional waveform synthesis,

    K. Kumar, R. Kumar, T. de Boissiere, L. Gestin, W. Z. Teoh, J. Sotelo, A. de Brebisson, Y . Bengio, and A. Courville, “Melgan: Generative adversarial networks for conditional waveform synthesis,” 2019

    2019

  25. [33]

    World: A vocoder- based high-quality speech synthesis system for real-time applications,

    M. MORISE, F. YOKOMORI, and K. OZAW A, “World: A vocoder- based high-quality speech synthesis system for real-time applications,” IEICE Transactions on Information and Systems , vol. E99.D, no. 7, pp. 1877–1884, 2016

    2016

  26. [34]

    Waveglow: A flow-based generative network for speech synthesis,

    R. Prenger, R. Valle, and B. Catanzaro, “Waveglow: A flow-based generative network for speech synthesis,” 2018

    2018

  27. [35]

    Neural Source-Filter Wave- form Models for Statistical Parametric Speech Synthesis,

    X. Wang, S. Takaki, and J. Yamagishi, “Neural Source-Filter Wave- form Models for Statistical Parametric Speech Synthesis,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 28, pp. 402–415, 2020

    2020

  28. [36]

    Can large-scale vocoded spoofed data improve speech spoofing countermeasure with a self-supervised front end?

    X. Wang and J. Yamagishi, “Can large-scale vocoded spoofed data improve speech spoofing countermeasure with a self-supervised front end?” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024, pp. 10 311–10 315

    2024

  29. [37]

    The lj speech dataset,

    K. Ito and L. Johnson, “The lj speech dataset,” https://keithito.com/LJ- Speech-Dataset/, 2017

    2017

  30. [38]

    fairseq: A fast, extensible toolkit for sequence modeling,

    M. Ott, S. Edunov, A. Baevski, A. Fan, S. Gross, N. Ng, D. Grangier, and M. Auli, “fairseq: A fast, extensible toolkit for sequence modeling,” in Proceedings of NAACL-HLT 2019: Demonstrations , 2019

    2019

  31. [39]

    Stc antispoofing systems for the asvspoof2021 challenge,

    A. Tomilov, A. Svishchev, M. V olkova, A. Chirkovskiy, A. Kondratev, and G. Lavrentyeva, “Stc antispoofing systems for the asvspoof2021 challenge,” in 2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge , 2021, pp. 61–67

    2021

  32. [40]

    Averaging weights leads to wider optima and better generalization,

    P. Izmailov, D. Podoprikhin, T. Garipov, D. Vetrov, and A. G. Wilson, “Averaging weights leads to wider optima and better generalization,” 2019

    2019

  33. [41]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” 2018

    2018

  34. [42]

    A discriminative feature learning approach for deep face recognition

    Y . Wen, K. Zhang, Z. Li, and Y . Qiao, “A discriminative feature learning approach for deep face recognition.” in ECCV, ser. Lecture Notes in Computer Science, vol. 9911. Springer, 2016, pp. 499–515

    2016

  35. [43]

    One-class learning towards synthetic voice spoofing detection,

    Y . Zhang, F. Jiang, and Z. Duan, “One-class learning towards synthetic voice spoofing detection,” IEEE Signal Processing Letters , vol. 28, p. 937–941, 2021

    2021

  36. [44]

    Asvspoof 2021: Towards spoofed and deepfake speech detection in the wild,

    X. Liu, X. Wang, M. Sahidullah, J. Patino, H. Delgado, T. Kinnunen, M. Todisco, J. Yamagishi, N. Evans, A. Nautsch, and K. A. Lee, “Asvspoof 2021: Towards spoofed and deepfake speech detection in the wild,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. ...

    2021

  37. [45]

    Rawboost: A raw data boosting and augmentation method applied to automatic speaker verification anti-spoofing,

    H. Tak, M. Kamble, J. Patino, M. Todisco, and N. Evans, “Rawboost: A raw data boosting and augmentation method applied to automatic speaker verification anti-spoofing,” 2022

    2022

  38. [46]

    A binaural room impulse response database for the evaluation of dereverberation algorithms,

    M. Jeub, M. Sch ¨afer, and P. Vary, “A binaural room impulse response database for the evaluation of dereverberation algorithms,” in Proceed- ings of International Conference on Digital Signal Processing (DSP) . IEEE, Jul. 2009, pp. 1–4

    2009

  39. [47]

    Ustc-kxdigit system description for asvspoof5 challenge,

    Y . Chen, H. Wu, N. Jiang, X. Xia, Q. Gu, Y . Hao, P. Cai, Y . Guan, J. Wang, W. Xie, L. Fang, S. Fang, Y . Song, W. Guo, L. Liu, and M. Xu, “Ustc-kxdigit system description for asvspoof5 challenge,”

  40. [48]

    High fidelity neural audio compression,

    A. D ´efossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,” 2022. [Online]. Available: https://arxiv.org/abs/2210.13438

    2022 arXiv

  41. [49]

    Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods,

    J. Platt et al. , “Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods,” Advances in large margin classifiers, vol. 10, no. 3, pp. 61–74, 1999

    1999

  42. [50]

    Towards measuring fairness in speech recognition: Fair-speech dataset,

    I.-E. Veliche, Z. Huang, V . A. Kochaniyan, F. Peng, O. Kalinli, and M. L. Seltzer, “Towards measuring fairness in speech recognition: Fair-speech dataset,” 2024

    2024

  43. [51]

    Torchaudio-squim: Reference-less speech quality and intelligi- bility measures in torchaudio,

    A. Kumar, K. Tan, Z. Ni, P. Manocha, X. Zhang, E. Henderson, and B. Xu, “Torchaudio-squim: Reference-less speech quality and intelligi- bility measures in torchaudio,” 2023

    2023

  44. [2024]

    Available: https://arxiv.org/abs/2409.01695

    [Online]. Available: https://arxiv.org/abs/2409.01695

    arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.