The DOI matches its Crossref record; 37 of the 37 references it cites were found in a registry or an index; 14 of the 18 sources its claims cite were checked for support.
Voice Deepfake Detection Using the Self-Supervised Pre-Training Model HuBERT
Open accessAbstract
In recent years, voice deepfake technology has developed rapidly, but current detection methods have the problems of insufficient detection generalization and insufficient feature extraction for unknown attacks. This paper presents a forged speech detection method (HuRawNet2_modified) based on a self-supervised pre-trained model (HuBERT) to improve detection (and address the above problems). A combination of impulsive signal-dependent additive noise and additive white Gaussian noise was adopted for data boosting and augmentation, and the HuBERT model was fine-tuned on different language databases. On this basis, the size of the extracted feature maps was modified independently by the α-feature map scaling (α-FMS) method, with a modified end-to-end method using the RawNet2 model as the backbone structure. The results showed that the HuBERT model could extract features more comprehensively and accurately. The best evaluation indicators were an equal error rate (EER) of 2.89% and a minimum tandem detection cost function (min t-DCF) of 0.2182 on the database of the ASVspoof2021 LA challenge, which verified the effectiveness of the detection method proposed in this paper. Compared with the baseline systems in databases of the ASVspoof 2021 LA challenge and the FMFCC-A, the values of EER and min t-DCF decreased. The results also showed that the self-supervised pre-trained model with fine-tuning can extract acoustic features across languages. And the detection can be slightly improved when the languages of the pre-trained database, and the fine-tuned and tested database are the same.
Overview
CoChat Summary
## TL;DR
This paper proposes HuRawNet2_modified, a voice deepfake detector that combines HuBERT self-supervised pre-training/fine-tuning with an improved RawNet2 back end and data augmentation. It reports best results of EER 2.89% and min t-DCF 0.2182 on ASVspoof 2021 LA, and finds that cross-language pre-training can work, with slightly better results when pre-training and testing languages match.
## Key findings
- On the ASVspoof 2021 LA evaluation set, HuRawNet2_modified (LibriSpeech) achieved EER 2.89% and min t-DCF 0.2182, outperforming the listed baselines: CQCC-GMM (15.62% / 0.4974), LFCC-GMM (19.30% / 0.5758), LFCC-LCNN (9.26% / 0.3445), and Raw audio-RawNet2 (9.50% / 0.4257).
- Compared with the RawNet2 baseline on ASVspoof 2021 LA, the paper says EER and min t-DCF were reduced by 69.5% and 48.7%, respectively.
- On FMFCC-A and FAD, the model improved over baselines: EER was reduced by 55.3% and 60.7%, and Log-loss by 42.2% and 61.7%, respectively.
- Table 5 shows cross-language pre-training/fine-tuning results:
- ASVspoof 2021 LA: 3.01% / 0.2278 (WenetSpeech) vs 2.89% / 0.2182 (LibriSpeech)
- ASVspoof 2019 LA: 2.12% / 0.1442 (WenetSpeech) vs 1.96% / 0.1393 (LibriSpeech)
- FMFCC-A: 3.25% / 0.3121 (WenetSpeech) vs 3.37% / 0.3378 (LibriSpeech)
- FAD: 8.31% / 0.1730 (WenetSpeech) vs 9.75% / 0.2292 (LibriSpeech)
## Methods
- The proposed system, HuRawNet2_modified, takes raw waveform input and performs binary classification into real vs. fake speech.
- Front end: HuBERT self-supervised pre-trained speech models trained on WenetSpeech (Chinese) and LibriSpeech (English) were fine-tuned on labeled datasets.
- Back end: an improved RawNet2 was used, with α-FMS in residual blocks and the number of residual blocks reduced from 6 to 3 to improve efficiency.
- Data and evaluation: experiments used ASVspoof 2021 LA, ASVspoof 2019 LA, FMFCC-A, and FAD; metrics were EER, min t-DCF, and Log-loss for FMFCC-A.
- Training details stated: audio was cut or copied to about 4 seconds, training used 40 epochs, Adam, batch size 14, and learning rate 0.000001 for the self-supervised front end.
## Limitations
- The paper states that fine-tuning requires a large amount of GPU memory.
- The paper also notes that although cross-language transfer works, performance is slightly better when the pre-trained, fine-tuned, and tested datasets are in the same language.
- Evident from the text: some reported results are presented only as selected tables and comparisons; the paper does not provide full detail for all baseline results on every dataset, so the comparison set is limited to what is shown.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
10 Valid · 3 Not valid · 5 Not checked · 28 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Current research shows that self-supervised speech models can extract robust acoustic features for unknown domains.”
What the sources say
- Partly supportedcites [9] Hsu, 2021, HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units, IEEEACM Trans. Audio Speech Lang. Process. sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) says self-supervised learning can produce universal speech representations and that its features generalize to multiple downstream applications. It also identifies new domains as a motivation, but the cited passages do not establish that the features are robust in unknown domains. (LLM, conf=0.82)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The training and development sets for the ASVspoof 2021 LA dataset are the same as those released for the ASVspoof 2019 challenge.”
What the sources say
- Supportedcites [7] Yamagishi, J., Wang, X., Todisco, M., Sahidullah, M., Patino, J., Nautsch, A., Liu, X., Lee, K.A., Kinnunen, T., and Evans, N. (2021, January 16). ASVspoof 2021: Accelerating Progress in Spoofed and Deepfake Speech Detection. Proceedings of the 2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge, Online. source
The Database section (p. 1) says no new training or development data were released for ASVspoof 2021 and that participants were expected to use the ASVspoof 2019 training and development partitions. This supports the claim for the LA task. (LLM, conf=0.95)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The results show that the performance of the HuBERT model is better than that of the wav2vec 2.0 model [9].”
What the sources say
- Partly supportedcites [9] Hsu, 2021, HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units, IEEEACM Trans. Audio Speech Lang. Process. sourceExperimental — accuracy not yet measured
The Introduction says HuBERT “either matches or improves upon” wav2vec 2.0 across the tested fine-tuning subsets. The Main Results section also says HuBERT is “on par” with two leading wav2vec 2.0-based pre-training results. Thus, the source supports comparable or sometimes better performance, but not an unqualified claim that HuBERT performs better. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“And the detection can be slightly improved when the languages of the pre-trained database, and the fine-tuned and tested database are the same.”
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The results showed that the HuBERT model could extract features more comprehensively and accurately.”
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
18 more claims
Create a free account to see all 46 claims and every source verdict.
- It may state two separate claims as one. 19
- Probably not a substantial claim. 9
- The quote only partly states this claim. 1
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
Distributed Representations of Words and Phrases and their Compositionality
Mikolov, T., Sutskever, I., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1310.4546 · Cited by 24,533
Representation Learning with Contrastive Predictive Coding
Oord, A. V. D., Li, Y., & Vinyals, O. (2018) · arXiv · doi:10.48550/arxiv.1807.03748 · Cited by 6,242
SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
Park, D. S., Chan, W., Zhang, Y., Chiu, C.-C., Zoph, B., Cubuk, E. D., & Le, Q. V. (2019) · Interspeech 2019 · doi:10.21437/interspeech.2019-2680 · Cited by 3,558
HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units
Hsu, W.-N., Bolte, B., Tsai, Y.-H. H., Lakhotia, K., Salakhutdinov, R., & Mohamed, A. (2021) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2021.3122291 · Cited by 2,983
wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations
Baevski, A., Zhou, H., Mohamed, A., & Auli, M. (2020) · arXiv · doi:10.48550/arxiv.2006.11477 · Cited by 2,369
WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing
Chen, S., Wang, C., Chen, Z., Wu, Y., Liu, S., Chen, Z., Li, J., Kanda, N., Yoshioka, T., Xiao, X., et al. (2022) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2022.3188113 · Cited by 1,926
Desplanques, B., Thienpondt, J., & Demuynck, K. (2020) · Interspeech 2020 · doi:10.21437/interspeech.2020-2650 · Cited by 1,568
wav2vec: Unsupervised Pre-Training for Speech Recognition
Schneider, S., Baevski, A., Collobert, R., & Auli, M. (2019) · Interspeech 2019 · doi:10.21437/interspeech.2019-1873 · Cited by 1,290
SpeechBrain: A General-Purpose Speech Toolkit
Ravanelli, M., Parcollet, T., Plantinga, P., Rouhe, A., Cornell, S., Lugosch, L., Subakan, C., Dawalatabad, N., Heba, A., Zhong, J., et al. (2021) · arXiv · doi:10.48550/arxiv.2106.04624 · Cited by 512
Research works printed but not linked: 1
- Wang, X., and Yamagishi, J. (2022). Frontiers in Fake Media Generation and Detection, Springer Nature.
Cited by
A Survey on Speech Deepfake Detection
Li, M., Ahmadiadli, Y., & Zhang, X.-P. (2025) · ACM Computing Surveys · doi:10.1145/3714458 · Cited by 101
Deepfake audio detection with spectral features and ResNeXt-based architecture
Tahaoglu, G., Baracchi, D., Shullani, D., Iuliani, M., & Piva, A. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113726 · Cited by 24
Anomaly Detection of Deepfake Audio Based on Real Audio Using Generative Adversarial Network Model
Song, D., Lee, N., Kim, J., & Choi, E. (2024) · IEEE Access · doi:10.1109/access.2024.3506973 · Cited by 28
Wani, T. M., Qadri, S. A. A., Wani, F. A., & Amerini, I. (2024) · Foundations and Trends® in Privacy and Security · doi:10.1561/3300000048 · Cited by 16
Leveraging Mixture of Experts for Improved Speech Deepfake Detection
Negroni, V., Salvi, D., Mezza, A. I., Bestagini, P., & Tubaro, S. (2025) · ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp49660.2025.10890398 · Cited by 14
Metadata
- Authors
- Li, L., Lu, T., Ma, X., Yuan, M., & Wan, D.
- Affiliations
- China People's Public Security University (CN)
- Corresponding author
- Lu, Tianliang
- Venue
- Applied Sciences
- Publisher
- MDPI AG
- Published
- 2023-07-22
- Volume
- 13
- Issue
- 14
- Pages
- 8488
- ISSN
- 2076-3417 (ISSN-L)
- OpenAlex
- W4385224813
- Cited by
- 33
- References
- 31
- Citations by year
- 2026: 9 · 2025: 15 · 2024: 9
- FWCI
- 3.10
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Applied Sciences (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Double First-Class Innovation Research Project for People’s Public Security University of China (2023SYL07)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Keywords
- speech spoofing detection, HuBERT, self-supervised pre-training, data augmentation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-30, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-30 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Voice Deepfake Detection Using the Self-Supervised Pre-Training Model HuBERT. https://app.cochat.ai/papers/10.3390/app13148488
- The paper
- Li, L., Lu, T., Ma, X., Yuan, M., & Wan, D. (2023). Voice Deepfake Detection Using the Self-Supervised Pre-Training Model HuBERT. Applied Sciences. https://doi.org/10.3390/app13148488
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.