Pith. sign in

REVIEW 4 major objections 4 minor 29 references

Generalizable Audio Deepfake Detection via Latent Space Refinement and Augmentation

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Adding learnable spoof prototypes and five embedding perturbations to a Wav2Vec2.0/AASIST detector yields equal error rates as low as 0.12% across four deepfake speech benchmarks, with no change to the model architecture.

desk verdict A plausible, cleanly ablated combination of multi-prototype refinement and latent-space augmentation that deserves refereeing, but the headline gains are confounded by an uncontrolled extra augmentation and single-run results. read the letter →

arxiv 2501.14240 v1 pith:CNQM5YJ4 submitted 2025-01-24 eess.AS cs.SD

Wen Huang , Yanmei Gu , Zhiming Wang , Huijia Zhu , Yanmin Qian This is my paper · ORCID
classification eess.AScs.SD
keywords audiodeepfakedetectionspoofingcountermeasuresgeneralizationtounseenattackslatentspaceaugmentationrefinementlearnableprototypesWav2Vec2.0XLSRAASIST
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

This paper claims that a spoofing detector can generalize to unseen deepfake attacks without any change to its architecture, by reshaping the feature space during training. The proposed strategy has two parts: Latent Space Refinement replaces the single 'spoof' class cluster with several learnable prototype vectors, and Latent Space Augmentation applies five perturbation operators exclusively to spoof embeddings so the model sees a wider range of fake patterns. Trained on the ASVspoof 2019 LA training partition, the best configuration (which also uses extra data augmentation) reports equal error rates of 0.12% on that set's evaluation partition, 1.05% on ASVspoof 2021 LA, 1.86% on ASVspoof 2021 DF, and 5.54% on In-The-Wild, matching or surpassing published systems. The practical stake is a drop-in training-time improvement for Wav2Vec2.0/AASIST-style detectors.

What carries the argument

The load-bearing object is a prototype-based latent space: $K$ learnable prototype vectors on the unit hypersphere for the spoof class (one for bonafide) with a smoothed max-cosine similarity scoring rule and an additive angular margin loss. Around it sit five augmentation operators applied only to spoof embeddings: Gaussian additive noise, affine scaling, batch mixup, linear interpolation toward the bonafide prototype, and linear extrapolation from the nearest spoof prototype, with norm-correction terms to move between cosine and Euclidean geometry. Augmented embeddings are concatenated with the originals and feed the same weighted cross-entropy plus prototype loss, while intra-class and inter-class regularization terms prevent the spoof prototypes from collapsing together or toward the bonafide center.

What would settle it

Train the same Wav2Vec2.0/AASIST system twice, once with LSR+LSA and once with the same five operators applied to randomly chosen embeddings of both classes under identical perturbation budgets, then compare equal error rates on a held-out family of recent TTS and voice-conversion attacks; if the gains do not depend on restricting augmentation to spoofs and on prototype-guided interpolation and extrapolation, the manifold-staying explanation fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the generalization bottleneck in audio deepfake detection is the oversimplified representation of the spoof class: a binary classifier with one prototype per class cannot capture the variety of synthesis artifacts, and a one-class boundary is too rigid. Modeling the spoof class with multiple learnable prototypes, kept diverse by an intra-class regularization term and separated from the bonafide prototype by an inter-class term, refines the latent space; perturbing only spoof embeddings with additive noise, affine transforms, batch mixup, linear interpolation toward the bonafide prototype, and linear extrapolation from the nearest spoof prototype then diversifies the spoof distribution. The ablation study isolates the active ingredients: intra-class prototype regularization and linear extrapolation are the strongest individual contributors, and combining all five augmentations gives the best results. The paper concludes that the integrated LSR+LSA training objective is what lets a single detector trained on ASVspoof 2019 LA reach the reported equal error rates across four benchmarks.

Load-bearing premise

The load-bearing premise is that the perturbations applied to spoof embeddings in latent space produce synthetic features that stay near the real distribution of spoofed speech, so that training on them transfers to unseen attacks instead of merely fitting noise.

Editorial extensions

If this is right

  • The method is orthogonal to architecture changes, so its gains should stack with newer front-ends and back-ends, not just Wav2Vec2.0/AASIST.
  • Because the strongest individual augmentation is linear extrapolation from the nearest spoof prototype, pushing spoof embeddings into unoccupied latent regions is the most productive form of diversity.
  • Latent-space augmentation beats equivalent input-space augmentation for noise and affine transforms, suggesting that low-dimensional feature manipulation is more efficient than raw waveform or spectrogram perturbation for these operators.
  • Combining all five augmentations performs best, which implies that the model benefits from a broad, heterogeneous expansion of the spoof region rather than any single perturbation family.
  • Prototype count matters: performance improves up to roughly eight spoof prototypes and degrades beyond, giving a tuning handle that is independent of the classifier architecture.

Reading between the lines

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

  • A testable extension is to make the number of spoof prototypes adaptive per attack family, using clustering or a Dirichlet process, since the paper's fixed-$K$ experiments show an optimum around eight prototypes.
  • The same multi-prototype plus negative-only augmentation recipe could transfer to other open-set audio tasks, such as speaker verification with unknown impostors or synthetic-voice based fraud detection, where the negative class is highly variable.
  • Because the augmentations preserve the label of the spoof embedding, the method implicitly bets that the spoof distribution is locally connected; one could stress-test this by applying the same operators to random regions of the latent space to verify the gains are tied to the spoof region.
  • The paper leaves open whether interpolation or extrapolation could also be applied to bonafide embeddings; if that helped, the gains would be about decision-surface shaping rather than spoof diversity alone.
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 / 4 minor

Summary. The paper proposes two training-time modifications for audio deepfake detectors built on Wav2Vec2.0-XLSR and AASIST: Latent Space Refinement (LSR), which replaces the single spoof prototype with K learnable prototypes and a prototype-based loss with intra- and inter-class regularizers, and Latent Space Augmentation (LSA), which applies five types of perturbations (additive noise, affine, batch mixup, linear interpolation, linear extrapolation) to spoof embeddings only. The method is evaluated by training on ASVspoof 2019 LA and testing on ASVspoof 2019 LA, ASVspoof 2021 LA/DF, and In-The-Wild, reporting EERs as low as 0.12%, 1.05%, 1.86%, and 5.54% respectively when an unspecified 'extra data augmentation' is added. The paper claims the integrated method matches or surpasses state-of-the-art countermeasures.

Significance. If the reported results are attributable to LSR and LSA and are reproducible, the method is a useful, architecture-agnostic training-time contribution: it requires no change at inference time and can be combined with other frontends. The internal ablations in Tables II and III consistently show that LSR and LSA each improve over the WCE baseline on average, and the t-SNE visualization in Fig. 2 provides intuitive support. However, the central empirical claim currently suffers from an uncontrolled augmentation comparison, missing variance estimates, and hyperparameter selection on test benchmarks, which makes the magnitude and even the direction of some individual gains uncertain.

major comments (4)
  1. [Section III-A, Table I] The 'extra data augmentation' in Table I is uncontrolled. Section III-A states that 'Rawboost is applied as basic augmentation and codec augmentation as extra augmentation,' but only the final row of Table I is marked with a star, and Tables II-IV never state which augmentation stack is used for any row. Consequently, the headline row (0.12/1.05/1.86/5.54) may confound the effect of LSR+LSA with the effect of the extra augmentation, and the comparison against published numbers is not recipe-matched. The authors should specify the augmentation configuration for every experiment and provide an ablation that isolates the contribution of LSR+LSA with and without the extra augmentation.
  2. [Tables II-IV] All results are single point estimates without error bars or significance tests. This matters because some individual comparisons are inconsistent with the average trend. For example, in Table II, on 21LA the LSR-only system (1.55% EER) is better than WCE+LSR (2.35% EER), even though WCE+LSR is better on average. Without repeated seeds, it is impossible to tell whether such differences reflect real improvements or run-to-run variance. The authors should report mean and standard deviation over at least three seeds, and ideally a paired significance test.
  3. [Section III-C, Fig. 3] The number of spoof prototypes K is selected using the evaluation sets. Section III-C and Fig. 3 report EER on 21LA, 21DF, and ITW as a function of K and conclude that K=8 is best and K=20 degrades performance. Since these are test benchmarks, choosing K based on them and then reporting results on the same benchmarks introduces optimism and makes the reported numbers not fully hold-out. The authors should either choose K on a validation split (e.g., the ASVspoof 2019 LA development set) or explicitly acknowledge that the reported numbers include test-set-based hyperparameter selection.
  4. [Section III-B, Table I] The claim of 'matching or surpassing current state-of-the-art methods' is not supported on all datasets. In Table I, the proposed LSR+LSA without extra augmentation has 1.19% EER on 21LA, which is worse than the 0.82% EER reported for Wav2Vec+AASIST [6]; even the starred row (1.05%) does not surpass that value. The comparison should be restricted to settings with matched training recipes, or the claim should be softened to indicate competitive but not uniformly superior performance.
minor comments (4)
  1. [Eq. (1)] The softmax-like smoothing in Eq. (1) is written with a denominator summing over j, but the numerator uses index i; please clarify that the denominator is the sum over all i of exp(<c_i,z>·gamma), or rewrite the expression with explicit index notation.
  2. [Section II-B, Eq. (6)] The additive noise scale beta is said to be 'sampled from N(0,1)', which allows negative values; if negative scaling is intended, please state so, otherwise use a positive distribution and clarify the role of beta.
  3. [Section III-A] The relationship between 'basic augmentation' (Rawboost) and 'extra augmentation' (codec) is described only in one sentence; please state clearly which of these are used in each table, including the baseline rows.
  4. [Fig. 3] The curve for K=20 shows performance decline, but there is no table or confidence interval; adding variance estimates or at least a table of the underlying EER values would make the claim more verifiable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reported EERs are empirical benchmark results, not quantities forced by the paper's equations or by a self-citation chain.

full rationale

This paper is an empirical method paper. The proposed LSR objective (Eq. 5) and the LSA transforms (Eqs. 6-10) define training losses and data-augmentation operations; they do not algebraically determine the equal-error rates reported in Tables I-IV. The reported EERs are measured on standard evaluation sets after training, so the central claim that the integration achieves competitive results is not derived from the method's own equations by construction. No parameter is fitted to a subset of data and then renamed as a prediction: the prototype count and the augmentation combination are model-selection choices, although if they were selected after inspecting the same benchmark results (Fig. 3, Table III), that would be a statistical selection-bias concern rather than a circular reduction. The cited prior work (Wav2Vec2.0, AASIST, RawBoost, ArcFace, and CV latent-space augmentation) is external to the authors and is used as building blocks, not as a self-citation chain that forces the conclusion. The main legitimate concerns are attribution and reproducibility: Table I's starred row includes extra data augmentation, the augmentation stack is not fully specified for all tables, and no error bars are reported. These are not circularity under the stated criteria.

Assumptions & free parameters 11 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard pretrained components, a set of hand-chosen hyperparameters, and the domain assumption that latent-space perturbations are semantically valid. No new theoretical entities or physical objects are introduced. The largest burden is the latent-manifold assumption, which is only supported by the reported EER numbers.

free parameters (11)
  • Number of spoof prototypes K = 8
    Chosen based on Fig. 3; performance peaks near 8 and declines at 20, suggesting selection on evaluation results.
  • Softmax temperature gamma = 10
    Set in Eq. 1; controls the sharpness of the smoothed maximum over prototype similarities.
  • Prototype loss scale s = 32
    Set in Section III-A; standard ArcFace-style scaling factor.
  • Angular margin m = 0.2
    Set in Section III-A; controls the margin in the prototype-based classification loss.
  • Inter-class regularization coefficient delta = 0.2
    Set in Section III-A; prevents the inter-class regularization loss from becoming negative.
  • Additive noise scaling distribution = beta sampled from N(0,1)
    Eq. 6; the noise magnitude is drawn from a standard normal distribution, a choice not justified beyond empirical performance.
  • Affine transformation scale range = a ~ U(0.9, 1.1)
    Eq. 7; range chosen by hand for the latent-space scaling operation.
  • Mixup coefficient distribution = alpha ~ Beta(0.5, 0.5)
    Eq. 8; standard mixup parameter choice.
  • Interpolation and extrapolation coefficient ranges = lambda_i, lambda_e ~ U(0, 0.1)
    Eqs. 9 and 10; small ranges chosen to keep augmented features near the original embeddings.
  • Weighted cross-entropy class weights = 0.9 (bonafide), 0.1 (spoof)
    Set in Section III-A; used to handle class imbalance in the combined WCE+LSR loss.
  • Learning rates = 1e-6 backbone, 1e-3 prototypes
    Set in Section III-A; not justified by a validation study.
assumptions (4)
  • domain assumption Pretrained Wav2Vec2.0 XLSR features are suitable and domain-agnostic inputs for anti-spoofing.
    Adopted from [3] and [6]; if the features are not sufficiently invariant, the prototype and augmentation geometry may not transfer across datasets. Invoked in Sections I and III-A.
  • domain assumption AASIST is a strong backend classifier for this task.
    Adopted from [20]; the capacity and inductive bias of the backend affect whether latent-space manipulations yield useful training signal. Invoked in Section III-A.
  • domain assumption Latent-space perturbations preserve spoof semantics.
    This is the core working hypothesis of LSA: adding noise, scaling, mixing, and interpolation in the embedding space creates realistic spoof variations. The paper provides empirical support but no formal justification. Invoked in Section II-B.
  • domain assumption The six spoofing attacks in the ASVspoof 2019 LA training set are diverse enough to learn generalizable prototypes.
    Training uses only the ASVspoof 2019 LA training set; the claim of generalization to unseen attacks depends on this diversity assumption. Invoked in Section III-A.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generalizable Audio Deepfake Detection via Latent Space Refinement and Augmentation." pith.science (2026). https://pith.science/paper/CNQM5YJ4

@misc{pith2026250114240,
  author       = {Pith},
  title        = {Pith review of: Generalizable Audio Deepfake Detection via Latent Space Refinement and Augmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CNQM5YJ4}},
  note         = {Machine review of arXiv:2501.14240}
}
read the original abstract

Advances in speech synthesis technologies, like text-to-speech (TTS) and voice conversion (VC), have made detecting deepfake speech increasingly challenging. Spoofing countermeasures often struggle to generalize effectively, particularly when faced with unseen attacks. To address this, we propose a novel strategy that integrates Latent Space Refinement (LSR) and Latent Space Augmentation (LSA) to improve the generalization of deepfake detection systems. LSR introduces multiple learnable prototypes for the spoof class, refining the latent space to better capture the intricate variations within spoofed data. LSA further diversifies spoofed data representations by applying augmentation techniques directly in the latent space, enabling the model to learn a broader range of spoofing patterns. We evaluated our approach on four representative datasets, i.e. ASVspoof 2019 LA, ASVspoof 2021 LA and DF, and In-The-Wild. The results show that LSR and LSA perform well individually, and their integration achieves competitive results, matching or surpassing current state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2501.14240 by the authors.

Figure 1
Figure 1. The pipeline of the proposed method, illustrating the process of Latent Space Refinement (LSR) and Latent Space Augmentation (LSA). [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. t-SNE visualization of the training dataset featuring various latent space augmentations. The green, blue, and red points represent the 2D projections [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The effect of the number of spoofed prototypes on EER (%) across [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 26 canonical work pages

  1. [6]

    Automatic speaker verification spoof- ing and deepfake detection using wav2vec 2.0 and data augmentation,

    Hemlata Tak, Massimiliano Todisco, Xin Wang, Jee-weon Jung, Junichi Yamagishi, and Nicholas Evans, “Automatic speaker verification spoof- ing and deepfake detection using wav2vec 2.0 and data augmentation,” in The Speaker and Language Recognition Workshop , 2022

    work page 2022

  2. [1]

    Asvspoof 2021: accelerating progress in spoofed and deepfake speech detection,

    Junichi Yamagishi, Xin Wang, Massimiliano Todisco, Md Sahidullah, Jose Patino, Andreas Nautsch, Xuechen Liu, Kong Aik Lee, Tomi Kinnunen, Nicholas Evans, et al., “Asvspoof 2021: accelerating progress in spoofed and deepfake speech detection,” in ASVspoof 2021 Workshop- Automatic Speaker Verification and Spoofing Coutermeasures Challenge, 2021

    work page 2021

  3. [2]

    Does audio deepfake detection general- ize?,

    Nicolas M M ¨uller, Pavel Czempin, Franziska Dieckmann, Adam Frogh- yar, and Konstantin B ¨ottinger, “Does audio deepfake detection general- ize?,” Interspeech, 2022

    work page 2022

  4. [3]

    Xls-r: Self-supervised cross-lingual speech representation learning at scale,

    Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, and Michael Auli, “Xls-r: Self-supervised cross-lingual speech representation learning at scale,” arXiv, vol. abs/2111.09296, 2021

    arXiv 2021

  5. [4]

    Robust speech recognition via large- scale weak supervision,

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever, “Robust speech recognition via large- scale weak supervision,” in International conference on machine learning. PMLR, 2023, pp. 28492–28518

    work page 2023

  6. [5]

    Wavlm: Large-scale self-supervised pre-training for full stack speech processing,

    Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, et al., “Wavlm: Large-scale self-supervised pre-training for full stack speech processing,” IEEE Journal of Selected Topics in Signal Process- ing, vol. 16, no. 6, pp. 1505–1518, 2022

    work page 2022

  7. [7]

    One-class learning towards synthetic voice spoofing detection,

    You Zhang, Fei Jiang, and Zhiyao Duan, “One-class learning towards synthetic voice spoofing detection,” IEEE Signal Processing Letters, vol. 28, pp. 937–941, 2021

    work page 2021

  8. [8]

    One-class learning with adaptive centroid shift for audio deepfake detection,

    Hyun Myung Kim, Kangwook Jang, and Hoirin Kim, “One-class learning with adaptive centroid shift for audio deepfake detection,” in Interspeech 2024, 2024, pp. 4853–4857

    work page 2024

Show all 29 references
  1. [9]

    Specaugment: A simple data augmentation method for automatic speech recognition,

    Daniel S Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, Ekin D Cubuk, and Quoc V Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” arXiv preprint arXiv:1904.08779, 2019

    1904 arXiv

  2. [10]

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

    Hemlata Tak, Madhu Kamble, Jose Patino, Massimiliano Todisco, and Nicholas Evans, “Rawboost: A raw data boosting and augmentation method applied to automatic speaker verification anti-spoofing,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal ...

    2022

  3. [11]

    Cpaug: Refining copy-paste augmentation for speech anti- spoofing,

    Linjuan Zhang, Kong Aik Lee, Lin Zhang, Longbiao Wang, and Baon- ing Niu, “Cpaug: Refining copy-paste augmentation for speech anti- spoofing,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 10996–11000

    2024

  4. [12]

    Targeted augmented data for audio deepfake detection,

    Marcella ASTRID, Enjie GHORBEL, and Djamila AOUADA, “Targeted augmented data for audio deepfake detection,” in 32nd European Signal Processing Conference (EUSIPCO 2024) , 2024

    2024

  5. [13]

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

    Xin Wang and Junichi Yamagishi, “Spoofed training data for speech spoofing countermeasure can be efficiently created using neural vocoders,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    2023

  6. [14]

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

    Xin Wang and Junichi Yamagishi, “Can large-scale vocoded spoofed data improve speech spoofing countermeasure with a self-supervised front end?,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 10311–10315

    2024

  7. [15]

    Data augmentation via latent space interpolation for image classification,

    Xiaofeng Liu, Yang Zou, Lingsheng Kong, Zhihui Diao, Junliang Yan, Jun Wang, Site Li, Ping Jia, and Jane You, “Data augmentation via latent space interpolation for image classification,” in 2018 24th International Conference on Pattern Recognition (ICPR) . IEEE, 2018, pp. 728–733

    2018

  8. [16]

    Transcending forgery specificity with latent space augmentation for generalizable deepfake detection,

    Zhiyuan Yan, Yuhao Luo, Siwei Lyu, Qingshan Liu, and Baoyuan Wu, “Transcending forgery specificity with latent space augmentation for generalizable deepfake detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 8984–8994

    2024

  9. [17]

    Asvspoof 2019: A large-scale public database of synthesized, converted and replayed speech,

    Xin Wang, Junichi Yamagishi, Massimiliano Todisco, H ´ector Delgado, Andreas Nautsch, Nicholas Evans, Md Sahidullah, Ville Vestman, Tomi Kinnunen, Kong Aik Lee, et al., “Asvspoof 2019: A large-scale public database of synthesized, converted and replayed speech,” Computer Speec...

    2019

  10. [18]

    Arcface: Additive angular margin loss for deep face recognition,

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4690–4699

    2019

  11. [19]

    mixup: Beyond empirical risk minimization,

    Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez- Paz, “mixup: Beyond empirical risk minimization,” arXiv preprint arXiv:1710.09412, 2017

    2017 arXiv

  12. [20]

    Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks,

    Jee-weon Jung, Hee-Soo Heo, Hemlata Tak, Hye-jin Shim, Joon Son Chung, Bong-Jin Lee, Ha-Jin Yu, and Nicholas Evans, “Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks,” in ICASSP 2022-2022 IEEE international conference on acoustics, speech ...

    2022

  13. [21]

    Attentive merging of hidden embeddings from pre-trained speech model for anti-spoofing detection,

    Zihan Pan, Tianchi Liu, Hardik B. Sailor, and Qiongqiong Wang, “Attentive merging of hidden embeddings from pre-trained speech model for anti-spoofing detection,” in Interspeech 2024, 2024, pp. 2090–2094

    2024

  14. [22]

    Towards generalisable and calibrated audio deepfake detection with self-supervised representations,

    Octavian Pascu, Adriana Stan, Dan Oneata, Elisabeta Oneata, and Horia Cucu, “Towards generalisable and calibrated audio deepfake detection with self-supervised representations,” in Interspeech 2024 , 2024, pp. 4828–4832

    2024

  15. [23]

    Audio deepfake detection with self-supervised wavlm and multi-fusion attentive classifier,

    Yinlin Guo, Haofan Huang, Xi Chen, He Zhao, and Yuehai Wang, “Audio deepfake detection with self-supervised wavlm and multi-fusion attentive classifier,” in ICASSP 2024-2024 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 12...

    2024

  16. [24]

    Anti- spoofing using transfer learning with variational information bottleneck,

    Youngsik Eom, Yeonghyeon Lee, Ji Sub Um, and Hoi Rin Kim, “Anti- spoofing using transfer learning with variational information bottleneck,” in Interspeech 2022, 2022, pp. 3568–3572

    2022

  17. [25]

    One-class knowledge distillation for spoofing speech detection,

    Jingze Lu, Yuxiang Zhang, Wenchao Wang, Zengqiang Shang, and Pengyuan Zhang, “One-class knowledge distillation for spoofing speech detection,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 11251–11255

    2024

  18. [26]

    Genuine-focused learning using mask autoencoder for generalized fake audio detection,

    Xiaopeng Wang, Ruibo Fu, Zhengqi Wen, Zhiyong Wang, Yuankun Xie, Yukun Liu, Jianhua Tao, Xuefei Liu, Yongwei Li, Xin Qi, Yi Lu, and Shuchen Shi, “Genuine-focused learning using mask autoencoder for generalized fake audio detection,” in Interspeech 2024, 2024, pp. 4848– 4852

    2024

  19. [27]

    Improving short utterance anti-spoofing with aasist2,

    Yuxiang Zhang, Jingze Lu, Zengqiang Shang, Wenchao Wang, and Pengyuan Zhang, “Improving short utterance anti-spoofing with aasist2,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 11636–11640

    2024

  20. [28]

    Temporal-channel modeling in multi-head self-attention for synthetic speech detection,

    Duc-Tuan Truong, Ruijie Tao, Tuan Nguyen, Hieu-Thi Luong, Kong Aik Lee, and Eng Siong Chng, “Temporal-channel modeling in multi-head self-attention for synthetic speech detection,” in Interspeech 2024, 2024, pp. 537–541

    2024

  21. [29]

    Spoofing speech detection by modeling local spectro- temporal and long-term dependency,

    Haochen Wu, Wu Guo, Zhentao Zhang, Wenting Zhao, Shengyu Peng, and Jie Zhang, “Spoofing speech detection by modeling local spectro- temporal and long-term dependency,” in Interspeech 2024 , 2024, pp. 507–511

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.