The DOI matches its Crossref record; 38 of the 38 references it cites were found in a registry or an index.
Measuring the Robustness of Audio Deepfake Detection under Real-World Corruption
Open accessAbstract
Deepfakes have emerged as a widespread and rapidly escalating concern in generative AI, spanning various media types such as images, audio, and videos. Among these, audio deepfakes are particularly alarming due to the growing accessibility of high-quality voice synthesis tools and the ease with which synthetic speech can be distributed via platforms like social media and robocalls. Consequently, detecting audio deepfakes is critical in combating the misuse of AI-synthesized speech. However, real-world audio is often subject to various corruptions, such as noise, modification, and compression, that may significantly impact detection performance. In this work, we systematically evaluate the robustness of 10 audio deepfake detection models against 18 common corruption types, grouped into categories: noise perturbation, audio modification, and compression. Using both traditional deep learning models and state-of-the-art foundation models, our study yields four key insights. (1) Most models demonstrate strong robustness to noise but they are notably more vulnerable to audio modifications and compression, especially when neural codecs are applied. (2) Speech foundation models generally outperform traditional models across most corruption scenarios, likely due to their extensive pre-training on large-scale and diverse audio datasets. (3) Increasing model size improves robustness, though with diminishing returns. (4) Robustness to unseen corruptions can be enhanced by targeted data augmentation during training or by applying speech enhancement techniques at inference time. These findings highlight the importance of comprehensive evaluation against diverse corruption types and developing more robust audio deepfake detection frameworks to ensure reliability in practical deployment settings. We further advocate that future research in deepfake detection across all media formats should account for the diverse and often unpredictable distortions common in real-world environments.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The text we found for this paper looks like another document, so the self-check wasn’t run.
Citations
Cites
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
Robust Speech Recognition via Large-Scale Weak Supervision
Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022) · arXiv · doi:10.48550/arxiv.2212.04356 · Cited by 1,953
Benchmarking Neural Network Robustness to Common Corruptions and Perturbations
Hendrycks, D., & Dietterich, T. (2019) · arXiv · doi:10.48550/arxiv.1903.12261 · Cited by 1,335
High-Fidelity Audio Compression with Improved RVQGAN
Kumar, I., Seetharaman, P., Luebs, A., Kumar, I., & Kumar, K. (2023) · neural information processing systems · doi:10.52202/075280-1214 · Cited by 982
SoundStream: An End-to-End Neural Audio Codec
Zeghidour, N., Luebs, A., Omran, A., Skoglund, J., & Tagliasacchi, M. (2022) · IEEE/ACM Transactions on Audio, Speech, and Language Processing · doi:10.1109/taslp.2021.3129994 · Cited by 610
NaturalSpeech 3: Zero-Shot Speech Synthesis with Factorized Codec and Diffusion ModelsPreprint
Ju, Z., Wang, Y., Shen, K., Tan, X., Xin, D., Yang, D., Liu, Y., Leng, Y., Song, K., Tang, S., et al. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2403.03100 · Cited by 548
Self-citationCLAP Learning Audio Concepts from Natural Language Supervision
Elizalde, B., Deshmukh, S., Ismail, M. A., & Wang, H. (2023) · ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp49357.2023.10095889 · Cited by 449
Data Augmentation Can Improve RobustnessPreprint
Rebuffi, S.-A., Gowal, S., Calian, D. A., Stimberg, F., Wiles, O., & Mann, T. (2021) · Neural Information Processing Systems · doi:10.48550/arxiv.2111.05328 · Cited by 441
End-to-End anti-spoofing with RawNet2
Tak, H., Patino, J., Todisco, M., Nautsch, A., Evans, N., & Larcher, A. (2021) · ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp39728.2021.9414234 · Cited by 402 · cited 1×
“[39], RawNet2 [30], AASIST [16], RawGATST [29] are specifically designed for synthetic speech detection.”
Research works printed but not linked: 2
- CT Cisco. 2021. Global 2021 forecast highlights. Cisco: San Jose, CA, USA (2021).
- Xiang Li, Pin-Yu Chen, and Wenqi Wei. 2024a. SONAR: A Synthetic AI-Audio Detection Framework and Benchmark. arXiv preprint arXiv:2410.04324 (2024). · arXiv
Cited by
Exposing and Mitigating Neural Codec Vulnerabilities in Audio Deepfake DetectionPreprint
Abdullah, Khan, A., & Malik, K. M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.07216
“For instance, Li et al. [3] reported that although self-supervised (SSL)-based foundation models perform comparatively well under moderate compression, their performance drops noticeably under extreme conditions, such as low-bitrate MP3 encoding.”
Neural Audio Codec for Robust Audio Deepfake Detection
Kim, J., Park, J., Koh, J., et al. (2026) · doi:10.48550/arxiv.2609.39651
“Robustness studies have further shown that audio compression can substantially degrade detection performance, with neural codecs often causing severe performance drops even when perceptual quality remains high [17, 18, 19].”
Robust Workflow Generation via Adversarial Learning for Audio Deepfake DetectionPreprint
Li, X., Chen, P.-Y., & Wei, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.20063
“Consistent with prior robustness studies [6], challenging perturbations such as EnCodec, Opus, quantization, pitch shifting, and time stretching receive high weights, whereas Gaussian noise and silence injection receive much lower weights.”
InfluentialSelf-citationShi, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.07981
Gris, L. R. S., Casanova, D., de Oliveira, F. S., Ferreira, A. I., Felício, B. A., Mata, R. C. R., & Soares, A. D. S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.28770
“Robustness studies further show that compression, additive noise, and reverberation can suppress synthesis artifacts and increase detection errors [13, 14].”
Metadata
- Authors
- Li, X., Chen, P.-Y., & Wei, W.
- Affiliations
- Fordham University (US)
- IBM (United States) (US)
- Venue
- Proceedings of the Sixteenth ACM Conference on Data and Application Security and Privacy
- Publisher
- ACM
- Published
- 2026-06-22
- Pages
- 145-158
- OpenAlex
- W7164519677
- Cited by
- 0
- References
- 5
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- ACM Conference on Data and Application Security and Privacy (CODASPY) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Keywords
- audio deepfake detection, corruption robustness, noise perturbation, audio compression, neural audio codecs, speech foundation models, model size, data augmentation, speech enhancement, synthetic speech, AI-generated audio
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.
A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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: Measuring the Robustness of Audio Deepfake Detection under Real-World Corruption. https://app.cochat.ai/papers/10.1145/3800506.3803494
- The paper
- Li, X., Chen, P.-Y., & Wei, W. (2026). Measuring the Robustness of Audio Deepfake Detection under Real-World Corruption. Proceedings of the Sixteenth ACM Conference on Data and Application Security and Privacy. https://doi.org/10.1145/3800506.3803494
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.