Collaborative Watermarking for Adversarial Speech Synthesis
Open accessThe DOI matches its Crossref record; 31 of the 33 references it cites were found in a registry or an index.
Abstract
Advances in neural speech synthesis have brought us technology that is not only close to human naturalness, but is also capable of instant voice cloning with little data, and is highly accessible with pre-trained models available. Naturally, the potential flood of generated content raises the need for synthetic speech detection and watermarking. Recently, considerable research effort in synthetic speech detection has been related to the Automatic Speaker Verification and Spoofing Countermeasure Challenge (ASVspoof), which focuses on passive countermeasures. This paper takes a complementary view to generated speech detection: a synthesis system should make an active effort to watermark the generated speech in a way that aids detection by another machine, but remains transparent to a human listener. We propose a collaborative training scheme for synthetic speech watermarking and show that a HiFi-GAN neural vocoder collaborating with the ASVspoof 2021 baseline countermeasure models consistently improves detection performance over conventional classifier training. Furthermore, we demonstrate how collaborative training can be paired with augmentation strategies for added robustness against noise and time-stretching. Finally, listening tests demonstrate that collaborative training has little adverse effect on perceptual quality of vocoded speech.
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
Labaca-Castro, R. (2023) · Machine Learning under Malware Attack · doi:10.1007/978-3-658-40442-0_9 · Cited by 19,272
Bender, W., Gruhl, D., Morimoto, N., & Lu, A. (1996) · IBM Systems Journal · doi:10.1147/sj.353.0313 · Cited by 2,774
A Light CNN for Deep Face Representation With Noisy Labels
Wu, X., He, R., Sun, Z., & Tan, T. (2018) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2018.2833032 · Cited by 1,186 · cited 1×
“The back end is a light convolution neural network (LCNN) [26], followed by two recurrent layers using long-short term memory units, global average pooling, and a linear output layer [27].”
BackgroundPerceptual Signal Analysis and Eigenvalue-Weighted Metrics for Musical Audio Quality Assessment
Majumder, S. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-27827-2_35 · Cited by 845
HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
Kong, J., Kim, J., & Bae, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.05646 · Cited by 722 · cited 4×
“Note that the HiFi-GAN paper [13] reports training for 2.5M iterations, which does yield higher synthesis quality.”
BackgroundMethodInfluentialSpoofing and countermeasures for speaker verification: A survey
Wu, Z., Evans, N., Kinnunen, T., Yamagishi, J., Alegre, F., & Li, H. (2015) · Speech Communication · doi:10.1016/j.specom.2014.10.005 · Cited by 601 · cited 1×
“Research on audio deepfake detection focuses mostly on passive protection using machine learning methods, which is also referred to as speech anti-spoofing [3].”
BackgroundWatermarking schemes evaluation
Petitcolas, F. A. P. (2000) · IEEE Signal Processing Magazine · doi:10.1109/79.879339 · Cited by 585 · cited 1×
“Watermarks can be evaluated using multiple characteristics, including robustness, perceptibility, and applications [14].”
BackgroundDiffWave: A Versatile Diffusion Model for Audio Synthesis
Kong, Z., Ping, W., Huang, J., Zhao, K., & Catanzaro, B. (2020) · arXiv · doi:10.48550/arxiv.2009.09761 · Cited by 569 · cited 1×
“Any deep generative model with differentiable sampling (including diffusion models [31, 32]) can collaborate with a detector model to improve the odds of detection.”
MethodTransfer learning from speaker verification to multispeaker text-to-speech synthesis
Jia, Y., Zhang, Y., Weiss, R. J., Wang, Q., Shen, J., Ren, F., Chen, Z., Nguyen, P., Pang, R., Moreno, I. L., et al. (2018) · arXiv · doi:10.48550/arxiv.1806.04558 · Cited by 519
Secure spread spectrum watermarking for images, audio and video
Cox, I. J., Kilian, J., Leighton, T., & Shamoon, T. (2002) · Proceedings of 3rd IEEE International Conference on Image Processing · doi:10.1109/icip.1996.560429 · Cited by 505 · cited 1×
“Typical algorithms using DSP include adding pseudo-noise in the temporal or spectral domain (a.k.a spread spectrum watermarking [15]), phase coding [16], and echo hiding [17].”
Method
Research works printed but not linked: 1
- Steinebach, 2004, Digitale Wasserzeichen fuer Audiodaten
Cited by
Audio Deepfake Detection: What Has Been Achieved and What Lies Ahead
Zhang, B., Cui, H., Nguyen, V., & Whitty, M. (2025) · Sensors · doi:10.3390/s25071989 · Cited by 72
“Juvela and Wang [170] introduced a collaborative training-based audio watermarking technique, which enables the generation model to embed watermarks actively, making the synthe-sised speech more easily identifiable by the collaboratively trained detection model while maintaining high audio quality…”
Proactive Detection of Voice Cloning with Localized WatermarkingPreprint
Roman, R. S., Fernandez, P., Défossez, A., Furon, T., Tran, T. K., & Elsahar, H. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2401.17264 · Cited by 183
“Our method aligns more closely with the concurrent work by Juvela & Wang (2023), which trains a detector, rather than a decoder.”
BackgroundMethodInfluentialSeamless: Multilingual Expressive and Streaming Speech Translation
Communication, S., Barrault, L., Chung, Y.-A., Meglioli, M. C., Dale, D., Dong, N., Duppenthaler, M., Duquenne, P.-A., Ellis, B., Elsahar, H., et al. (2023) · arXiv · doi:10.48550/arxiv.2312.05187 · Cited by 45
Latent Watermarking of Audio Generative Models
Roman, R. S., Fernandez, P., Deleforge, A., Adi, Y., & Serizel, R. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10889782 · Cited by 26
“Examples include watermarking: image GANs by training a hyper-network model [35], latent diffusion models by a quick fine-tuning of their decoder [13], and HiFi-GAN decoders that take mel-spectrograms and output waveforms [36].”
Blind and Low-Vision Individuals' Detection of Audio Deepfakes
Sharevski, F., Zeidieh, A., Loop, J. V., & Jachim, P. (2024) · ACM SIGSAC Conference on Computer and Communications Security (CCS) · doi:10.1145/3658644.3690305 · Cited by 14
“A recent trend in response to audio deepfake deception is the use of watermarking, a technique where a detection clue is included in the generated speech to mark it as a “deepfake” in a way that aids detection by another machine, but remains transparent to a human listener [18].”
Metadata
- Authors
- Juvela, L., & Wang, X.
- Affiliations
- Aalto University (FI)
- National Institute of Informatics (JP)
- Venue
- ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
- Publisher
- IEEE
- Published
- 2024-04-14
- Pages
- 11231-11235
- ISSN
- 1520-6149 (ISSN-L)
- OpenAlex
- W4392903621
- Cited by
- 13
- References
- 33
- Citations by year
- 2026: 3 · 2025: 7 · 2024: 3
- FWCI
- 3.82
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE International Conference on Acoustics Speech and Signal Processing (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Aaltodoc (Aalto University) (submittedVersion, other-oa)
- arXiv (Cornell University)
- Funders
- Ministry of Education, Culture, Sports, Science and Technology (CREST)
- Japan Society for the Promotion of Science (21H04906, JPMJCR18A6, CREST)
- Core Research for Evolutional Science and Technology (JPMJCR18A6)
- Precursory Research for Embryonic Science and Technology
- 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 › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- voice cloning, synthetic speech detection, collaborative training, HiFi-GAN, neural vocoder, ASVspoof, spoofing countermeasures, data augmentation, noise robustness, perceptual speech quality
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.
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: Collaborative Watermarking for Adversarial Speech Synthesis. https://app.cochat.ai/papers/10.1109/icassp48485.2024.10448134
- The paper
- Juvela, L., & Wang, X. (2024). Collaborative Watermarking for Adversarial Speech Synthesis. ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). https://doi.org/10.1109/icassp48485.2024.10448134
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.