mlsanityData to 14 Oct 2026

Are Deep Speech Denoising Models Robust to Adversarial Noise?

arXiv 2503.11627 ↗First posted 14 Mar 2025cs.SD cs.LG eess.AScited by 0

Not on the map. Why?

Abstract (arXiv)

Deep noise suppression (DNS) models enjoy widespread use throughout a variety of high-stakes speech applications. However, we show that four recent DNS models can each be reduced to outputting unintelligible gibberish through the addition of psychoacoustically hidden adversarial noise, even in low-background-noise and simulated over-the-air settings. For three of the models, a small transcription study with audio and multimedia experts confirms unintelligibility of the attacked audio; simultaneously, an ABX study shows that the adversarial noise is generally imperceptible, with some variance between participants and samples. While we also establish several negative results around targeted attacks and model transfer, our results nevertheless highlight the need for practical countermeasures before open-source DNS systems can be used in safety-critical applications.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 51 of its 77 references that are in the map. First posted on arXiv 14 Mar 2025.

Builds on

51 of 77 references.

References · 51

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  3. 2013Intriguing properties of neural networkscited by 5,155
  4. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,057
  5. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
  6. 2016Universal adversarial perturbationscited by 784
  7. 2020On Adaptive Attacks to Adversarial Example Defensescited by 555
  8. 2018Audio Adversarial Examples: Targeted Attacks on Speech-to-Textcited by 274
  9. 2021NISQA: A Deep CNN-Self-Attention Model for Multidimensional Speech Quality Prediction with Crowdsourced Datasetscited by 192
  10. 2020Real Time Speech Enhancement in the Waveform Domaincited by 186
  11. 2021DNSMOS P.835: A Non-Intrusive Perceptual Objective Speech Quality Metric to Evaluate Noise Suppressorscited by 172
  12. 2024MaskGCT: Zero-Shot Text-to-Speech with Masked Generative Codec Transformercited by 161
38 more
  1. 2019Imperceptible, Robust, and Targeted Adversarial Examples for Automatic Speech Recognitioncited by 124
  2. 2018Adversarial Attacks Against Automatic Speech Recognition Systems via Psychoacoustic Hidingcited by 85
  3. 2015Analysis of classifiers' robustness to adversarial perturbationscited by 76
  4. 2017DolphinAtack: Inaudible Voice Commandscited by 69
  5. 2019Who is Real Bob? Adversarial Attacks on Speaker Recognition Systemscited by 67
  6. 2023ICASSP 2023 Deep Noise Suppression Challengecited by 65
  7. 2022ICASSP 2022 Deep Noise Suppression Challengecited by 60
  8. 2023MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectracited by 52
  9. 2018Did you hear that? Adversarial Examples Against Automatic Speech Recognitioncited by 48
  10. 2019Universal Adversarial Perturbations for Speech Recognition Systemscited by 41
  11. 2021WaveGuard: Understanding and Mitigating Audio Adversarial Examplescited by 40
  12. 2018Fooling End-to-end Speaker Verification by Adversarial Examplescited by 39
  13. 2018Characterizing Audio Adversarial Examples Using Temporal Dependencycited by 36
  14. 2023SELM: Speech Enhancement Using Discrete Tokens and Language Modelscited by 36
  15. 2020Enabling Fast and Universal Audio Adversarial Attack Using Generative Modelcited by 32
  16. 2017Crafting Adversarial Examples For Speech Paralinguistics Applicationscited by 30
  17. 2022FullSubNet+: Channel Attention FullSubNet with Complex Spectrograms for Speech Enhancementcited by 29
  18. 2022FRCRN: Boosting Feature Representation using Frequency Recurrence for Monaural Speech Enhancementcited by 26
  19. 2019Adversarial Attacks on GMM i-vector based Speaker Verification Systemscited by 23
  20. 2022Towards Understanding and Mitigating Audio Adversarial Examples for Speaker Recognitioncited by 23
  21. 2023Defending against Adversarial Audio via Diffusion Modelcited by 23
  22. 2020Adversarial Attack and Defense Strategies for Deep Speaker Recognition Systemscited by 21
  23. 2021Attack on practical speaker verification system using universal adversarial perturbationscited by 18
  24. 2019Imperio: Robust Over-the-Air Adversarial Examples for Automatic Speech Recognition Systemscited by 16
  25. 2020Inaudible Adversarial Perturbations for Targeted Attack in Speaker Recognitioncited by 13
  26. 2021Sequential Randomized Smoothing for Adversarially Robust Speech Recognitioncited by 13
  27. 2022Towards Adversarially Robust Deep Image Denoisingcited by 13
  28. 2023Explicit Estimation of Magnitude and Phase Spectra in Parallel for High-Quality Speech Enhancementcited by 13
  29. 2019Universal Adversarial Audio Perturbationscited by 11
  30. 2020FoolHD: Fooling speaker identification by Highly imperceptible adversarial Disturbancescited by 7
  31. 2023On the Behavior of Intrusive and Non-intrusive Speech Enhancement Metrics in Predictive and Generative Settingscited by 6
  32. 2023Evaluating Similitude and Robustness of Deep Image Denoising Models via Adversarial Attackcited by 6
  33. 2020Adversarial attacks on audio source separationcited by 5
  34. 2020Adversarially Training for Audio Classifierscited by 3
  35. 2024Real-time multichannel deep speech enhancement in hearing aids: Comparing monaural and binaural processing in complex acoustic scenarioscited by 3
  36. 2023ROSE: A Recognition-Oriented Speech Enhancement Framework in Air Traffic Control Using Multi-Objective Learningcited by 2
  37. 2024Towards Evaluating the Robustness of Automatic Speech Recognition Systems via Audio Style Transfercited by 2
  38. 2024Audio Enhancement for Computer Audition -- An Iterative Training Paradigm Using Sample Importancecited by 2

50 most cited of 51 All 51 (JSON)

Built on by

Not cited yet.