AI RESEARCH
Are Deep Speech Denoising Models Robust to Adversarial Noise?
arXiv CS.LG
•
ArXi:2503.11627v2 Announce Type: replace-cross 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.