We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Sequential Randomized Smoothing for Adversarially Robust Speech Recognition

Abstract: While Automatic Speech Recognition has been shown to be vulnerable to adversarial attacks, defenses against these attacks are still lagging. Existing, naive defenses can be partially broken with an adaptive attack. In classification tasks, the Randomized Smoothing paradigm has been shown to be effective at defending models. However, it is difficult to apply this paradigm to ASR tasks, due to their complexity and the sequential nature of their outputs. Our paper overcomes some of these challenges by leveraging speech-specific tools like enhancement and ROVER voting to design an ASR model that is robust to perturbations. We apply adaptive versions of state-of-the-art attacks, such as the Imperceptible ASR attack, to our model, and show that our strongest defense is robust to all attacks that use inaudible noise, and can only be broken with very high distortion.
Comments: This update adds some relevant references to past and concurrent work
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Journal reference: 2021.emnlp-main.514 (2021) 6372-6386
DOI: 10.18653/v1/2021.emnlp-main.514
Cite as: arXiv:2112.03000 [cs.CL]
  (or arXiv:2112.03000v2 [cs.CL] for this version)

Submission history

From: Raphael Olivier [view email]
[v1] Fri, 5 Nov 2021 21:51:40 GMT (151kb,D)
[v2] Mon, 10 Jan 2022 19:38:33 GMT (152kb,D)

Link back to: arXiv, form interface, contact.