Applying momentum and variance-tuned gradient attacks in the time domain with a VAD mask improves transferable attack success against Whisper and Speech2Text ASR models.
{WaveGuard}: Understanding and mitigating audio adversarial exam- ples,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Transferable Adversarial Attacks against ASR
Applying momentum and variance-tuned gradient attacks in the time domain with a VAD mask improves transferable attack success against Whisper and Speech2Text ASR models.