Matches in SemOpenAlex for { <https://semopenalex.org/work/W4223520345> ?p ?o ?g. }
Showing items 1 to 71 of
71
with 100 items per page.
- W4223520345 abstract "Adversarial attacks are a threat to automatic speech recognition (ASR) systems, and it becomes imperative to propose defenses to protect them. In this paper, we perform experiments to show that K2 conformer hybrid ASR is strongly affected by white-box adversarial attacks. We propose three defenses--denoiser pre-processor, adversarially fine-tuning ASR model, and adversarially fine-tuning joint model of ASR and denoiser. Our evaluation shows denoiser pre-processor (trained on offline adversarial examples) fails to defend against adaptive white-box attacks. However, adversarially fine-tuning the denoiser using a tandem model of denoiser and ASR offers more robustness. We evaluate two variants of this defense--one updating parameters of both models and the second keeping ASR frozen. The joint model offers a mean absolute decrease of 19.3% ground truth (GT) WER with reference to baseline against fast gradient sign method (FGSM) attacks with different $L_infty$ norms. The joint model with frozen ASR parameters gives the best defense against projected gradient descent (PGD) with 7 iterations, yielding a mean absolute increase of 22.3% GT WER with reference to baseline; and against PGD with 500 iterations, yielding a mean absolute decrease of 45.08% GT WER and an increase of 68.05% adversarial target WER." @default.
- W4223520345 created "2022-04-15" @default.
- W4223520345 creator A5014580424 @default.
- W4223520345 creator A5014918159 @default.
- W4223520345 creator A5027217976 @default.
- W4223520345 creator A5050632169 @default.
- W4223520345 creator A5052564654 @default.
- W4223520345 creator A5057010207 @default.
- W4223520345 creator A5062630507 @default.
- W4223520345 date "2022-04-08" @default.
- W4223520345 modified "2023-10-14" @default.
- W4223520345 title "Defense against Adversarial Attacks on Hybrid Speech Recognition using Joint Adversarial Fine-tuning with Denoiser" @default.
- W4223520345 doi "https://doi.org/10.48550/arxiv.2204.03851" @default.
- W4223520345 hasPublicationYear "2022" @default.
- W4223520345 type Work @default.
- W4223520345 citedByCount "0" @default.
- W4223520345 crossrefType "posted-content" @default.
- W4223520345 hasAuthorship W4223520345A5014580424 @default.
- W4223520345 hasAuthorship W4223520345A5014918159 @default.
- W4223520345 hasAuthorship W4223520345A5027217976 @default.
- W4223520345 hasAuthorship W4223520345A5050632169 @default.
- W4223520345 hasAuthorship W4223520345A5052564654 @default.
- W4223520345 hasAuthorship W4223520345A5057010207 @default.
- W4223520345 hasAuthorship W4223520345A5062630507 @default.
- W4223520345 hasBestOaLocation W42235203451 @default.
- W4223520345 hasConcept C104317684 @default.
- W4223520345 hasConcept C11413529 @default.
- W4223520345 hasConcept C127413603 @default.
- W4223520345 hasConcept C153180895 @default.
- W4223520345 hasConcept C153258448 @default.
- W4223520345 hasConcept C154945302 @default.
- W4223520345 hasConcept C170154142 @default.
- W4223520345 hasConcept C18555067 @default.
- W4223520345 hasConcept C185592680 @default.
- W4223520345 hasConcept C28490314 @default.
- W4223520345 hasConcept C37736160 @default.
- W4223520345 hasConcept C41008148 @default.
- W4223520345 hasConcept C50644808 @default.
- W4223520345 hasConcept C55493867 @default.
- W4223520345 hasConcept C63479239 @default.
- W4223520345 hasConceptScore W4223520345C104317684 @default.
- W4223520345 hasConceptScore W4223520345C11413529 @default.
- W4223520345 hasConceptScore W4223520345C127413603 @default.
- W4223520345 hasConceptScore W4223520345C153180895 @default.
- W4223520345 hasConceptScore W4223520345C153258448 @default.
- W4223520345 hasConceptScore W4223520345C154945302 @default.
- W4223520345 hasConceptScore W4223520345C170154142 @default.
- W4223520345 hasConceptScore W4223520345C18555067 @default.
- W4223520345 hasConceptScore W4223520345C185592680 @default.
- W4223520345 hasConceptScore W4223520345C28490314 @default.
- W4223520345 hasConceptScore W4223520345C37736160 @default.
- W4223520345 hasConceptScore W4223520345C41008148 @default.
- W4223520345 hasConceptScore W4223520345C50644808 @default.
- W4223520345 hasConceptScore W4223520345C55493867 @default.
- W4223520345 hasConceptScore W4223520345C63479239 @default.
- W4223520345 hasLocation W42235203451 @default.
- W4223520345 hasOpenAccess W4223520345 @default.
- W4223520345 hasPrimaryLocation W42235203451 @default.
- W4223520345 hasRelatedWork W1529681343 @default.
- W4223520345 hasRelatedWork W2973512984 @default.
- W4223520345 hasRelatedWork W2979999556 @default.
- W4223520345 hasRelatedWork W3005617766 @default.
- W4223520345 hasRelatedWork W3007517706 @default.
- W4223520345 hasRelatedWork W3046550781 @default.
- W4223520345 hasRelatedWork W3082456152 @default.
- W4223520345 hasRelatedWork W3096517430 @default.
- W4223520345 hasRelatedWork W4287868736 @default.
- W4223520345 hasRelatedWork W4292267426 @default.
- W4223520345 isParatext "false" @default.
- W4223520345 isRetracted "false" @default.
- W4223520345 workType "article" @default.