Passer à la navigation principale Passer à la recherche Passer au contenu principal

Adversarially training for audio classifiers

  • École de technologie supérieure

Résultats de recherche: Chapitre dans un livre, rapport, actes de conférence › Participation à un ouvrage collectif lié à un colloque ou une conférence › Revue par des pairs

9 Citations (Scopus)

Résumé

In this paper, we investigate the potential effect of the adversarially training on the robustness of six advanced deep neural networks against a variety of targeted and non-targeted adversarial attacks. We firstly show that, the ResNet-56 model trained on the 2D representation of the discrete wavelet transform appended with the tonnetz chromagram outperforms other models in terms of recognition accuracy. Then we demonstrate the positive impact of adversarially training on this model as well as other deep architectures against six types of attack algorithms (white and black-box) with the cost of the reduced recognition accuracy and limited adversarial perturbation. We run our experiments on two benchmarking environmental sound datasets and show that without any imposed limitations on the budget allocations for the adversary, the fooling rate of the adversarially trained models can exceed 90%. In other words, adversarial attacks exist in any scales, but they might require higher adversarial perturbations compared to non-adversarially trained models.

Langue originaleAnglais
titreProceedings of ICPR 2020 - 25th International Conference on Pattern Recognition
EditeurInstitute of Electrical and Electronics Engineers Inc.
Pages9569-9576
Nombre de pages8
ISBN (Electronique)9781728188089
Les DOIs
ÉtatPublié - 2020
Evénement25th International Conference on Pattern Recognition, ICPR 2020 - Virtual, Online, Italie
Durée: 10 janv. 2021 → 15 janv. 2021

Série de publications

NomProceedings - International Conference on Pattern Recognition
ISSN (imprimé)1051-4651

Conférence

Conférence25th International Conference on Pattern Recognition, ICPR 2020
Pays/TerritoireItalie
La villeVirtual, Online
Période10/01/21 → 15/01/21

Classification disciplinaire ASJC Scopus

  • Vision par ordinateur et reconnaissance de formes

Empreinte digitale

Voici les principaux termes ou expressions associés à « Adversarially training for audio classifiers ». Ces libellés thématiques sont générés à partir du titre et du résumé de la publication. Ensemble, ils forment une empreinte digitale unique.

Citer cette ressource