Matches in SemOpenAlex for { <https://semopenalex.org/work/W3151299405> ?p ?o ?g. }
Showing items 1 to 81 of
81
with 100 items per page.
- W3151299405 abstract "Recent studies have been shown the vulnerability of Deep neural networks (DNNs) by a small adversarial perturbation in images that humans cannot distinguish and a well-trained neural network misclassifies. There are many defense methods, e.g. randomness, ensemble, and adversarial training that can improve the robustness of neural networks with respect to adversaries. Among them, adversarial training has been being the most outstanding defense. However, the training cost becomes a big challenge to adversarial learning. In this paper, we propose a new defense algorithm called Bayes without Bayesian Learning, which does not add the training phase in resisting adversarial attacks. Our method has based on the stochastic components of Bayesian Convolutional Neural Network (BCNN) to prevent the forceful gradient based attacks and generate the ensemble model to enhance the performance of models. In order to avoid adversarial training, we utilize pretrained CNN models on both natural and perturbed data. Therefore, this approach can be applied to any model with learned parameters to accelerate the computation of the defensive method. Our defensive algorithm has significantly improved the accuracy of pretrained CNN models under the high level of attacks. For instance, on ImageNet with Resnet50 network, we achieve 57% prediction accuracy in the top5, under strong PGD attack. We also confirm 3% improvement when combining our method with adversarial training. Experimental results prove the efficiency of our BCNN model that does not need Bayesian learning but still resists adversarial perturbation on both small and large scale datasets as CIFAR10, CIFAR100, and ImageNet." @default.
- W3151299405 created "2021-04-13" @default.
- W3151299405 creator A5042516424 @default.
- W3151299405 creator A5046136958 @default.
- W3151299405 creator A5074853381 @default.
- W3151299405 date "2020-11-01" @default.
- W3151299405 modified "2023-09-24" @default.
- W3151299405 title "Bayes without Bayesian Learning for Resisting Adversarial Attacks" @default.
- W3151299405 cites W1571206560 @default.
- W3151299405 cites W2302255633 @default.
- W3151299405 cites W2557283755 @default.
- W3151299405 cites W2603766943 @default.
- W3151299405 cites W2774644650 @default.
- W3151299405 cites W2795727551 @default.
- W3151299405 cites W2891519874 @default.
- W3151299405 cites W2895033754 @default.
- W3151299405 cites W2916286792 @default.
- W3151299405 cites W2951266961 @default.
- W3151299405 cites W2951595529 @default.
- W3151299405 cites W2962710014 @default.
- W3151299405 cites W2963207607 @default.
- W3151299405 cites W2963626858 @default.
- W3151299405 cites W2963857521 @default.
- W3151299405 cites W2964082701 @default.
- W3151299405 cites W2964253222 @default.
- W3151299405 cites W2970317235 @default.
- W3151299405 cites W2996564870 @default.
- W3151299405 doi "https://doi.org/10.1109/candar51075.2020.00038" @default.
- W3151299405 hasPublicationYear "2020" @default.
- W3151299405 type Work @default.
- W3151299405 sameAs 3151299405 @default.
- W3151299405 citedByCount "0" @default.
- W3151299405 crossrefType "proceedings-article" @default.
- W3151299405 hasAuthorship W3151299405A5042516424 @default.
- W3151299405 hasAuthorship W3151299405A5046136958 @default.
- W3151299405 hasAuthorship W3151299405A5074853381 @default.
- W3151299405 hasConcept C104317684 @default.
- W3151299405 hasConcept C107673813 @default.
- W3151299405 hasConcept C108583219 @default.
- W3151299405 hasConcept C119857082 @default.
- W3151299405 hasConcept C154945302 @default.
- W3151299405 hasConcept C185592680 @default.
- W3151299405 hasConcept C207201462 @default.
- W3151299405 hasConcept C2778403875 @default.
- W3151299405 hasConcept C37736160 @default.
- W3151299405 hasConcept C41008148 @default.
- W3151299405 hasConcept C50644808 @default.
- W3151299405 hasConcept C55493867 @default.
- W3151299405 hasConcept C63479239 @default.
- W3151299405 hasConcept C81363708 @default.
- W3151299405 hasConceptScore W3151299405C104317684 @default.
- W3151299405 hasConceptScore W3151299405C107673813 @default.
- W3151299405 hasConceptScore W3151299405C108583219 @default.
- W3151299405 hasConceptScore W3151299405C119857082 @default.
- W3151299405 hasConceptScore W3151299405C154945302 @default.
- W3151299405 hasConceptScore W3151299405C185592680 @default.
- W3151299405 hasConceptScore W3151299405C207201462 @default.
- W3151299405 hasConceptScore W3151299405C2778403875 @default.
- W3151299405 hasConceptScore W3151299405C37736160 @default.
- W3151299405 hasConceptScore W3151299405C41008148 @default.
- W3151299405 hasConceptScore W3151299405C50644808 @default.
- W3151299405 hasConceptScore W3151299405C55493867 @default.
- W3151299405 hasConceptScore W3151299405C63479239 @default.
- W3151299405 hasConceptScore W3151299405C81363708 @default.
- W3151299405 hasLocation W31512994051 @default.
- W3151299405 hasOpenAccess W3151299405 @default.
- W3151299405 hasPrimaryLocation W31512994051 @default.
- W3151299405 hasRelatedWork W11339170 @default.
- W3151299405 hasRelatedWork W12016155 @default.
- W3151299405 hasRelatedWork W349990 @default.
- W3151299405 hasRelatedWork W5005022 @default.
- W3151299405 hasRelatedWork W5410279 @default.
- W3151299405 hasRelatedWork W5508898 @default.
- W3151299405 hasRelatedWork W7303821 @default.
- W3151299405 hasRelatedWork W8021486 @default.
- W3151299405 hasRelatedWork W9190101 @default.
- W3151299405 hasRelatedWork W9657784 @default.
- W3151299405 isParatext "false" @default.
- W3151299405 isRetracted "false" @default.
- W3151299405 magId "3151299405" @default.
- W3151299405 workType "article" @default.