Matches in SemOpenAlex for { <https://semopenalex.org/work/W3128748085> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W3128748085 abstract "While machine learning models today can achieve high accuracies on classification tasks, they can be deceived by minor imperceptible distortions to the data. These are known as adversarial attacks and can be lethal in the black-box setting which does not require knowledge of the target model type or its parameters. Binary neural networks that have sign activation and are trained with gradient descent have been shown to be harder to attack than conventional sigmoid activation networks but their improvements are marginal. We instead train sign activation networks with a novel gradient-free stochastic coordinate descent algorithm and propose an ensemble of such networks as a defense model. We evaluate the robustness of our model (a hard problem in itself) on image, text, and medical ECG data and find it to be more robust than ensembles of binary, full precision, and convolutional neural networks, and than random forests while attaining comparable clean test accuracy. In order to explain our model's robustness we show that an adversary targeting a single network in our ensemble fails to attack (and thus non-transferable to) other networks in the ensemble. Thus a datapoint requires a large distortion to fool the majority of networks in our ensemble and is likely to be detected in advance. This property of non-transferability arises naturally from the non-convexity of sign activation networks and randomization in our gradient-free training algorithm without any adversarial defense effort." @default.
- W3128748085 created "2021-02-15" @default.
- W3128748085 creator A5036148022 @default.
- W3128748085 creator A5068347202 @default.
- W3128748085 creator A5083898401 @default.
- W3128748085 creator A5087990990 @default.
- W3128748085 date "2021-05-04" @default.
- W3128748085 modified "2023-09-23" @default.
- W3128748085 title "Defending against black-box adversarial attacks with gradient-free trained sign activation neural networks" @default.
- W3128748085 hasPublicationYear "2021" @default.
- W3128748085 type Work @default.
- W3128748085 sameAs 3128748085 @default.
- W3128748085 citedByCount "0" @default.
- W3128748085 crossrefType "journal-article" @default.
- W3128748085 hasAuthorship W3128748085A5036148022 @default.
- W3128748085 hasAuthorship W3128748085A5068347202 @default.
- W3128748085 hasAuthorship W3128748085A5083898401 @default.
- W3128748085 hasAuthorship W3128748085A5087990990 @default.
- W3128748085 hasConcept C104317684 @default.
- W3128748085 hasConcept C108583219 @default.
- W3128748085 hasConcept C11413529 @default.
- W3128748085 hasConcept C119857082 @default.
- W3128748085 hasConcept C12267149 @default.
- W3128748085 hasConcept C153180895 @default.
- W3128748085 hasConcept C153258448 @default.
- W3128748085 hasConcept C154945302 @default.
- W3128748085 hasConcept C185592680 @default.
- W3128748085 hasConcept C206688291 @default.
- W3128748085 hasConcept C33923547 @default.
- W3128748085 hasConcept C38365724 @default.
- W3128748085 hasConcept C41008148 @default.
- W3128748085 hasConcept C48372109 @default.
- W3128748085 hasConcept C50644808 @default.
- W3128748085 hasConcept C55493867 @default.
- W3128748085 hasConcept C63479239 @default.
- W3128748085 hasConcept C66905080 @default.
- W3128748085 hasConcept C81363708 @default.
- W3128748085 hasConcept C81388566 @default.
- W3128748085 hasConcept C94375191 @default.
- W3128748085 hasConceptScore W3128748085C104317684 @default.
- W3128748085 hasConceptScore W3128748085C108583219 @default.
- W3128748085 hasConceptScore W3128748085C11413529 @default.
- W3128748085 hasConceptScore W3128748085C119857082 @default.
- W3128748085 hasConceptScore W3128748085C12267149 @default.
- W3128748085 hasConceptScore W3128748085C153180895 @default.
- W3128748085 hasConceptScore W3128748085C153258448 @default.
- W3128748085 hasConceptScore W3128748085C154945302 @default.
- W3128748085 hasConceptScore W3128748085C185592680 @default.
- W3128748085 hasConceptScore W3128748085C206688291 @default.
- W3128748085 hasConceptScore W3128748085C33923547 @default.
- W3128748085 hasConceptScore W3128748085C38365724 @default.
- W3128748085 hasConceptScore W3128748085C41008148 @default.
- W3128748085 hasConceptScore W3128748085C48372109 @default.
- W3128748085 hasConceptScore W3128748085C50644808 @default.
- W3128748085 hasConceptScore W3128748085C55493867 @default.
- W3128748085 hasConceptScore W3128748085C63479239 @default.
- W3128748085 hasConceptScore W3128748085C66905080 @default.
- W3128748085 hasConceptScore W3128748085C81363708 @default.
- W3128748085 hasConceptScore W3128748085C81388566 @default.
- W3128748085 hasConceptScore W3128748085C94375191 @default.
- W3128748085 hasLocation W31287480851 @default.
- W3128748085 hasOpenAccess W3128748085 @default.
- W3128748085 hasPrimaryLocation W31287480851 @default.
- W3128748085 hasRelatedWork W2892487364 @default.
- W3128748085 hasRelatedWork W2895033754 @default.
- W3128748085 hasRelatedWork W2913317039 @default.
- W3128748085 hasRelatedWork W2914520039 @default.
- W3128748085 hasRelatedWork W2952666634 @default.
- W3128748085 hasRelatedWork W2956085480 @default.
- W3128748085 hasRelatedWork W2962710014 @default.
- W3128748085 hasRelatedWork W2967955854 @default.
- W3128748085 hasRelatedWork W3010258799 @default.
- W3128748085 hasRelatedWork W3016549807 @default.
- W3128748085 hasRelatedWork W3024713099 @default.
- W3128748085 hasRelatedWork W3041782238 @default.
- W3128748085 hasRelatedWork W3048203754 @default.
- W3128748085 hasRelatedWork W3090800462 @default.
- W3128748085 hasRelatedWork W3091172437 @default.
- W3128748085 hasRelatedWork W3093572737 @default.
- W3128748085 hasRelatedWork W3094480364 @default.
- W3128748085 hasRelatedWork W3132910243 @default.
- W3128748085 hasRelatedWork W3161790530 @default.
- W3128748085 hasRelatedWork W3192751986 @default.
- W3128748085 isParatext "false" @default.
- W3128748085 isRetracted "false" @default.
- W3128748085 magId "3128748085" @default.
- W3128748085 workType "article" @default.