Matches in SemOpenAlex for { <https://semopenalex.org/work/W4200609671> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W4200609671 endingPage "3251" @default.
- W4200609671 startingPage "3243" @default.
- W4200609671 abstract "The novel multiexit deep neural network (DNN) architectures provide a new optimization solution for efficient model inference in edge systems. Inference of most samples can be completed within the first few layers on an edge device without the need to transmit them to a remote server. This can significantly increase the inference speed and system throughput, which is particularly beneficial to the resource-constrained scenarios. Unfortunately, researchers proposed an inference slow-down attack against this technique, where an external adversary can add imperceptible perturbations on clean samples to invalidate the multiexit mechanism. In this article, we propose a defensive quantization ( <monospace xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>DefQ</monospace> ) method as the first defense against the inference slow-down attack. It is designed to be lightweight and can be easily implemented in off-the-shelf camera sensors. Particularly, <monospace xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>DefQ</monospace> introduces a novel quantization operation to preprocess the input images. It is capable of removing the perturbations from the malicious samples and preserving the correct inference exit points and prediction accuracy. Meanwhile, it has little impact on the clean samples. Extensive evaluations show that <monospace xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>DefQ</monospace> can effectively defeat the inference slow-down attack and well protect the efficiency of edge systems." @default.
- W4200609671 created "2021-12-31" @default.
- W4200609671 creator A5019692903 @default.
- W4200609671 creator A5024417199 @default.
- W4200609671 creator A5028270700 @default.
- W4200609671 creator A5052813770 @default.
- W4200609671 creator A5083330935 @default.
- W4200609671 date "2023-02-15" @default.
- W4200609671 modified "2023-09-26" @default.
- W4200609671 title "DefQ: Defensive Quantization Against Inference Slow-Down Attack for Edge Computing" @default.
- W4200609671 cites W1979387426 @default.
- W4200609671 cites W1982025193 @default.
- W4200609671 cites W2001996312 @default.
- W4200609671 cites W2033622708 @default.
- W4200609671 cites W2043407937 @default.
- W4200609671 cites W2143612262 @default.
- W4200609671 cites W2145339207 @default.
- W4200609671 cites W2154555188 @default.
- W4200609671 cites W2302255633 @default.
- W4200609671 cites W2586654419 @default.
- W4200609671 cites W2733765803 @default.
- W4200609671 cites W2804503210 @default.
- W4200609671 cites W2913259440 @default.
- W4200609671 cites W2934843808 @default.
- W4200609671 cites W2947969447 @default.
- W4200609671 cites W2963087201 @default.
- W4200609671 cites W2963125010 @default.
- W4200609671 cites W2963384482 @default.
- W4200609671 cites W2963393494 @default.
- W4200609671 cites W2963446712 @default.
- W4200609671 cites W2981114133 @default.
- W4200609671 cites W2989885118 @default.
- W4200609671 cites W3037830434 @default.
- W4200609671 cites W3084432478 @default.
- W4200609671 cites W3097297321 @default.
- W4200609671 cites W3163711566 @default.
- W4200609671 cites W4236099117 @default.
- W4200609671 doi "https://doi.org/10.1109/jiot.2021.3138935" @default.
- W4200609671 hasPublicationYear "2023" @default.
- W4200609671 type Work @default.
- W4200609671 citedByCount "1" @default.
- W4200609671 countsByYear W42006096712022 @default.
- W4200609671 crossrefType "journal-article" @default.
- W4200609671 hasAuthorship W4200609671A5019692903 @default.
- W4200609671 hasAuthorship W4200609671A5024417199 @default.
- W4200609671 hasAuthorship W4200609671A5028270700 @default.
- W4200609671 hasAuthorship W4200609671A5052813770 @default.
- W4200609671 hasAuthorship W4200609671A5083330935 @default.
- W4200609671 hasConcept C111919701 @default.
- W4200609671 hasConcept C11413529 @default.
- W4200609671 hasConcept C124101348 @default.
- W4200609671 hasConcept C138236772 @default.
- W4200609671 hasConcept C154945302 @default.
- W4200609671 hasConcept C162307627 @default.
- W4200609671 hasConcept C2776214188 @default.
- W4200609671 hasConcept C28855332 @default.
- W4200609671 hasConcept C41008148 @default.
- W4200609671 hasConcept C79974875 @default.
- W4200609671 hasConceptScore W4200609671C111919701 @default.
- W4200609671 hasConceptScore W4200609671C11413529 @default.
- W4200609671 hasConceptScore W4200609671C124101348 @default.
- W4200609671 hasConceptScore W4200609671C138236772 @default.
- W4200609671 hasConceptScore W4200609671C154945302 @default.
- W4200609671 hasConceptScore W4200609671C162307627 @default.
- W4200609671 hasConceptScore W4200609671C2776214188 @default.
- W4200609671 hasConceptScore W4200609671C28855332 @default.
- W4200609671 hasConceptScore W4200609671C41008148 @default.
- W4200609671 hasConceptScore W4200609671C79974875 @default.
- W4200609671 hasFunder F4320321001 @default.
- W4200609671 hasIssue "4" @default.
- W4200609671 hasLocation W42006096711 @default.
- W4200609671 hasOpenAccess W4200609671 @default.
- W4200609671 hasPrimaryLocation W42006096711 @default.
- W4200609671 hasRelatedWork W2949701228 @default.
- W4200609671 hasRelatedWork W3013760193 @default.
- W4200609671 hasRelatedWork W3110967717 @default.
- W4200609671 hasRelatedWork W3131458535 @default.
- W4200609671 hasRelatedWork W3162668736 @default.
- W4200609671 hasRelatedWork W3166603227 @default.
- W4200609671 hasRelatedWork W3214097103 @default.
- W4200609671 hasRelatedWork W4281678247 @default.
- W4200609671 hasRelatedWork W4295854938 @default.
- W4200609671 hasRelatedWork W4323922040 @default.
- W4200609671 hasVolume "10" @default.
- W4200609671 isParatext "false" @default.
- W4200609671 isRetracted "false" @default.
- W4200609671 workType "article" @default.