Matches in SemOpenAlex for { <https://semopenalex.org/work/W3157391214> ?p ?o ?g. }
Showing items 1 to 81 of
81
with 100 items per page.
- W3157391214 endingPage "102280" @default.
- W3157391214 startingPage "102280" @default.
- W3157391214 abstract "Backdoor data poisoning (a.k.a. Trojan attack) is an emerging form of adversarial attack usually against deep neural network image classifiers. The attacker poisons the training set with a relatively small set of images from one (or several) source class(es), embedded with a backdoor pattern and labeled to a target class. For a successful attack, during operation, the trained classifier will: 1) misclassify a test image from the source class(es) to the target class whenever the backdoor pattern is present; 2) maintain high classification accuracy for backdoor-free test images. In this paper, we make a breakthrough in defending backdoor attacks with imperceptible backdoor patterns (e.g. watermarks) before/during the classifier training phase. This is a challenging problem because it is a priori unknown which subset (if any) of the training set has been poisoned. We propose an optimization-based reverse engineering defense that jointly: 1) detects whether the training set is poisoned; 2) if so, accurately identifies the target class and the training images with the backdoor pattern embedded; and 3) additionally, reverse engineers an estimate of the backdoor pattern used by the attacker. In benchmark experiments on CIFAR-10 (as well as four other data sets), considering a variety of attacks, our defense achieves a new state-of-the-art by reducing the attack success rate to no more than 4.9% after removing detected suspicious training images." @default.
- W3157391214 created "2021-05-10" @default.
- W3157391214 creator A5049853144 @default.
- W3157391214 creator A5063903486 @default.
- W3157391214 creator A5085283385 @default.
- W3157391214 date "2021-07-01" @default.
- W3157391214 modified "2023-10-03" @default.
- W3157391214 title "Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing" @default.
- W3157391214 cites W1989898472 @default.
- W3157391214 cites W2067713319 @default.
- W3157391214 cites W2107397716 @default.
- W3157391214 cites W2112796928 @default.
- W3157391214 cites W2942091739 @default.
- W3157391214 cites W2963196925 @default.
- W3157391214 cites W2973217491 @default.
- W3157391214 cites W3007264885 @default.
- W3157391214 cites W3132644649 @default.
- W3157391214 cites W4247200422 @default.
- W3157391214 doi "https://doi.org/10.1016/j.cose.2021.102280" @default.
- W3157391214 hasPublicationYear "2021" @default.
- W3157391214 type Work @default.
- W3157391214 sameAs 3157391214 @default.
- W3157391214 citedByCount "7" @default.
- W3157391214 countsByYear W31573912142021 @default.
- W3157391214 countsByYear W31573912142022 @default.
- W3157391214 countsByYear W31573912142023 @default.
- W3157391214 crossrefType "journal-article" @default.
- W3157391214 hasAuthorship W3157391214A5049853144 @default.
- W3157391214 hasAuthorship W3157391214A5063903486 @default.
- W3157391214 hasAuthorship W3157391214A5085283385 @default.
- W3157391214 hasBestOaLocation W31573912141 @default.
- W3157391214 hasConcept C119857082 @default.
- W3157391214 hasConcept C153180895 @default.
- W3157391214 hasConcept C154945302 @default.
- W3157391214 hasConcept C169903167 @default.
- W3157391214 hasConcept C174333608 @default.
- W3157391214 hasConcept C177264268 @default.
- W3157391214 hasConcept C199360897 @default.
- W3157391214 hasConcept C2777212361 @default.
- W3157391214 hasConcept C2781045450 @default.
- W3157391214 hasConcept C38652104 @default.
- W3157391214 hasConcept C41008148 @default.
- W3157391214 hasConcept C50644808 @default.
- W3157391214 hasConcept C51632099 @default.
- W3157391214 hasConcept C95623464 @default.
- W3157391214 hasConceptScore W3157391214C119857082 @default.
- W3157391214 hasConceptScore W3157391214C153180895 @default.
- W3157391214 hasConceptScore W3157391214C154945302 @default.
- W3157391214 hasConceptScore W3157391214C169903167 @default.
- W3157391214 hasConceptScore W3157391214C174333608 @default.
- W3157391214 hasConceptScore W3157391214C177264268 @default.
- W3157391214 hasConceptScore W3157391214C199360897 @default.
- W3157391214 hasConceptScore W3157391214C2777212361 @default.
- W3157391214 hasConceptScore W3157391214C2781045450 @default.
- W3157391214 hasConceptScore W3157391214C38652104 @default.
- W3157391214 hasConceptScore W3157391214C41008148 @default.
- W3157391214 hasConceptScore W3157391214C50644808 @default.
- W3157391214 hasConceptScore W3157391214C51632099 @default.
- W3157391214 hasConceptScore W3157391214C95623464 @default.
- W3157391214 hasFunder F4320338279 @default.
- W3157391214 hasLocation W31573912141 @default.
- W3157391214 hasLocation W31573912142 @default.
- W3157391214 hasOpenAccess W3157391214 @default.
- W3157391214 hasPrimaryLocation W31573912141 @default.
- W3157391214 hasRelatedWork W2563096758 @default.
- W3157391214 hasRelatedWork W2792951589 @default.
- W3157391214 hasRelatedWork W2989852175 @default.
- W3157391214 hasRelatedWork W3113332968 @default.
- W3157391214 hasRelatedWork W3135366566 @default.
- W3157391214 hasRelatedWork W3157391214 @default.
- W3157391214 hasRelatedWork W3157486893 @default.
- W3157391214 hasRelatedWork W3201070945 @default.
- W3157391214 hasRelatedWork W4287281082 @default.
- W3157391214 hasRelatedWork W4288797976 @default.
- W3157391214 hasVolume "106" @default.
- W3157391214 isParatext "false" @default.
- W3157391214 isRetracted "false" @default.
- W3157391214 magId "3157391214" @default.
- W3157391214 workType "article" @default.