Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313447762> ?p ?o ?g. }
Showing items 1 to 61 of
61
with 100 items per page.
- W4313447762 abstract "Adversarial training is an effective approach to make deep neural networks robust against adversarial attacks. Recently, different adversarial training defenses are proposed that not only maintain a high clean accuracy but also show significant robustness against popular and well studied adversarial attacks such as PGD. High adversarial robustness can also arise if an attack fails to find adversarial gradient directions, a phenomenon known as `gradient masking'. In this work, we analyse the effect of label smoothing on adversarial training as one of the potential causes of gradient masking. We then develop a guided mechanism to avoid local minima during attack optimization, leading to a novel attack dubbed Guided Projected Gradient Attack (G-PGA). Our attack approach is based on a `match and deceive' loss that finds optimal adversarial directions through guidance from a surrogate model. Our modified attack does not require random restarts, large number of attack iterations or search for an optimal step-size. Furthermore, our proposed G-PGA is generic, thus it can be combined with an ensemble attack strategy as we demonstrate for the case of Auto-Attack, leading to efficiency and convergence speed improvements. More than an effective attack, G-PGA can be used as a diagnostic tool to reveal elusive robustness due to gradient masking in adversarial defenses." @default.
- W4313447762 created "2023-01-06" @default.
- W4313447762 creator A5012655817 @default.
- W4313447762 creator A5038550641 @default.
- W4313447762 creator A5079112987 @default.
- W4313447762 creator A5083105774 @default.
- W4313447762 date "2022-12-30" @default.
- W4313447762 modified "2023-09-26" @default.
- W4313447762 title "Guidance Through Surrogate: Towards a Generic Diagnostic Attack" @default.
- W4313447762 doi "https://doi.org/10.48550/arxiv.2212.14875" @default.
- W4313447762 hasPublicationYear "2022" @default.
- W4313447762 type Work @default.
- W4313447762 citedByCount "0" @default.
- W4313447762 crossrefType "posted-content" @default.
- W4313447762 hasAuthorship W4313447762A5012655817 @default.
- W4313447762 hasAuthorship W4313447762A5038550641 @default.
- W4313447762 hasAuthorship W4313447762A5079112987 @default.
- W4313447762 hasAuthorship W4313447762A5083105774 @default.
- W4313447762 hasBestOaLocation W43134477621 @default.
- W4313447762 hasConcept C104317684 @default.
- W4313447762 hasConcept C126255220 @default.
- W4313447762 hasConcept C134306372 @default.
- W4313447762 hasConcept C154945302 @default.
- W4313447762 hasConcept C185592680 @default.
- W4313447762 hasConcept C186633575 @default.
- W4313447762 hasConcept C2984842247 @default.
- W4313447762 hasConcept C33923547 @default.
- W4313447762 hasConcept C37736160 @default.
- W4313447762 hasConcept C41008148 @default.
- W4313447762 hasConcept C50644808 @default.
- W4313447762 hasConcept C55493867 @default.
- W4313447762 hasConcept C63479239 @default.
- W4313447762 hasConceptScore W4313447762C104317684 @default.
- W4313447762 hasConceptScore W4313447762C126255220 @default.
- W4313447762 hasConceptScore W4313447762C134306372 @default.
- W4313447762 hasConceptScore W4313447762C154945302 @default.
- W4313447762 hasConceptScore W4313447762C185592680 @default.
- W4313447762 hasConceptScore W4313447762C186633575 @default.
- W4313447762 hasConceptScore W4313447762C2984842247 @default.
- W4313447762 hasConceptScore W4313447762C33923547 @default.
- W4313447762 hasConceptScore W4313447762C37736160 @default.
- W4313447762 hasConceptScore W4313447762C41008148 @default.
- W4313447762 hasConceptScore W4313447762C50644808 @default.
- W4313447762 hasConceptScore W4313447762C55493867 @default.
- W4313447762 hasConceptScore W4313447762C63479239 @default.
- W4313447762 hasLocation W43134477621 @default.
- W4313447762 hasOpenAccess W4313447762 @default.
- W4313447762 hasPrimaryLocation W43134477621 @default.
- W4313447762 hasRelatedWork W2903311573 @default.
- W4313447762 hasRelatedWork W2943368551 @default.
- W4313447762 hasRelatedWork W2950183588 @default.
- W4313447762 hasRelatedWork W2963894298 @default.
- W4313447762 hasRelatedWork W3072584680 @default.
- W4313447762 hasRelatedWork W3094843325 @default.
- W4313447762 hasRelatedWork W3207178610 @default.
- W4313447762 hasRelatedWork W3211782752 @default.
- W4313447762 hasRelatedWork W4303857474 @default.
- W4313447762 hasRelatedWork W4309156448 @default.
- W4313447762 isParatext "false" @default.
- W4313447762 isRetracted "false" @default.
- W4313447762 workType "article" @default.