FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4296519903> ?p ?o ?g. }

• W4296519903 endingPage "28" @default.
• W4296519903 startingPage "13" @default.
• W4296519903 abstract "Adversarial attacks pose a security challenge for deep neural networks, motivating researchers to build various defense methods. Consequently, the performance of black-box attacks turns down under defense scenarios. A significant observation is that some feature-level attacks achieve an excellent success rate to fool undefended models, while their transferability is severely degraded when encountering defenses, which give a false sense of security. In this paper, we explain one possible reason caused this phenomenon is the domain-overfitting effect, which degrades the capabilities of feature perturbed images and makes them hardly fool adversarially trained defenses. To this end, we study a novel feature-level method, referred to as Decoupled Feature Attack (DEFEAT). Unlike the current attacks that use a round-robin procedure to estimate gradient estimation and update perturbation, DEFEAT decouples adversarial example generation from the optimization process. In the first stage, DEFEAT learns an distribution full of perturbations with high adversarial effects. And it then iteratively samples the noises from learned distribution to assemble adversarial examples. On top of that, we can apply transformations of existing methods into the DEFEAT framework to produce more robust perturbations. We also provide insights into the relationship between transferability and latent features that helps the community to understand the intrinsic mechanism of adversarial attacks. Extensive experiments evaluated on a variety of black-box models suggest the superiority of DEFEAT, i.e., our method fools defenses at an average success rate of 88.4%, remarkably outperforming state-of-the-art transferable attacks by a large margin of 11.5%. The code is publicly available at https://github.com/mesunhlf/DEFEAT." @default.
• W4296519903 created "2022-09-21" @default.
• W4296519903 creator A5027058176 @default.
• W4296519903 creator A5029844813 @default.
• W4296519903 creator A5040743941 @default.
• W4296519903 date "2022-12-01" @default.
• W4296519903 modified "2023-10-17" @default.
• W4296519903 title "DEFEAT: Decoupled feature attack across deep neural networks" @default.
• W4296519903 cites W1849277567 @default.
• W4296519903 cites W2183341477 @default.
• W4296519903 cites W2194775991 @default.
• W4296519903 cites W2543927648 @default.
• W4296519903 cites W2746600820 @default.
• W4296519903 cites W2774018344 @default.
• W4296519903 cites W2774644650 @default.
• W4296519903 cites W2896078964 @default.
• W4296519903 cites W2925582157 @default.
• W4296519903 cites W2947874337 @default.
• W4296519903 cites W2962847335 @default.
• W4296519903 cites W2962858109 @default.
• W4296519903 cites W2962933288 @default.
• W4296519903 cites W2963446712 @default.
• W4296519903 cites W2963771536 @default.
• W4296519903 cites W2963857521 @default.
• W4296519903 cites W2964171870 @default.
• W4296519903 cites W2964350391 @default.
• W4296519903 cites W2969542116 @default.
• W4296519903 cites W2984699060 @default.
• W4296519903 cites W2987019399 @default.
• W4296519903 cites W2990237009 @default.
• W4296519903 cites W2997583194 @default.
• W4296519903 cites W3017485054 @default.
• W4296519903 cites W3018911559 @default.
• W4296519903 cites W3034190247 @default.
• W4296519903 cites W3091154097 @default.
• W4296519903 cites W3140102771 @default.
• W4296519903 cites W3150919918 @default.
• W4296519903 cites W3186991201 @default.
• W4296519903 cites W4200612769 @default.
• W4296519903 cites W4281698656 @default.
• W4296519903 cites W4282963876 @default.
• W4296519903 doi "https://doi.org/10.1016/j.neunet.2022.09.009" @default.
• W4296519903 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/36228335" @default.
• W4296519903 hasPublicationYear "2022" @default.
• W4296519903 type Work @default.
• W4296519903 citedByCount "0" @default.
• W4296519903 crossrefType "journal-article" @default.
• W4296519903 hasAuthorship W4296519903A5027058176 @default.
• W4296519903 hasAuthorship W4296519903A5029844813 @default.
• W4296519903 hasAuthorship W4296519903A5040743941 @default.
• W4296519903 hasConcept C111919701 @default.
• W4296519903 hasConcept C119857082 @default.
• W4296519903 hasConcept C138885662 @default.
• W4296519903 hasConcept C140331021 @default.
• W4296519903 hasConcept C153180895 @default.
• W4296519903 hasConcept C154945302 @default.
• W4296519903 hasConcept C22019652 @default.
• W4296519903 hasConcept C2776401178 @default.
• W4296519903 hasConcept C2984842247 @default.
• W4296519903 hasConcept C37736160 @default.
• W4296519903 hasConcept C41008148 @default.
• W4296519903 hasConcept C41895202 @default.
• W4296519903 hasConcept C50644808 @default.
• W4296519903 hasConcept C61272859 @default.
• W4296519903 hasConcept C774472 @default.
• W4296519903 hasConcept C94966114 @default.
• W4296519903 hasConcept C98045186 @default.
• W4296519903 hasConceptScore W4296519903C111919701 @default.
• W4296519903 hasConceptScore W4296519903C119857082 @default.
• W4296519903 hasConceptScore W4296519903C138885662 @default.
• W4296519903 hasConceptScore W4296519903C140331021 @default.
• W4296519903 hasConceptScore W4296519903C153180895 @default.
• W4296519903 hasConceptScore W4296519903C154945302 @default.
• W4296519903 hasConceptScore W4296519903C22019652 @default.
• W4296519903 hasConceptScore W4296519903C2776401178 @default.
• W4296519903 hasConceptScore W4296519903C2984842247 @default.
• W4296519903 hasConceptScore W4296519903C37736160 @default.
• W4296519903 hasConceptScore W4296519903C41008148 @default.
• W4296519903 hasConceptScore W4296519903C41895202 @default.
• W4296519903 hasConceptScore W4296519903C50644808 @default.
• W4296519903 hasConceptScore W4296519903C61272859 @default.
• W4296519903 hasConceptScore W4296519903C774472 @default.
• W4296519903 hasConceptScore W4296519903C94966114 @default.
• W4296519903 hasConceptScore W4296519903C98045186 @default.
• W4296519903 hasFunder F4320321001 @default.
• W4296519903 hasFunder F4320337111 @default.
• W4296519903 hasLocation W42965199031 @default.
• W4296519903 hasLocation W42965199032 @default.
• W4296519903 hasOpenAccess W4296519903 @default.
• W4296519903 hasPrimaryLocation W42965199031 @default.
• W4296519903 hasRelatedWork W2901701333 @default.
• W4296519903 hasRelatedWork W2914158293 @default.
• W4296519903 hasRelatedWork W2963923490 @default.
• W4296519903 hasRelatedWork W3120578866 @default.
• W4296519903 hasRelatedWork W3156525504 @default.
• W4296519903 hasRelatedWork W3206423720 @default.
• W4296519903 hasRelatedWork W4213022598 @default.
• W4296519903 hasRelatedWork W4292794158 @default.

• next