Matches in SemOpenAlex for { <https://semopenalex.org/work/W3034190247> ?p ?o ?g. }
- W3034190247 abstract "Adversarial examples can cause catastrophic mistakes in Deep Neural Network (DNNs) based vision systems e.g., for classification, segmentation and object detection. The vulnerability of DNNs against such attacks can prove a major roadblock towards their real-world deployment. Transferability of adversarial examples demand generalizable defenses that can provide cross-task protection. Adversarial training that enhances robustness by modifying target model's parameters lacks such generalizability. On the other hand, different input processing based defenses fall short in the face of continuously evolving attacks. In this paper, we take the first step to combine the benefits of both approaches and propose a self-supervised adversarial training mechanism in the input space. By design, our defense is a generalizable approach and provides significant robustness against the textbf{unseen} adversarial attacks (eg by reducing the success rate of translation-invariant textbf{ensemble} attack from 82.6% to 31.9% in comparison to previous state-of-the-art). It can be deployed as a plug-and-play solution to protect a variety of vision systems, as we demonstrate for the case of classification, segmentation and detection." @default.
- W3034190247 created "2020-06-19" @default.
- W3034190247 creator A5000300751 @default.
- W3034190247 creator A5038550641 @default.
- W3034190247 creator A5041437150 @default.
- W3034190247 creator A5079112987 @default.
- W3034190247 creator A5083105774 @default.
- W3034190247 date "2020-06-01" @default.
- W3034190247 modified "2023-10-06" @default.
- W3034190247 title "A Self-supervised Approach for Adversarial Robustness" @default.
- W3034190247 cites W2117539524 @default.
- W3034190247 cites W2171943915 @default.
- W3034190247 cites W2183341477 @default.
- W3034190247 cites W2194775991 @default.
- W3034190247 cites W2618530766 @default.
- W3034190247 cites W2774018344 @default.
- W3034190247 cites W2774644650 @default.
- W3034190247 cites W2887603965 @default.
- W3034190247 cites W2962785568 @default.
- W3034190247 cites W2963150697 @default.
- W3034190247 cites W2963312584 @default.
- W3034190247 cites W2963470893 @default.
- W3034190247 cites W2963626858 @default.
- W3034190247 cites W2963857521 @default.
- W3034190247 cites W2963881378 @default.
- W3034190247 cites W2965669158 @default.
- W3034190247 cites W2969542116 @default.
- W3034190247 doi "https://doi.org/10.1109/cvpr42600.2020.00034" @default.
- W3034190247 hasPublicationYear "2020" @default.
- W3034190247 type Work @default.
- W3034190247 sameAs 3034190247 @default.
- W3034190247 citedByCount "77" @default.
- W3034190247 countsByYear W30341902472020 @default.
- W3034190247 countsByYear W30341902472021 @default.
- W3034190247 countsByYear W30341902472022 @default.
- W3034190247 countsByYear W30341902472023 @default.
- W3034190247 crossrefType "proceedings-article" @default.
- W3034190247 hasAuthorship W3034190247A5000300751 @default.
- W3034190247 hasAuthorship W3034190247A5038550641 @default.
- W3034190247 hasAuthorship W3034190247A5041437150 @default.
- W3034190247 hasAuthorship W3034190247A5079112987 @default.
- W3034190247 hasAuthorship W3034190247A5083105774 @default.
- W3034190247 hasBestOaLocation W30341902472 @default.
- W3034190247 hasConcept C104317684 @default.
- W3034190247 hasConcept C105339364 @default.
- W3034190247 hasConcept C105795698 @default.
- W3034190247 hasConcept C108583219 @default.
- W3034190247 hasConcept C111919701 @default.
- W3034190247 hasConcept C119857082 @default.
- W3034190247 hasConcept C140331021 @default.
- W3034190247 hasConcept C154945302 @default.
- W3034190247 hasConcept C185592680 @default.
- W3034190247 hasConcept C27158222 @default.
- W3034190247 hasConcept C2776151529 @default.
- W3034190247 hasConcept C2984842247 @default.
- W3034190247 hasConcept C33923547 @default.
- W3034190247 hasConcept C37736160 @default.
- W3034190247 hasConcept C41008148 @default.
- W3034190247 hasConcept C55493867 @default.
- W3034190247 hasConcept C61272859 @default.
- W3034190247 hasConcept C63479239 @default.
- W3034190247 hasConcept C89600930 @default.
- W3034190247 hasConcept C94124525 @default.
- W3034190247 hasConceptScore W3034190247C104317684 @default.
- W3034190247 hasConceptScore W3034190247C105339364 @default.
- W3034190247 hasConceptScore W3034190247C105795698 @default.
- W3034190247 hasConceptScore W3034190247C108583219 @default.
- W3034190247 hasConceptScore W3034190247C111919701 @default.
- W3034190247 hasConceptScore W3034190247C119857082 @default.
- W3034190247 hasConceptScore W3034190247C140331021 @default.
- W3034190247 hasConceptScore W3034190247C154945302 @default.
- W3034190247 hasConceptScore W3034190247C185592680 @default.
- W3034190247 hasConceptScore W3034190247C27158222 @default.
- W3034190247 hasConceptScore W3034190247C2776151529 @default.
- W3034190247 hasConceptScore W3034190247C2984842247 @default.
- W3034190247 hasConceptScore W3034190247C33923547 @default.
- W3034190247 hasConceptScore W3034190247C37736160 @default.
- W3034190247 hasConceptScore W3034190247C41008148 @default.
- W3034190247 hasConceptScore W3034190247C55493867 @default.
- W3034190247 hasConceptScore W3034190247C61272859 @default.
- W3034190247 hasConceptScore W3034190247C63479239 @default.
- W3034190247 hasConceptScore W3034190247C89600930 @default.
- W3034190247 hasConceptScore W3034190247C94124525 @default.
- W3034190247 hasLocation W30341902471 @default.
- W3034190247 hasLocation W30341902472 @default.
- W3034190247 hasOpenAccess W3034190247 @default.
- W3034190247 hasPrimaryLocation W30341902471 @default.
- W3034190247 hasRelatedWork W10050540 @default.
- W3034190247 hasRelatedWork W11738893 @default.
- W3034190247 hasRelatedWork W12219208 @default.
- W3034190247 hasRelatedWork W3865299 @default.
- W3034190247 hasRelatedWork W4112095 @default.
- W3034190247 hasRelatedWork W4972971 @default.
- W3034190247 hasRelatedWork W5470710 @default.
- W3034190247 hasRelatedWork W7619760 @default.
- W3034190247 hasRelatedWork W9657784 @default.
- W3034190247 hasRelatedWork W13761807 @default.
- W3034190247 isParatext "false" @default.
- W3034190247 isRetracted "false" @default.
- W3034190247 magId "3034190247" @default.