FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3110934648> ?p ?o ?g. }

• W3110934648 abstract "Although deep neural networks have shown promising performances on various tasks, they are susceptible to incorrect predictions induced by imperceptibly small perturbations in inputs. A large number of previous works proposed to detect adversarial attacks. Yet, most of them cannot effectively detect them against adaptive whitebox attacks where an adversary has the knowledge of the model and the defense method. In this paper, we propose a new probabilistic adversarial detector motivated by a recently introduced non-robust feature. We consider the non-robust features as a common property of adversarial examples, and we deduce it is possible to find a cluster in representation space corresponding to the property. This idea leads us to probability estimate distribution of adversarial representations in a separate cluster, and leverage the distribution for a likelihood based adversarial detector." @default.
• W3110934648 created "2020-12-21" @default.
• W3110934648 creator A5008305493 @default.
• W3110934648 creator A5070302452 @default.
• W3110934648 creator A5077052539 @default.
• W3110934648 creator A5082893706 @default.
• W3110934648 creator A5085659523 @default.
• W3110934648 date "2020-12-07" @default.
• W3110934648 modified "2023-09-25" @default.
• W3110934648 title "Learning to Separate Clusters of Adversarial Representations for Robust Adversarial Detection" @default.
• W3110934648 cites W1522301498 @default.
• W3110934648 cites W1959608418 @default.
• W3110934648 cites W2078622638 @default.
• W3110934648 cites W2112796928 @default.
• W3110934648 cites W2166160300 @default.
• W3110934648 cites W2187089797 @default.
• W3110934648 cites W2194775991 @default.
• W3110934648 cites W2342045095 @default.
• W3110934648 cites W2543296129 @default.
• W3110934648 cites W2561975083 @default.
• W3110934648 cites W2590523583 @default.
• W3110934648 cites W2594867206 @default.
• W3110934648 cites W2594877703 @default.
• W3110934648 cites W2605631833 @default.
• W3110934648 cites W2607219512 @default.
• W3110934648 cites W2721006554 @default.
• W3110934648 cites W2760733685 @default.
• W3110934648 cites W2765384636 @default.
• W3110934648 cites W2767962654 @default.
• W3110934648 cites W2768927604 @default.
• W3110934648 cites W2774644650 @default.
• W3110934648 cites W2795725650 @default.
• W3110934648 cites W2796892552 @default.
• W3110934648 cites W2801079363 @default.
• W3110934648 cites W2802557767 @default.
• W3110934648 cites W2867167548 @default.
• W3110934648 cites W2888399003 @default.
• W3110934648 cites W2898963688 @default.
• W3110934648 cites W2911634294 @default.
• W3110934648 cites W2947133760 @default.
• W3110934648 cites W2950106672 @default.
• W3110934648 cites W2950468330 @default.
• W3110934648 cites W2950864148 @default.
• W3110934648 cites W2954629067 @default.
• W3110934648 cites W2962729158 @default.
• W3110934648 cites W2962759300 @default.
• W3110934648 cites W2962851953 @default.
• W3110934648 cites W2963034614 @default.
• W3110934648 cites W2963054787 @default.
• W3110934648 cites W2963062382 @default.
• W3110934648 cites W2963070423 @default.
• W3110934648 cites W2963076808 @default.
• W3110934648 cites W2963143631 @default.
• W3110934648 cites W2963154688 @default.
• W3110934648 cites W2963158386 @default.
• W3110934648 cites W2963207607 @default.
• W3110934648 cites W2963431851 @default.
• W3110934648 cites W2963446712 @default.
• W3110934648 cites W2963539306 @default.
• W3110934648 cites W2963556353 @default.
• W3110934648 cites W2963564844 @default.
• W3110934648 cites W2963600714 @default.
• W3110934648 cites W2963612069 @default.
• W3110934648 cites W2963747550 @default.
• W3110934648 cites W2963857521 @default.
• W3110934648 cites W2963952467 @default.
• W3110934648 cites W2964116600 @default.
• W3110934648 cites W2964153729 @default.
• W3110934648 cites W2964197269 @default.
• W3110934648 cites W2964253222 @default.
• W3110934648 cites W2964294232 @default.
• W3110934648 cites W2966108112 @default.
• W3110934648 cites W2970115835 @default.
• W3110934648 cites W2970971581 @default.
• W3110934648 cites W2971109239 @default.
• W3110934648 cites W2995106777 @default.
• W3110934648 cites W2995876497 @default.
• W3110934648 cites W2996296329 @default.
• W3110934648 cites W2996303980 @default.
• W3110934648 cites W2996690718 @default.
• W3110934648 cites W3013520104 @default.
• W3110934648 cites W3103340107 @default.
• W3110934648 cites W3105599650 @default.
• W3110934648 cites W3118608800 @default.
• W3110934648 cites W9657784 @default.
• W3110934648 doi "https://doi.org/10.48550/arxiv.2012.03483" @default.
• W3110934648 hasPublicationYear "2020" @default.
• W3110934648 type Work @default.
• W3110934648 sameAs 3110934648 @default.
• W3110934648 citedByCount "0" @default.
• W3110934648 crossrefType "posted-content" @default.
• W3110934648 hasAuthorship W3110934648A5008305493 @default.
• W3110934648 hasAuthorship W3110934648A5070302452 @default.
• W3110934648 hasAuthorship W3110934648A5077052539 @default.
• W3110934648 hasAuthorship W3110934648A5082893706 @default.
• W3110934648 hasAuthorship W3110934648A5085659523 @default.
• W3110934648 hasBestOaLocation W31109346481 @default.
• W3110934648 hasConcept C111472728 @default.
• W3110934648 hasConcept C119857082 @default.
• W3110934648 hasConcept C138885662 @default.

• next