Matches in SemOpenAlex for { <https://semopenalex.org/work/W3096024872> ?p ?o ?g. }
- W3096024872 abstract "This paper proposes a new defense against neural network backdooring attacks that are maliciously trained to mispredict in the presence of attacker-chosen triggers. Our defense is based on the intuition that the feature extraction layers of a backdoored network embed new features to detect the presence of a trigger and the subsequent classification layers learn to mispredict when triggers are detected. Therefore, to detect backdoors, the proposed defense uses two synergistic anomaly detectors trained on clean validation data: the first is a novelty detector that checks for anomalous features, while the second detects anomalous mappings from features to outputs by comparing with a separate classifier trained on validation data. The approach is evaluated on a wide range of backdoored networks (with multiple variations of triggers) that successfully evade state-of-the-art defenses. Additionally, we evaluate the robustness of our approach on imperceptible perturbations, scalability on large-scale datasets, and effectiveness under domain shift. This paper also shows that the defense can be further improved using data augmentation." @default.
- W3096024872 created "2020-11-09" @default.
- W3096024872 creator A5010950688 @default.
- W3096024872 creator A5025135982 @default.
- W3096024872 creator A5031351839 @default.
- W3096024872 creator A5054769060 @default.
- W3096024872 creator A5082413942 @default.
- W3096024872 date "2020-11-04" @default.
- W3096024872 modified "2023-09-27" @default.
- W3096024872 title "Detecting Backdoors in Neural Networks Using Novel Feature-Based Anomaly Detection." @default.
- W3096024872 cites W1998808035 @default.
- W3096024872 cites W2019464758 @default.
- W3096024872 cites W2067713319 @default.
- W3096024872 cites W2103018059 @default.
- W3096024872 cites W2108598243 @default.
- W3096024872 cites W2119112357 @default.
- W3096024872 cites W2132870739 @default.
- W3096024872 cites W2144182447 @default.
- W3096024872 cites W2145287260 @default.
- W3096024872 cites W2158899491 @default.
- W3096024872 cites W2517229335 @default.
- W3096024872 cites W2543927648 @default.
- W3096024872 cites W2774423163 @default.
- W3096024872 cites W2798302089 @default.
- W3096024872 cites W2804964665 @default.
- W3096024872 cites W2867167548 @default.
- W3096024872 cites W2900018096 @default.
- W3096024872 cites W2921870669 @default.
- W3096024872 cites W2934843808 @default.
- W3096024872 cites W2963037989 @default.
- W3096024872 cites W2963143631 @default.
- W3096024872 cites W2963207607 @default.
- W3096024872 cites W2963431851 @default.
- W3096024872 cites W2963446712 @default.
- W3096024872 cites W2963771448 @default.
- W3096024872 cites W2963857521 @default.
- W3096024872 cites W2963911037 @default.
- W3096024872 cites W2964041528 @default.
- W3096024872 cites W2964153729 @default.
- W3096024872 cites W2964222437 @default.
- W3096024872 cites W2964308564 @default.
- W3096024872 cites W2966187620 @default.
- W3096024872 cites W2970200861 @default.
- W3096024872 cites W2979332623 @default.
- W3096024872 cites W2979417040 @default.
- W3096024872 cites W3003676953 @default.
- W3096024872 cites W3035808436 @default.
- W3096024872 cites W3118608800 @default.
- W3096024872 cites W639708223 @default.
- W3096024872 hasPublicationYear "2020" @default.
- W3096024872 type Work @default.
- W3096024872 sameAs 3096024872 @default.
- W3096024872 citedByCount "1" @default.
- W3096024872 countsByYear W30960248722021 @default.
- W3096024872 crossrefType "posted-content" @default.
- W3096024872 hasAuthorship W3096024872A5010950688 @default.
- W3096024872 hasAuthorship W3096024872A5025135982 @default.
- W3096024872 hasAuthorship W3096024872A5031351839 @default.
- W3096024872 hasAuthorship W3096024872A5054769060 @default.
- W3096024872 hasAuthorship W3096024872A5082413942 @default.
- W3096024872 hasConcept C104317684 @default.
- W3096024872 hasConcept C111472728 @default.
- W3096024872 hasConcept C119857082 @default.
- W3096024872 hasConcept C124101348 @default.
- W3096024872 hasConcept C132010649 @default.
- W3096024872 hasConcept C138885662 @default.
- W3096024872 hasConcept C153180895 @default.
- W3096024872 hasConcept C154945302 @default.
- W3096024872 hasConcept C185592680 @default.
- W3096024872 hasConcept C27206212 @default.
- W3096024872 hasConcept C2776145971 @default.
- W3096024872 hasConcept C2778738651 @default.
- W3096024872 hasConcept C2778924833 @default.
- W3096024872 hasConcept C41008148 @default.
- W3096024872 hasConcept C48044578 @default.
- W3096024872 hasConcept C50644808 @default.
- W3096024872 hasConcept C52622490 @default.
- W3096024872 hasConcept C55493867 @default.
- W3096024872 hasConcept C63479239 @default.
- W3096024872 hasConcept C739882 @default.
- W3096024872 hasConcept C76155785 @default.
- W3096024872 hasConcept C77088390 @default.
- W3096024872 hasConcept C94915269 @default.
- W3096024872 hasConcept C95623464 @default.
- W3096024872 hasConceptScore W3096024872C104317684 @default.
- W3096024872 hasConceptScore W3096024872C111472728 @default.
- W3096024872 hasConceptScore W3096024872C119857082 @default.
- W3096024872 hasConceptScore W3096024872C124101348 @default.
- W3096024872 hasConceptScore W3096024872C132010649 @default.
- W3096024872 hasConceptScore W3096024872C138885662 @default.
- W3096024872 hasConceptScore W3096024872C153180895 @default.
- W3096024872 hasConceptScore W3096024872C154945302 @default.
- W3096024872 hasConceptScore W3096024872C185592680 @default.
- W3096024872 hasConceptScore W3096024872C27206212 @default.
- W3096024872 hasConceptScore W3096024872C2776145971 @default.
- W3096024872 hasConceptScore W3096024872C2778738651 @default.
- W3096024872 hasConceptScore W3096024872C2778924833 @default.
- W3096024872 hasConceptScore W3096024872C41008148 @default.
- W3096024872 hasConceptScore W3096024872C48044578 @default.
- W3096024872 hasConceptScore W3096024872C50644808 @default.