Matches in SemOpenAlex for { <https://semopenalex.org/work/W2997131184> ?p ?o ?g. }
Showing items 1 to 74 of
74
with 100 items per page.
- W2997131184 abstract "We propose the scheme that mitigates the adversarial perturbation $epsilon$ on the adversarial example $X_{adv}$ ($=$ $X$ $pm$ $epsilon$, $X$ is a benign sample) by subtracting the estimated perturbation $hat{epsilon}$ from $X$ $+$ $epsilon$ and adding $hat{epsilon}$ to $X$ $-$ $epsilon$. The estimated perturbation $hat{epsilon}$ comes from the difference between $X_{adv}$ and its moving-averaged outcome $W_{avg}*X_{adv}$ where $W_{avg}$ is $N times N$ moving average kernel that all the coefficients are one. Usually, the adjacent samples of an image are close to each other such that we can let $X$ $approx$ $W_{avg}*X$ (naming this relation after X-MAS[X minus Moving Averaged Samples]). By doing that, we can make the estimated perturbation $hat{epsilon}$ falls within the range of $epsilon$. The scheme is also extended to do the multi-level mitigation by configuring the mitigated adversarial example $X_{adv}$ $pm$ $hat{epsilon}$ as a new adversarial example to be mitigated. The multi-level mitigation gets $X_{adv}$ closer to $X$ with a smaller (i.e. mitigated) perturbation than original unmitigated perturbation by setting the moving averaged adversarial sample $W_{avg} * X_{adv}$ (which has the smaller perturbation than $X_{adv}$ if $X$ $approx$ $W_{avg}*X$) as the boundary condition that the multi-level mitigation cannot cross over (i.e. decreasing $epsilon$ cannot go below and increasing $epsilon$ cannot go beyond). With the multi-level mitigation, we can get high prediction accuracies even in the adversarial example having a large perturbation (i.e. $epsilon$ $>$ $16$). The proposed scheme is evaluated with adversarial examples crafted by the FGSM (Fast Gradient Sign Method) based attacks on ResNet-50 trained with ImageNet dataset." @default.
- W2997131184 created "2020-01-10" @default.
- W2997131184 creator A5040004945 @default.
- W2997131184 creator A5066063772 @default.
- W2997131184 creator A5081950684 @default.
- W2997131184 creator A5090081544 @default.
- W2997131184 date "2019-12-19" @default.
- W2997131184 modified "2023-09-25" @default.
- W2997131184 title "Mitigating large adversarial perturbations on X-MAS (X minus Moving Averaged Samples)" @default.
- W2997131184 cites W2108598243 @default.
- W2997131184 cites W2174868984 @default.
- W2997131184 cites W2503523779 @default.
- W2997131184 cites W2516574342 @default.
- W2997131184 cites W2786118190 @default.
- W2997131184 cites W2891719429 @default.
- W2997131184 cites W2919386767 @default.
- W2997131184 cites W2949650786 @default.
- W2997131184 cites W2962933288 @default.
- W2997131184 cites W2963001136 @default.
- W2997131184 cites W2963207607 @default.
- W2997131184 cites W2963389226 @default.
- W2997131184 cites W2963542245 @default.
- W2997131184 doi "https://doi.org/10.48550/arxiv.1912.12170" @default.
- W2997131184 hasPublicationYear "2019" @default.
- W2997131184 type Work @default.
- W2997131184 sameAs 2997131184 @default.
- W2997131184 citedByCount "0" @default.
- W2997131184 crossrefType "posted-content" @default.
- W2997131184 hasAuthorship W2997131184A5040004945 @default.
- W2997131184 hasAuthorship W2997131184A5066063772 @default.
- W2997131184 hasAuthorship W2997131184A5081950684 @default.
- W2997131184 hasAuthorship W2997131184A5090081544 @default.
- W2997131184 hasBestOaLocation W29971311841 @default.
- W2997131184 hasConcept C111919701 @default.
- W2997131184 hasConcept C114614502 @default.
- W2997131184 hasConcept C121332964 @default.
- W2997131184 hasConcept C134306372 @default.
- W2997131184 hasConcept C154945302 @default.
- W2997131184 hasConcept C177918212 @default.
- W2997131184 hasConcept C2777894999 @default.
- W2997131184 hasConcept C33923547 @default.
- W2997131184 hasConcept C37736160 @default.
- W2997131184 hasConcept C37914503 @default.
- W2997131184 hasConcept C41008148 @default.
- W2997131184 hasConcept C62520636 @default.
- W2997131184 hasConceptScore W2997131184C111919701 @default.
- W2997131184 hasConceptScore W2997131184C114614502 @default.
- W2997131184 hasConceptScore W2997131184C121332964 @default.
- W2997131184 hasConceptScore W2997131184C134306372 @default.
- W2997131184 hasConceptScore W2997131184C154945302 @default.
- W2997131184 hasConceptScore W2997131184C177918212 @default.
- W2997131184 hasConceptScore W2997131184C2777894999 @default.
- W2997131184 hasConceptScore W2997131184C33923547 @default.
- W2997131184 hasConceptScore W2997131184C37736160 @default.
- W2997131184 hasConceptScore W2997131184C37914503 @default.
- W2997131184 hasConceptScore W2997131184C41008148 @default.
- W2997131184 hasConceptScore W2997131184C62520636 @default.
- W2997131184 hasLocation W29971311841 @default.
- W2997131184 hasOpenAccess W2997131184 @default.
- W2997131184 hasPrimaryLocation W29971311841 @default.
- W2997131184 hasRelatedWork W1978042415 @default.
- W2997131184 hasRelatedWork W2017331178 @default.
- W2997131184 hasRelatedWork W2060680670 @default.
- W2997131184 hasRelatedWork W2114987976 @default.
- W2997131184 hasRelatedWork W2123829666 @default.
- W2997131184 hasRelatedWork W2976797620 @default.
- W2997131184 hasRelatedWork W2997131184 @default.
- W2997131184 hasRelatedWork W3086542228 @default.
- W2997131184 hasRelatedWork W3103234735 @default.
- W2997131184 hasRelatedWork W4301484659 @default.
- W2997131184 isParatext "false" @default.
- W2997131184 isRetracted "false" @default.
- W2997131184 magId "2997131184" @default.
- W2997131184 workType "article" @default.