AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Certified Defense to Image Transformations via Randomized Smoothing

Marc Fischer, Maximilian Baader, Martin Vechev

NeurIPS 2020 · Adversarial robustness and security › Certified and empirical robustness · approach: Randomized smoothing certification · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply BaseSPT smoothing to the base classifierMNIST classifier · MNIST rotation 30deg, worst-of-100adversarial accuracy: 0.73 -> 0.99Improved
apply BaseSPT smoothing to the base classifierCIFAR-10 classifier · CIFAR-10 rotation 30degadversarial accuracy: 0.26 -> 0.85Improved
apply BaseSPT smoothing to the base classifierImageNet classifier · ImageNet rotation 30degadversarial accuracy: 0.56 -> 0.76Improved
apply BaseSPT smoothing under translationMNIST classifier · MNIST translation 4pxadversarial accuracy: 0.03 -> 0.53Improved
increase n_gamma from 50 to 200DistSPT global · CIFAR-10 rotationcertified accuracy at 0deg: 0.31 -> 0.32Improved
increase n_gamma from 50 to 200DistSPT global · CIFAR-10 rotationrun time (s): 89.75 -> 351.47Hurt
raise E from 1.20 to 1.35 (ablation)DistSPT global · RImageNet rotationcertified accuracy at 0deg: 0.74 -> 0.64Hurt
raise E from 0.95 to 1.35 (ablation)DistSPT global · ImageNet rotationcertified accuracy at 0deg: 0.30 -> 0.16Hurt
use per-image error bound E (DistSPT^x) instead of global EDistSPT · CIFAR-10 rotation 0degcertified accuracy: 0.31 -> 0.35Improved
use per-image error bound E (DistSPT^x) instead of global EDistSPT · ImageNet rotation 0degcertified accuracy: 0.23 -> 0.31Improved
add vignetting to base model training/eval (ablation)DistSPT global base model · CIFAR-10 rotationbase accuracy: 0.34 -> 0.56Improved
add Gaussian blur (ablation)DistSPT global base model · ImageNet rotationbase accuracy: 0.38 -> 0.32Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy