AI Paper Atlas

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

Extending Adversarial Attacks to Produce Adversarial Class Probability Distributions

Jon Vadillo, Roberto Santana, Jose A. Lozano

JMLR 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Attack evaluation and defense critique

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
use informed transition matrices (EWTM) instead of AMtarget distribution approximation · Speech Commands, DeepFoolmax/mean abs diff, KL, SpearmanImproved
use MAB baseline instead of proposed AM/UBM(eps>0.01)/CRMfooling rate · Speech Commands, DeepFool, eps=0.15fooling rate (%): 94.69 -> 79.64Hurt
use EWTM instead of AMfooling rate · Speech Commands, DeepFool, eps=0.15fooling rate (%): 94.69 -> 87.84Hurt
add upper-bound relaxation via eta to UBM (ablation)UBM · Speech Commands, DeepFool, eps=0.01valid matrix success (%): 67.9 -> 100.0Improved
add Laplace smoothing to CRM (ablation)CRM · Speech Commands, DeepFool, eps=0.05valid matrix success (%): 14.0 -> 70.6Improved
add zeroing of P({y_i}|y_i) to CRM (ablation)CRM · Speech Commands, DeepFool, eps=0.05valid matrix success (%): 70.6 -> 100.0Improved
increase maximum distortion eps (ablation)all methods · Speech Commands, DeepFooltarget approximation: max abs diff below 0.09, mean abs diff 0.03, Spearman above 0.80 for eps >= 0.05Improved
use EWTM instead of other methods at high distortionEWTM · Speech Commands, DeepFool, eps>=0.05distribution similarityImproved
increase eps when reproducing the original distribution (ablation)all methods · Speech Commands, DeepFool, original distributionapproximation accuracyHurt
reduce samples per class N to 1 (ablation)EWTM · Speech Commands, DeepFool, eps=0.01valid matrix success (%): 100 -> 0.8Hurt
reduce samples per class N to 50 (ablation)AM/UBM/EWTM · Speech Commands, DeepFool, eps=0.01effectivenessNo clear effect
use sparse-reach attack FGSM instead of C&W/PGDproposed methods · Speech Commands, DeepFooldifferences between methodsMixed

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