AI Paper Atlas

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

Scaling up the Randomized Gradient-Free Adversarial Attack Reveals Overestimation of Robustness Using Established Attacks

Francesco Croce, Jonas Rauber, Matthias Hein

IJCV 2019 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness-accuracy trade-off analysis · 33 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
use the randomized gradient-free attack instead of PGD-10kPGD-10k · MNISTaverage difference to best l2 robust accuracy: 0.1011 -> 0.0051Improved
use the randomized gradient-free attack instead of PGD-10kPGD-10k · MNISTmaximum difference to best l2 robust accuracy: 0.5040 -> 0.0500Improved
use the randomized gradient-free attack instead of CW-100kCW-100k · GTSaverage difference to best l2 robust accuracy: 0.0172 -> 0Improved
use the randomized gradient-free attack instead of CW-100kCW-100k · CIFAR-10average difference to best l2 robust accuracy: 0.0037 -> 0.0060Hurt
use the randomized gradient-free attack instead of CW-100kCW-100k · CIFAR-10maximum difference to best l2 robust accuracy: 0.0180 -> 0.0140Improved
use 10000 random restarts instead of a single start for PGDPGD-1 · MNISTaverage difference to best l2 robust accuracy: 0.2367 -> 0.1011Improved
increase CW iterations from 10k to 100kCW-10k · MNISTaverage difference to best l2 robust accuracy: 0.1701 -> 0.1681No clear effect
use the randomized gradient-free attack instead of PGD-10k on an l_inf-adversarially trained modelPGD-10k · MNIST l_inf-at, eps=2.0l2 robust accuracy upper bound: 0.536 -> 0.032Improved
use the randomized gradient-free attack instead of DeepFoolDeepFool · MNIST l_inf-at, eps=2.0l2 robust accuracy upper bound: 0.918 -> 0.032Improved
use the randomized gradient-free attack instead of PGD-1k on provably robust l_inf-MMR modelPGD-1k · MNIST l_inf-MMR-at, eps=2.0l2 robust accuracy upper bound: 0.310 -> 0.048Improved
use the randomized gradient-free attack instead of PGD-1k on l_inf-KW modelPGD-1k · MNIST l_inf-KW, eps=2.0l2 robust accuracy upper bound: 0.226 -> 0.148Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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