AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Bit-Flip Attack: Crushing Neural Network With Progressive Bit Search

Adnan Siraj Rakin, Zhezhi He, Deliang Fan

ICCV 2019 · Adversarial robustness and security › Attacks on text and graphs · approach: Attacks and defenses in sequential settings · 298 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 progressive bit search instead of random bit flips (ablation)bit-flip attack on 8-bit ResNet-18 · ImageNettop-1 accuracy: 100 random flips cost under 1% vs 13 PBS flips reach 0.1%Improved
attack only the first conv layer instead of only the last linear layer (ablation)BFA on VGG16 · CIFAR-10top-1 accuracy (%): 84.61 -> 10.06Improved
use attack sample size 128 instead of 16 (ablation)BFA with PBS · CIFAR-10 / ImageNetbit-flips neededImproved
use a very large or very small attack sample size (ablation)BFA with PBS · ImageNet ResNet-18attack strengthMixed
randomly flip one bit in a full-precision floating-point weight (ablation)ResNet-18 · ImageNetaccuracyHurt
lower quantization bit-width from 8 to 4BFA with PBS · CIFAR-10accuracy after attackNo clear effect
increase network depth/params from ResNet-18 to ResNet-34BFA with PBS · ImageNetbit-flips to crush: 13 -> 11Improved

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