AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

Generalizable Data-Free Objective for Crafting Universal Adversarial Perturbations

Konda Reddy Mopuri, Aditya Ganeshan, R. Venkatesh Babu

TPAMI 2018 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transfer attacks on face recognition · 228 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
add range prior (mean and dynamic range) (ablation)GD-UAP on GoogLeNet · ILSVRC fooling ratefooling rate (%): 58.62 -> 71.44Improved
add training data samples as prior (ablation)GD-UAP on GoogLeNet · ILSVRC fooling ratefooling rate (%): 71.44 -> 83.54Improved
add training data samples as prior (ablation)GD-UAP on ResNet-152 · ILSVRC fooling ratefooling rate (%): 37.3 -> 66.68Improved
use data w/ less background instead of all data (ablation)GD-UAP on FCN-8s-VGG · Pascal VOC segmentationGFR (%): 61.15 -> 67.19Improved
use data w/ less background instead of all data (ablation)GD-UAP on DL-RN101 · Pascal VOC segmentationGFR (%): 58.62 -> 56.61Hurt
use l2-norm activation objective instead of mean activation (ablation)GD-UAP with data on CaffeNet · ILSVRC fooling ratefooling rate (%): 88.35 -> 91.54Improved
use UAP crafted on mismatched data (Places-205 to ILSVRC)UAP on GoogLeNet · ILSVRC fooling ratefooling rate (%): 78.5 -> 28.17Hurt
use range prior noise instead of real data for UAPUAP on GoogLeNet · ILSVRC fooling ratefooling rate (%): 78.5 -> 10.56Hurt
add data prior over range prior on depth estimation (ablation)GD-UAP on Monodepth-ResNet50 · KITTI GFR on delta<1.25GFR (%): 47.6% -> 24.3%Hurt
apply PRN defense against UAP perturbationsGD-UAP data prior on GoogLeNet · ILSVRC fooling ratefooling rate (%): 83.54 -> 52.50Hurt
use scattering-based hybrid network instead of ResNet-18GD-UAP transfer attack from VGG-13, range prior · ILSVRC fooling ratefooling rate (%): 39.14 -> 30.96Hurt

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