AI Paper Atlas

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

Hiding Visual Information via Obfuscating Adversarial Perturbations

Zhigang Su, Dawei Zhou, Nannan Wang, Decheng Liu, Zhen Wang, Xinbo Gao

ICCV 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Diffusion-based adversarial purification · 16 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
replace distance loss L_d with variance consistency loss L_v (ablation)AVIH · LFW face imagesvisual hiding vs recovery qualityImproved
use variance consistency loss rather than distance loss for resisting external keys (ablation)AVIH · LFW face imagesrecoverability by other keysImproved
vary variance loss weight lambda from 0.001 to 0.01 (ablation)AVIH · LFW face imagesprotection and recovery qualityNo clear effect
increase iterations from 100 to 800 (ablation)AVIH · Iterations study (10 samples)SSIM of recovered image: 0.721 -> 0.912Improved
increase iterations from 600 to 800 (ablation)AVIH · Iterations study (10 samples)COS: 0.997 -> 0.997No clear effect
train key model on COCO instead of CelebA (ablation)AVIH · LFWaccuracy and SSIM_d: SphereFace 80.0 -> 79.9No clear effect
increase block size of variance loss with highly overlapping blocks (ablation)AVIH · LFW face imagesrecovery qualityHurt
protect gallery with AVIH, one imageFace recognition models · LFWaccuracy (%): SphereFace 80.3 -> 80.0No clear effect
replace all gallery images with AVIH-protected imagesSphereFace · LFWaccuracy (%): 80.3 -> 80.6Improved
AVIH versus ITP on classificationResNet50 · CIFAR-10accuracy (%): 90.16 -> 95.28Improved
recover protected images using a key model with different initializationAVIH · key model security, 16 modelsrecoverabilityHurt
add Gaussian noise to protected image up to sigma 10/255AVIH · LFWSSIM_d: 0.93 -> 0.81Hurt

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