AI Paper Atlas

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

Adversarial Robustness of Discriminative Self-Supervised Learning in Vision

Ömer Veysel Çağatan, Ömer Faruk Tal, M. Emre Gursoy

ICCV 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training recipes · 1 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 self-supervised pretraining instead of supervised pretraining (frozen backbone)ResNet50 · ImageNet adversarial robustnessaccuracy under attack: 34.5 (supervised) vs 40.5 (MoCoV3), 32.6 (SwAV), 32.1 (SimCLR)Mixed
add non-linear projector and stronger augmentation (MoCoV1 to MoCoV2) (ablation)MoCo · ImageNet overall adversarialaccuracy drop (%): 71 -> 52Improved
add prediction head and larger batch size (MoCoV2 to MoCoV3) (ablation)MoCo · ImageNet overall adversarialaccuracy drop (%): 52 -> 46Improved
use ViT backbone instead of ResNet (ablation)DINO · ImageNet overall adversarialaccuracy drop (%): 55% -> 42% dropImproved
use ViT backbone instead of ResNet (ablation)MoCoV3 · ImageNet overall adversarialaccuracy drop (%): 46% -> 61% dropHurt
train longer (more epochs) (ablation)SwAV and MoCoV3 · ImageNet IAAadversarial robustness: about 1% marginal improvementNo clear effect
train longer (beyond 100 epochs) (ablation)SwAV and MoCoV3 · ImageNet UAPadversarial accuracy (%): SwAV 33 -> 40; MoCoV3 41 -> 48Improved
use full augmentation set instead of limited augmentation (ablation)BYOL · ImageNet UAPaccuracy drop (%): BYOL-NC 46, BYOL-CC 48 -> standard 39Improved
vary batch size (128 vs 512 vs standard) (ablation)BYOL · ImageNet IAA/UAPaccuracy drop (%): IAA 57, 56, 56; UAP 41, 43, 39No clear effect
fine-tune instead of linear probeSSL and supervised ResNet50 · Transfer learning (adversarial drop)accuracy drop (%): IAA 60 -> 49; UAP 49 -> 43Improved
fine-tune instead of linear probeImageNet-to-downstream IAA robustness correlation · Transfer learningSpearman correlation of drops: 0.79 -> 0.12Hurt
fine-tune instead of linear probeSwAV · CIFAR-10 FGSMaccuracy drop: 79% -> 17%Improved

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