AI Paper Atlas

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

Certified but Fooled! Breaking Certified Defenses with Ghost Certificates

Quoc Viet Vo, Tashreque Mohammed Haq, Paul Montague, Tamas Abraham, Ehsan Abbasnejad, Damith C. Ranasinghe

AAAI 2026 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness to poisoning and multiple attacks · 0 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
restrict perturbation to SAM+GradCAM salient-region mask (GhostCert) instead of global perturbation (Shadow Attack)Randomized Smoothing ensemble · ImageNet untargetedattack success rate (%): 40% -> ~100% (sigma=0.25)Improved
GhostCert vs Shadow Attack in targeted attack on ensembleRandomized Smoothing ensemble · ImageNet targetedattack success rate (%): 20% -> over 80% (sigma=0.25)Improved
GhostCert vs Shadow Attack on DensePureDensePure · ImageNet untargeted sigma=0.25spoofed radiusNo clear effect
GhostCert vs Shadow Attack on DensePureDensePure · ImageNet untargeted sigma=0.25attack success rate (%): 30-65% -> 30-100%Improved
increase distortion budget from 2 to 6 in user study (ablation)GhostCert · User study naturalness% choosing GhostCert: 74 -> 58Hurt
switch ensemble defense to larger noise (sigma 0.5, 1.0) (ablation)GhostCert targeted attack on ensemble · ImageNet targetedattack success rateHurt

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