AI Paper Atlas

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

Towards Unified Robustness Against Both Backdoor and Adversarial Attacks

Zhenxing Niu, Yuyao Sun, Qiguang Miao, Rong Jin, Gang Hua

TPAMI 2024 · Adversarial robustness and security › Certified and empirical robustness · approach: Attack evaluation and defense critique · 18 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 mean-teacher (EMA of student models) to PBE (ablation)PBE · CIFAR-10 Blend ablationRACC: 50.64 -> 56.74Improved
add mean-teacher to PBE (ablation)PBE · CIFAR-10 Blend ablationASR: 1.15 -> 0.84Improved
add mean-teacher to PBE (ablation)PBE · CIFAR-10 Blend ablationACC: 90.88 -> 90.36Hurt
add SPECTRE data filtering to PBE (ablation)PBE · CIFAR-10 Blend ablationRACC: 50.64 -> 61.13Improved
add SPECTRE data filtering to PBE (ablation)PBE · CIFAR-10 Blend ablationACC: 90.88 -> 92.36Improved
add backdoor unlearning with poisoned images to PBE (ablation)PBE · CIFAR-10 Blend ablationASR: 1.15 -> 0.48Improved
add backdoor unlearning to PBE (ablation)PBE · CIFAR-10 Blend ablationRACC: 50.64 -> 50.67No clear effect
combine all three (PUD) versus PBEPBE · CIFAR-10 BlendASR: 1.15 -> 0.55Improved
PUD instead of PBEPBE · CIFAR-10 BadNets (all-to-one)ASR: 3.41 -> 1.44Improved
PUD instead of PBEPBE · CIFAR-10 SIG (all-to-one)ASR: 3.53 -> 0.31Improved
combine prediction-based and SPECTRE purification (ablation)prediction-based purification (PBE) · CIFAR-10 BadNets poisoned image filteringfiltered poisoned images: 455/500 -> 499/500Improved
combine prediction-based and SPECTRE purification (ablation)SPECTRE · CIFAR-10 BadNets poisoned image filteringfiltered poisoned images: 488/500 -> 499/500Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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