Research map / Adversarial robustness and security / Black-box and transfer attacks
Unfooling Perturbation-Based Post Hoc Explainers
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Deploy the scaffolding attack | LIME · COMPAS | fidelity_g: 0.31 -> 0.23 | Hurt |
| Use KNNCAD to filter abnormal perturbations during explanation | LIME · COMPAS, attack with N_hl=1 | fidelity_DEFENSE,f: 0.23 -> 0.30 | Helped |
| Use KNNCAD to filter abnormal perturbations during explanation | LIME · German Credit, attack | fidelity_DEFENSE,f: 0.18 -> 0.27 | Helped |
| Use KNNCAD to filter abnormal perturbations during explanation | LIME · Communities & Crime, attack with N_hl=1 | fidelity_DEFENSE,f: 0.16 -> 0.20 | Helped |
| Deploy the scaffolding attack | SHAP · German Credit | fidelity_g: 0.24 -> 0.12 | Hurt |
| Use KNNCAD to filter abnormal perturbations during explanation | SHAP · German Credit, attack | fidelity_DEFENSE,f: 0.12 -> 0.27 | Helped |
| Use KNNCAD to filter abnormal perturbations during explanation | SHAP · Communities & Crime, attack with N_hl=1 | fidelity_DEFENSE,f: 0.07 -> 0.19 | Helped |
| Use KNNCAD with DETECT to compare anomaly-score distributions for natural and perturbed samples | LIME and SHAP · All evaluated datasets | attack detection | Helped |
| Increase the number of explainer perturbations above 1,000 | DETECT with LIME or SHAP · COMPAS | detection-score variance: n_p > 1,000 | Helped |
| Increase the KNN neighbor count k (ablation) | KNNCAD / DETECT · Communities & Crime | attack-detection margin | Hurt |
| Use SHAP's perturbation scheme instead of LIME's (ablation) | DETECT · Communities & Crime | attack-detection margin: ~5× | Helped |
Closest papers by meaning
- Auditing Local Explanations is Hard (NeurIPS 2024 · similarity 0.84)
- Adversarial Machine Unlearning (ICLR 2025 · similarity 0.83)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.83)
- Factor Graph-based Interpretable Neural Networks (ICLR 2025 · similarity 0.83)
- X-Hacking: The Threat of Misguided AutoML (ICML 2025 · similarity 0.83)
- Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibration (NeurIPS 2025 · similarity 0.83)
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.