Research map / Adversarial robustness and security / Certified and empirical robustness
Exploiting Missing Data Remediation Strategies Using Adversarial Missingness Attacks
Deniz Koyuncu, Alex Gittens, Bülent Yener, Moti Yung
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use BLAMM adversarial missingness instead of MCAR with same missingness rate (ablation) | mean-imputation modeler · ca-housing, wine-quality, german-credit | normalized L1 error to adversarial coefficients: ca-housing 0.76 -> 0.01 | Improved |
| attack with mean-imputation proxy against a CCA modeler (mismatch) (ablation) | BLAMM mean attack · ca-housing, wine-quality | normalized L1 error: cca/mean ca-housing 1.12, wine 0.71 | Hurt |
| use CCA attack against mean-imputation modeler (ablation) | BLAMM CCA attack · ca-housing, wine-quality, german-credit | normalized L1 error: mean/cca 0.01, 0.04, 0.01 | Improved |
| modeler switches to MICE imputation (ablation) | modeler · CCA attack on ca-housing, wine-quality | normalized L1 error: cca/cca 0.06 -> mice/cca 0.27 (ca-housing) | Improved |
| reduce adversary data access from 100% to 25% (ablation) | BLAMM on TARnet + mean · Twins ATE | estimated ATE (%): 10.52 -> 1.62 | Hurt |
| reduce adversary data access from 100% to 25% (ablation) | BLAMM on Tnet + mean · Twins ATE | estimated ATE (%): 10.42 -> -1.5 | Hurt |
| reduce adversary data access from 100% to 25% (ablation) | BLAMM on linear and CF estimators · Twins ATE | estimated ATE (%): linear 9.86 -> 2.26; CF 7.65 -> 1.54 | Hurt |
| use BLAMM attack versus MCAR on Twins ATE with full access | regression ATE estimators · Twins ATE | estimated ATE (%): TARnet -1.0 -> 10.52 | Improved |
Closest papers by meaning
- From Trojan Horses to Castle Walls: Unveiling Bilateral Data Poisoning Effects in Diffusion Models (NeurIPS 2024 · similarity 0.86)
- Contamination-Resilient Anomaly Detection via Adversarial Learning on Partially-Observed Normal and Anomalous Data (ICML 2024 · similarity 0.85)
- Missing Data Imputation by Reducing Mutual Information with Rectified Flows (NeurIPS 2025 · similarity 0.84)
- Reconstruction Attacks on Machine Unlearning: Simple Models are Vulnerable (NeurIPS 2024 · similarity 0.84)
- Machine Unlearning Fails to Remove Data Poisoning Attacks (ICLR 2025 · similarity 0.83)
- Adversarial Attacks on Data Attribution (ICLR 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.
