Research map / Adversarial robustness and security / Backdoor and poisoning
Backdoor Defense via Deconfounded Representation Learning
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 |
|---|---|---|---|
| Train using CBD instead of DBD | CIFAR-10, six-attack average · CIFAR-10 | ASR (%): 4.53 -> 1.60 | Helped |
| Train using CBD instead of DBD | GTSRB, six-attack average · GTSRB | ASR (%): 3.42 -> 1.82 | Helped |
| Train using CBD instead of DBD | ImageNet Subset, four-attack average · ImageNet Subset | ASR (%): 2.36 -> 0.91 | Helped |
| Use CBD on poisoned data instead of no defense | CIFAR-10, six-attack average · CIFAR-10 | CA (%): 85.14 -> 87.00 | Helped |
| Use CBD on poisoned data instead of no defense | GTSRB, six-attack average · GTSRB | CA (%): 96.33 -> 95.17 | Hurt |
| Use CBD on poisoned data instead of no defense | ImageNet Subset, four-attack average · ImageNet Subset | CA (%): 85.74 -> 87.90 | Helped |
| Use CBD rather than no defense at 50% poisoning | CIFAR-10, BadNets · CIFAR-10, BadNets, 50% poisoning | ASR (%): 100 -> 1.47 | Helped |
| Use CBD rather than no defense at 50% poisoning | CIFAR-10, Blend · CIFAR-10, Blend, 50% poisoning | CA (%): 69.67 -> 85.56 | Helped |
| Apply CBD against the adaptive noise attack | CIFAR-10, BadNets · CIFAR-10, BadNets adaptive attack | ASR (%): 99.62 -> 4.31 | Helped |
| Apply CBD to the stealthier WaNet attack | CIFAR-10 · CIFAR-10, WaNet | ASR (%): 98.55 -> 4.24 | Mixed |
Closest papers by meaning
- Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness (NeurIPS 2024 · similarity 0.89)
- BaDExpert: Extracting Backdoor Functionality for Accurate Backdoor Input Detection (ICLR 2024 · similarity 0.89)
- Backdoor Defense via Test-Time Detecting and Repairing (CVPR 2024 · similarity 0.89)
- REFINE: Inversion-Free Backdoor Defense via Model Reprogramming (ICLR 2025 · similarity 0.89)
- Adversarial-Inspired Backdoor Defense via Bridging Backdoor and Adversarial Attacks (AAAI 2025 · similarity 0.88)
- BAN: Detecting Backdoors Activated by Adversarial Neuron Noise (NeurIPS 2024 · similarity 0.88)
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.