Research map / Adversarial robustness and security / Backdoor and poisoning
Injection, Attack and Erasure: Revocable Backdoor Attacks via Machine Unlearning
Baogang Song, Dongdong Zhao, Jianwen Xiang, Qiben Xu, Zizhuo Yu
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 |
|---|---|---|---|
| add unlearning loss to trigger optimization (ablation) | revocable backdoor trigger generator · CIFAR-10 (First-Order) | ASR after unlearning: 32.52 -> 25.12 | Improved |
| add fixed sample partition and PCGrad (ablation) | revocable backdoor trigger generator · CIFAR-10 (First-Order) | ASR after unlearning: 68.60 -> 25.12 | Improved |
| add fixed sample partition and PCGrad (ablation) | revocable backdoor trigger generator · CIFAR-10 | ASR: 98.57 -> 98.36 | No clear effect |
| add fixed sample partition and PCGrad (ablation) | gradient conflict · CIFAR-10 | cosine similarity of gradients: -0.65 -> -0.35 | Improved |
| use revocable-trigger optimization vs COMBAT | backdoor attack · CIFAR-10 (First-Order) | ASR after unlearning: 45.97 -> 25.12 | Improved |
| use revocable-trigger optimization vs BadNets | backdoor attack · CIFAR-10 | ASR: 29.8 -> 98.36 | Improved |
| raise poisoning rate from 0.5% to 5% (ablation) | revocable backdoor · CIFAR-10 | ASR: 69.73 -> 98.36 | Improved |
| raise poisoning rate from 0.5% to 5% (ablation) | revocable backdoor · CIFAR-10 | ASR after unlearning: 5.34 -> 25.12 | Hurt |
| raise trigger strength eta from 0.04 to 0.12 (ablation) | revocable backdoor · CIFAR-10 | ASR: 71.82 -> 99.91 | Improved |
| raise trigger strength eta from 0.04 to 0.12 (ablation) | revocable backdoor · CIFAR-10 | ASR after unlearning: 9.71 -> 44.40 | Hurt |
| unlearning with UnrollSGD instead of First-Order | revocable backdoor · CIFAR-10 / ImageNet-10 | benign accuracy after unlearning: 90.05 -> 82.93 | Mixed |
| apply fine-pruning defense | revocable backdoor · CIFAR-10 (First-Order) | ASR | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- COMBAT — COMBAT: Alternated Training for Effective Clean-Label Backdoor Attacks (AAAI 2024) · better on 2 of 4 shared results
- Sleeper Agent — Sleeper Agent: Scalable Hidden Trigger Backdoors for Neural Networks Trained from Scratch (NeurIPS 2022) · better on 4 of 4 shared results
- Wanet — WaNet - Imperceptible Warping-based Backdoor Attack (ICLR 2021) · better on 4 of 4 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
- Backdoor Attacks via Machine Unlearning (AAAI 2024 · similarity 0.90)
- Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness (NeurIPS 2024 · similarity 0.90)
- REFINE: Inversion-Free Backdoor Defense via Model Reprogramming (ICLR 2025 · similarity 0.90)
- Rethinking Backdoor Attacks on Dataset Distillation: A Kernel Method Perspective (ICLR 2024 · similarity 0.89)
- Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense (NeurIPS 2024 · similarity 0.89)
- Demystifying Poisoning Backdoor Attacks from a Statistical Perspective (ICLR 2024 · similarity 0.89)
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.
