Research map / Adversarial robustness and security / Backdoor and poisoning
IBA: Towards Irreversible Backdoor Attacks in Federated Learning
Dung Thuy Nguyen, Tuan Minh Nguyen, Anh Tuan Tran, Khoa D Doan, KOK SENG WONG
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 partial model poisoning (PGD) to IBA (ablation) | IBA · CIFAR-10, Krum | BA: 27.07% -> 50.33% | Improved |
| add PGD and model replacement (MR) to IBA (ablation) | IBA · CIFAR-10, Krum | BA: 27.07% -> 65.98% | Improved |
| add PGD to IBA (ablation) | IBA · CIFAR-10, Multi-krum | BA: 79.5% -> 97.16% | Improved |
| add PGD and MR to IBA (ablation) | IBA · CIFAR-10, RLR | BA: 63.94% -> 47.07% | Hurt |
| add PGD to IBA (ablation) | IBA · CIFAR-10, RFA | BA: 98.03% -> 96.7% | Hurt |
| add PGD and MR to IBA (ablation) | IBA · CIFAR-10, No-defense | BA: 85.63% -> 97.79% | Improved |
| add PGD to IBA (ablation) | IBA · MNIST, Krum | BA: 14.13% -> 87.48% | Improved |
| add PGD and MR to IBA (ablation) | IBA · MNIST, Krum | BA: 14.13% -> 17.29% | No clear effect |
| add PGD and MR to IBA (ablation) | IBA · MNIST, FoolsGold | BA: 80.08% -> 99.83% | Improved |
| increase compromising ratio of attackers from 5% to 20% (fixed-pool) | IBA · fixed-pool attacks | BA | Improved |
| use IBA with adversarial perturbation trigger instead of DBA patch trigger | IBA vs DBA · Neural Cleanse, CIFAR-10/MNIST/Tiny ImageNet | Anomaly Index | Improved |
| replace DBA with IBA after removing the attacker | IBA vs DBA · MNIST, 250 rounds after removal | BA as % of peak: 38.76 -> 99.14 | Improved |
Closest papers by meaning
- Beyond Traditional Threats: A Persistent Backdoor Attack on Federated Learning (AAAI 2024 · similarity 0.92)
- Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective (AAAI 2024 · similarity 0.92)
- SADBA: Self-Adaptive Distributed Backdoor Attack Against Federated Learning (AAAI 2025 · similarity 0.92)
- Infighting in the Dark: Multi-Label Backdoor Attack in Federated Learning (CVPR 2025 · similarity 0.91)
- Backdoor Federated Learning by Poisoning Backdoor-Critical Layers (ICLR 2024 · similarity 0.90)
- IBAS:Imperceptible Backdoor Attacks in Split Learning with Limited Information (AAAI 2025 · similarity 0.90)
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.
