Research map / Federated learning and privacy / Privacy attacks and secure learning
Neurotoxin: Durable Backdoors in Federated Learning
Zhengming Zhang, Ashwinee Panda, Linyue Song, Yaoqing Yang, Michael Mahoney, Prateek Mittal, Ramchandran Kannan, Joseph Gonzalez
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 |
|---|---|---|---|
| Mask poisoned gradients away from coordinates with large benign updates | Baseline backdoor attack on Reddit with LSTM · Reddit, trigger len = 1 | lifespan: 32 -> 122 | Improved |
| Mask poisoned gradients away from coordinates with large benign updates | Baseline backdoor attack · Reddit, trigger 2, differential privacy | lifespan: 13 -> 41 | Improved |
| Mask poisoned gradients away from coordinates with large benign updates | Baseline backdoor attack · CIFAR10, base-case trigger | lifespan: 116 -> 279 | Improved |
| Mask poisoned gradients away from coordinates with large benign updates | Baseline backdoor attack · CIFAR100, base-case trigger | lifespan: 943 -> 1723 | Improved |
| Use mask ratio 5% rather than the baseline ratio 0% (ablation) | Neurotoxin · Reddit, trigger set 1 | lifespan: 44 -> 197 | Improved |
| Use mask ratio 3% rather than the baseline ratio 0% (ablation) | Neurotoxin · Reddit, trigger set 2 | lifespan: 78 -> 187 | Improved |
| Use mask ratio 1% rather than the baseline ratio 0% (ablation) | Neurotoxin · Reddit, trigger set 3 | lifespan: 124 -> 302 | Improved |
| Reduce attacker participation to every other active round (ablation) | Baseline and Neurotoxin backdoor attacks · Reddit, trigger 2 | lifespan: 17 to 11; 70 to 51 | Mixed |
| Increase trigger difficulty by shortening it from three words to one (ablation) | Baseline and Neurotoxin backdoor attacks · Reddit, LSTM | lifespan: 78 -> 32; 123 -> 122 | Mixed |
| Use a larger AttackNum (ablation) | Neurotoxin · Reddit, trigger set 2 | lifespan: 70 -> 105 -> 123 -> 173 | Improved |
| Apply Neurotoxin gradient masking | Baseline backdoor attack · Sentiment140 | lifespan: 278 -> 416 | Improved |
| Apply Neurotoxin gradient masking | Baseline backdoor attack · CIFAR10 | top eigenvalue: 899.37 -> 210.14 | Improved |
Later papers that compared with this method
3 later accepted papers report Neurotoxin as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Batman: Benign Knowledge Alignment Through Malicious Null Space in Federated Backdoor Attack (CVPR 2026 · better on 4 of 4 shared results)
- Bad-PFL: Exploiting Backdoor Attacks against Personalized Federated Learning (ICLR 2025 · better on 1 of 1 shared result)
- Chameleon: Adapting to Peer Images for Planting Durable Backdoors in Federated Learning (ICML 2023 · 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 Federated Learning by Poisoning Backdoor-Critical Layers (ICLR 2024 · similarity 0.88)
- Beyond Traditional Threats: A Persistent Backdoor Attack on Federated Learning (AAAI 2024 · similarity 0.88)
- Demystifying Poisoning Backdoor Attacks from a Statistical Perspective (ICLR 2024 · similarity 0.87)
- Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective (AAAI 2024 · similarity 0.87)
- BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models (NeurIPS 2025 · similarity 0.87)
- The Ripple Effect: On Unforeseen Complications of Backdoor Attacks (ICML 2025 · similarity 0.86)
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.
