Research map / Federated learning and privacy / Privacy attacks and secure learning
zPROBE: Zero Peek Robustness Checks for Federated Learning
Zahra Ghodsi, Mojan Javaheripi, Nojan Sheybani, Xinqiao Zhang, Ke Huang, Farinaz Koushanfar
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 zPROBE robustness checks under Byzantine attacks | federated learning on MNIST · MNIST IID | test accuracy: 10.2%-11.2% -> within 0.0%-0.6% of benign | Improved |
| add zPROBE robustness checks under Byzantine attacks | federated learning on CIFAR-10 · CIFAR-10 IID | accuracy gap to benign: 45%-90% -> 3%-7% | Improved |
| use rank-based dynamic thresholds in zPROBE instead of static thresholds with public data (EIFFeL) | robust and private FL · MNIST, FMNIST, CIFAR-10 | accuracy: +1.2%, +0.5%, +2.8% | Improved |
| use per-user threshold filtering instead of the median of cluster means as final aggregator (ablation) | zPROBE aggregation · F-MNIST | accuracy: 28.6% degradation avoided | Improved |
| decrease portion of malicious updates S_m (ablation) | zPROBE runtime · F-MNIST LeNet5 | runtime (ms): 316.5 (S_m=1.0) -> 777.9 (S_m=0.1) | Hurt |
| increase number of clusters (ablation) | zPROBE · F-MNIST | accuracy: less than 0.18% variation | No clear effect |
| increase number of clients from 30 to 200 (ablation) | zPROBE · F-MNIST | accuracy | No clear effect |
| add random user dropouts (ablation) | zPROBE · F-MNIST | accuracy: fluctuations <0.13% | No clear effect |
| increase attack magnitude kappa (ablation) | zPROBE defense · F-MNIST sign flip | accuracy loss: accuracy loss 0.2%-9.8% vs ~88% without defense | Improved |
| add zPROBE robustness and correctness checks on top of baseline secure aggregation | client runtime · MNIST | runtime (ms): 208.0 -> 444.7 | Hurt |
Later papers that compared with this method
1 later accepted paper reports zPROBE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Fortifying Federated Learning Towards Trustworthiness via Auditable Data Valuation and Verifiable Client Contribution (CVPR 2025 · better on 1 of 1 shared result)
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
- Towards the Robustness of Differentially Private Federated Learning (AAAI 2024 · similarity 0.87)
- Towards Trustworthy Federated Learning with Untrusted Participants (ICML 2025 · similarity 0.86)
- Certifiably Byzantine-Robust Federated Conformal Prediction (ICML 2024 · similarity 0.86)
- Certifiably Robust Model Evaluation in Federated Learning under Meta-Distributional Shifts (ICML 2025 · similarity 0.86)
- One-shot Empirical Privacy Estimation for Federated Learning (ICLR 2024 · similarity 0.86)
- Self-Driven Entropy Aggregation for Byzantine-Robust Heterogeneous Federated Learning (ICML 2024 · 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.
