Research map / Federated learning and privacy / Privacy attacks and secure learning
Exploring the Vulnerabilities of Federated Learning: A Deep Dive Into Gradient Inversion Attacks
Pengxin Guo, Runxi Wang, Shuang Zeng, Jinjing Zhu, Haoning Jiang, Yanran Wang, Yuyin Zhou, Feifei Wang, Hui Xiong, Liangqiong Qu
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 server-side simulation with known local training info (strong simulation) versus none | IG under FedAvg · CIFAR-100, E=2, B=16 | PSNR (dB): 11.62 -> 15.17 | Improved |
| use FedAvg with multiple local epochs instead of single step | IG · CIFAR-100 | reconstruction quality | Hurt |
| add more identical labels in one batch (ablation) | IG · CIFAR-10, batch of 4 | gradient cosine similarity: -0.0305 -> 0.9541 | Hurt |
| increase batch size (ablation) | IG · CIFAR-10/100, ImageNet | attack performance | Hurt |
| increase image resolution (ablation) | IG · ImageNet vs CIFAR | attack performance | Hurt |
| use a more complicated network (ResNet-18 to ResNet-101) (ablation) | IG · CIFAR-100 | attack performance | Hurt |
| attack a well-trained model instead of an untrained one (ablation) | IG · CIFAR-100 | attack performance | Hurt |
| use Sigmoid instead of other activations (ReLU, Tanh) (ablation) | CI-Net · CIFAR/ImageNet | reconstruction | Improved |
| use random Gaussian noise in place of real gradients (ablation) | GGL · ImageNet | semantic reconstruction | No clear effect |
| increase batch size from 1 to 64 | Robbing the Fed (1000 bins) · CIFAR-10 | images exactly reconstructed: 64/64 -> 60/64 | Hurt |
| increase batch size from 1 to 64 | Robbing the Fed (1000 bins) · ImageNet | images exactly reconstructed: 64/64 -> 61/64 | Hurt |
| increase image resolution (ablation) | Fishing · ImageNet | attack performance | Hurt |
Closest papers by meaning
- A New Federated Learning Framework Against Gradient Inversion Attacks (AAAI 2025 · similarity 0.93)
- Gradient Inversion Attacks on Parameter-Efficient Fine-Tuning (CVPR 2025 · similarity 0.89)
- Concealing Sensitive Samples against Gradient Leakage in Federated Learning (AAAI 2024 · similarity 0.89)
- Foreseeing Reconstruction Quality of Gradient Inversion: An Optimization Perspective (AAAI 2024 · similarity 0.89)
- Revisiting Gradient Pruning: A Dual Realization for Defending against Gradient Attacks (AAAI 2024 · similarity 0.89)
- Efficient Federated Learning against Byzantine Attacks and Data Heterogeneity via Aggregating Normalized Gradients (NeurIPS 2025 · 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.
