Research map / Federated learning and privacy / Privacy attacks and secure learning
Enhanced Privacy Leakage from Noise-Perturbed Gradients via Gradient-Guided Conditional Diffusion Models
Jiayang Meng, Tao Huang, Hong Chen, Chen Hou, Guolong Zheng
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 |
|---|---|---|---|
| use gradient-guided diffusion sampling instead of fine-tuned diffusion | GGSS-R vs FinetunedDiff · CelebA CNN Gaussian noise 0.01 | PSNR: 11.6675 -> 14.9784 | Improved |
| use gradient-guided diffusion sampling | GGSS-R vs FinetunedDiff · CelebA CNN original gradients | PSNR: 29.4490 -> 41.1229 | Improved |
| increase Gaussian noise variance from 1e-4 to 1e-1 (ablation) | GGSS-R · CelebA CNN | PSNR: 28.0343 -> 9.0236 | Hurt |
| use Laplacian instead of Gaussian noise (ablation) | GGSS-R · CelebA CNN | PSNR: 24.0567 -> 24.2073 (var 1e-3) | No clear effect |
| increase gradient batch size from 1 to 8 (ablation) | GGSS-R · CelebA | PSNR: 41.1229 -> 12.1425 | Hurt |
| set guidance rate to 0.20 instead of 0.01 (ablation) | GGSS-R · CelebA | PSNR: 26.9023 -> 41.1229 | Improved |
| raise guidance rate from 0.20 to 0.80 (ablation) | GGSS-R · CelebA | PSNR: 41.1229 -> 35.6101 | Hurt |
| use higher-RV attacked model (ablation) | GGSS-R · CelebA, ResNet18 vs MLP-3 | PSNR: 10.79 -> 43.42 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GIFD — GIFD: A Generative Gradient Inversion Method with Feature Domain Optimization (ICCV 2023) · better on 2 of 2 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
- Concealing Sensitive Samples against Gradient Leakage in Federated Learning (AAAI 2024 · similarity 0.90)
- Mjölnir: Breaking the Shield of Perturbation-Protected Gradients via Adaptive Diffusion (AAAI 2025 · similarity 0.90)
- High-Fidelity Gradient Inversion in Distributed Learning (AAAI 2024 · similarity 0.90)
- Leak and Learn: An Attacker's Cookbook to Train Using Leaked Data from Federated Learning (CVPR 2024 · similarity 0.89)
- DRAG: Data Reconstruction Attack using Guided Diffusion (ICML 2025 · similarity 0.88)
- Foreseeing Reconstruction Quality of Gradient Inversion: An Optimization Perspective (AAAI 2024 · 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.
