Research map / Federated learning and privacy / Privacy attacks and secure learning
Understanding Deep Gradient Leakage via Inversion Influence Functions
Haobo Zhang, Junyuan Hong, Yuyang Deng, Mehrdad Mahdavi, Jiayu Zhou
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 |
|---|---|---|---|
| Estimate recovery risk with I²F power iteration instead of directly minimizing inversion loss | CIFAR10, ResNet18 · CIFAR10 | computation time: 5 seconds at most; 150 seconds | Improved |
| Perturb gradients along Jacobian directions with smaller eigenvalues | DGL defenses · MNIST and CIFAR10 | recovery error | Improved |
| Use homogeneous Gaussian gradient noise | DGL defenses · MNIST and CIFAR10 | recovery error across samples and classes | Mixed |
| Initialize with Kaiming or Xavier instead of uniform | LeNet · MNIST and CIFAR10 | recovery error under Gaussian gradient noise | Improved |
| Initialize with normal rather than Kaiming or Xavier | LeNet · MNIST and CIFAR10 | recovery error under Gaussian gradient noise | Hurt |
| Use the I²F lower bound rather than the numerator-only approximation (ablation) | I²F · Different models and datasets | approximation of recovery error | Improved |
| Use I²F to estimate recovery error across model and dataset settings | DGL and GS attacks · MNIST and CIFAR10 | correlation with recovery error | Mixed |
| Apply I²F to text-gradient inversion | BERT and GPT-2 with TAG · WikiText-103 | correlation with text similarity metrics | Improved |
Closest papers by meaning
- High-Fidelity Gradient Inversion in Distributed Learning (AAAI 2024 · similarity 0.87)
- A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks (AAAI 2025 · similarity 0.86)
- Mjölnir: Breaking the Shield of Perturbation-Protected Gradients via Adaptive Diffusion (AAAI 2025 · similarity 0.86)
- Differentially private optimization for non-decomposable objective functions (ICLR 2025 · similarity 0.85)
- Concealing Sensitive Samples against Gradient Leakage in Federated Learning (AAAI 2024 · similarity 0.85)
- Theoretical Insights in Model Inversion Robustness and Conditional Entropy Maximization for Collaborative Inference Systems (CVPR 2025 · similarity 0.85)
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.
