Research map / Adversarial robustness and security / Attacks on text and graphs
Variational Model Inversion Attacks
Kuan-Chieh Wang, YAN FU, Ke Li, Ashish J Khisti, Richard Zemel, Alireza Makhzani
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 |
|---|---|---|---|
| Replace General MI with VMI using DCGAN | Model inversion attack · MNIST | accuracy: 0 -> 0.95{ $\pm$ 0.02 } | Improved |
| Replace Generative MI with VMI using DCGAN | Model inversion attack · CelebA | accuracy: 0.07{ $\pm$ 0.02 } -> 0.37{ $\pm$ 0.07 } | Improved |
| Use StyleGAN instead of DCGAN | VMI · CelebA | accuracy: 0.37{ $\pm$ 0.07 } -> 0.55{ $\pm$ 0.06 } | Improved |
| Use StyleGAN instead of DCGAN | VMI · CXR | FID: 265.14 { \ (96.3)} -> 123.17 { \ (57.0)} | Improved |
| Increase gamma from 1e-3 to 1e-1 (ablation) | VMI with StyleGAN and Gaussian variational family · CelebA | diversity: 0.61{ $\pm$ 0.03 } -> 0.70}{ $\pm$ 0.03 } | Improved |
| Increase gamma from 1e-3 to 1e-1 (ablation) | VMI with StyleGAN and Gaussian variational family · CelebA | accuracy: 0.56}{ $\pm$ 0.05 } -> 0.23{ $\pm$ 0.03 } | Hurt |
| Use a flow variational family instead of a Gaussian (ablation) | VMI with StyleGAN, gamma=0 · CelebA | accuracy: 0.57}{ $\pm$ 0.06 } -> 0.58}{ $\pm$ 0.06 } | No clear effect |
| Use a flow variational family instead of a Gaussian (ablation) | VMI with StyleGAN, gamma=1e-1 · CelebA | diversity: 0.70}{ $\pm$ 0.03 } -> 0.59}{ $\pm$ 0.05 } | Hurt |
| Use a flow variational family instead of a Gaussian (ablation) | VMI with DCGAN · MNIST | accuracy: 0.93{ $\pm$ 0.06 } -> 0.95{ $\pm$ 0.02 } | Improved |
| Use StyleGAN instead of DCGAN | VMI · CelebA | FID: 40.89 { \ (31.3)} -> 17.41 { \ (19.2)} | Improved |
Closest papers by meaning
- A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks (AAAI 2025 · similarity 0.88)
- Model Inversion Robustness: Can Transfer Learning Help? (CVPR 2024 · similarity 0.87)
- Pseudo-Private Data Guided Model Inversion Attacks (NeurIPS 2024 · similarity 0.86)
- From Head to Tail: Efficient Black-box Model Inversion Attack via Long-tailed Learning (CVPR 2025 · similarity 0.86)
- High-Fidelity Gradient Inversion in Distributed Learning (AAAI 2024 · similarity 0.86)
- Generative Model Inversion Through the Lens of the Manifold Hypothesis (NeurIPS 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.
