Research map / Adversarial robustness and security / Attacks on text and graphs
Knowledge-Enriched Distributional Model Inversion Attacks
Si Chen, Mostafa Kahla, Ruoxi Jia, Guo-Jun Qi
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 soft-label discrimination (SD) (ablation) | GMI · CelebA VGG16 | attack accuracy: .21 -> .35 | Improved |
| add soft-label discrimination (SD) (ablation) | GMI · CelebA VGG16 | FID: 52.51 -> 33.75 | Improved |
| add entropy minimization (EM) to SD (ablation) | SD · CelebA VGG16 | attack accuracy: .35 -> .43 | Improved |
| add distributional recovery (DR) (ablation) | GMI · CelebA VGG16 | attack accuracy: .21 -> .47 | Improved |
| add distributional recovery (DR) (ablation) | GMI · CelebA VGG16 | FID: 52.51 -> 45.28 | Improved |
| add DR to SD (ablation) | SD · CelebA VGG16 | attack accuracy: .35 -> .62 | Improved |
| add EM to SD+DR (ablation) | SD+DR · CelebA VGG16 | attack accuracy: .62 -> .72 | Improved |
| add EM to SD+DR (ablation) | SD+DR · CelebA VGG16 | FID: 23.82 -> 23.72 | No clear effect |
| replace GMI with proposed attack | face.evolve · CelebA | attack accuracy: .31 -> .81 | Improved |
| replace GMI with proposed attack | MNIST digit classifier · MNIST | attack accuracy: .08 -> .68 | Improved |
| train GAN on a shifted public dataset (FaceScrub instead of in-distribution) | Ours · FaceScrub->CelebA | attack accuracy: .72 -> .13 | Hurt |
| train GAN on FFHQ instead of CelebA-disjoint public data | Ours · FFHQ->CelebA | attack accuracy: .72 -> .36 | Hurt |
Closest papers by meaning
- Privacy Leakage on DNNs: A Survey of Model Inversion Attacks and Defenses (IJCV 2026 · similarity 0.92)
- A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks (AAAI 2025 · similarity 0.90)
- Pseudo-Private Data Guided Model Inversion Attacks (NeurIPS 2024 · similarity 0.89)
- Model Inversion Robustness: Can Transfer Learning Help? (CVPR 2024 · similarity 0.88)
- Trap-MID: Trapdoor-based Defense against Model Inversion Attacks (NeurIPS 2024 · similarity 0.88)
- From Head to Tail: Efficient Black-box Model Inversion Attack via Long-tailed Learning (CVPR 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.
