Research map / Adversarial robustness and security / Black-box and transfer attacks
Reinforcement Learning-Based Black-Box Model Inversion Attacks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 RLB-MI instead of the strongest listed baseline | Black-box model inversion · CelebA with VGG16, ResNet-152, and Face.evoLVe | Attack Acc: 0.703 -> 0.659; 0.765 -> 0.804; 0.741 -> 0.793 | Mixed |
| Use RLB-MI instead of the strongest listed baseline | Black-box model inversion · PubFig83 and FaceScrub | Attack Acc: 0.400 -> 0.420; 0.400 -> 0.475 | Helped |
| Use RLB-MI instead of the strongest listed baseline | Black-box model inversion · CelebA with Face.evoLVe | Attack Acc: 0.741 -> 0.793 | Helped |
| Use RLB-MI instead of the strongest listed baseline | Black-box model inversion with FFHQ public data · FFHQ→PubFig83 and FFHQ→FaceScrub | Attack Acc: 0.360 -> 0.400; 0.285 -> 0.385 | Mixed |
| Use RLB-MI instead of the strongest listed baseline | Black-box model inversion with FFHQ public data · FFHQ→CelebA | Attack Acc: 0.427 -> 0.433 | Helped |
| Increase diversity factor α (ablation) | RLB-MI · CelebA with Face.evoLVe | Attack accuracy | Hurt |
| Increase diversity factor α (ablation) | RLB-MI · CelebA with Face.evoLVe | Coverage | Helped |
| Increase diversity factor α (ablation) | RLB-MI · CelebA with Face.evoLVe | Density | No clear effect |
| Train for more episodes past the saturation point (ablation) | RLB-MI · PubFig83 and FaceScrub | Attack accuracy | Hurt |
| Use DDPG instead of SAC (ablation) | RLB-MI · CelebA with Face.evoLVe | Attack Acc: 0.793 -> 0.380 | Hurt |
| Use TD3 instead of SAC (ablation) | RLB-MI · CelebA with Face.evoLVe | Attack Acc: 0.793 -> 0.677 | Hurt |
Closest papers by meaning
- Pseudo-Private Data Guided Model Inversion Attacks (NeurIPS 2024 · similarity 0.89)
- A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks (AAAI 2025 · similarity 0.88)
- From Head to Tail: Efficient Black-box Model Inversion Attack via Long-tailed Learning (CVPR 2025 · similarity 0.88)
- High-Fidelity Gradient Inversion in Distributed Learning (AAAI 2024 · similarity 0.86)
- Trap-MID: Trapdoor-based Defense against Model Inversion Attacks (NeurIPS 2024 · similarity 0.86)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · 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.