Research map / Vision-language and multimodal / Cross-modal learning
M$^4$I: Multi-modal Models Membership Inference
Pingyi Hu, Zihan Wang, Ruoxi Sun, Hu Wang, Minhui Xue
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 feature-based instead of metric-based attack | M4I · unrestricted scenario image captioning | attack success rate | Improved |
| restrict attacker to data-only knowledge (ablation) | metric-based M4I · image captioning | attack success rate | Hurt |
| attacker lacks architecture and dataset knowledge (constrained) (ablation) | metric-based M4I · image captioning | attack success rate: 81 -> 51% - 76% | Hurt |
| attacker lacks architecture knowledge (ablation) | feature-based M4I · image captioning | attack success rate | Hurt |
| target trained on MSCOCO | metric-based M4I · image captioning | attack success rate | Hurt |
| data augmentation and l2 regularization on target | M4I attacks · image captioning | attack success rate | Mixed |
| differential privacy training with epsilon = 1.3 | M4I attacks · FR target model | attack success rate | Hurt |
| differential privacy training with epsilon = 1.3 | target captioning model · FR target model | average ROUGE-L: 0.245 -> 0.147 | Hurt |
Closest papers by meaning
- Variance-Based Membership Inference Attacks Against Large-Scale Image Captioning Models (CVPR 2025 · similarity 0.89)
- Membership Inference Attacks against Large Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- Membership Inference Attacks against Fine-tuned Large Language Models via Self-prompt Calibration (NeurIPS 2024 · similarity 0.86)
- MMCert: Provable Defense against Adversarial Attacks to Multi-modal Models (CVPR 2024 · similarity 0.85)
- Low-Cost High-Power Membership Inference Attacks (ICML 2024 · similarity 0.85)
- Estimating Model-Level Membership Inference Vulnerability Without Reference Models (NeurIPS 2026 · 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.
