Research map / Adversarial robustness and security / Attacks on text and graphs
PureProof: Diffusion-Resistant Black-box Targeted Attack on Large Vision-Language Models
Yiming Cao, Dong Wang, Xinqi Lyu, Bin Xiao
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 full PureProof (SRA+ARA+SCR) vs diffusion-unaware VLM attacks | targeted VLM attacks · DiffPure defense, LLaVA-1.5 | ASR (Fool) %: 41.4 -> 76.8 (vs DH-cos) | Improved |
| use single-step reverse alignment instead of full-chain backpropagation through diffusion | PureProof · time per step | seconds per step: 44.717 -> 1.761 | Improved |
| add Adaptive Re-noising Augmentation (ARA) (ablation) | PureProof · Qwen3-VL and LLaVA-1.5 vs DiffPure | attack efficacy | Improved |
| add SRA and SCR components (ablation) | PureProof · Qwen3-VL and LLaVA-1.5 vs DiffPure | attack efficacy | Improved |
| increase re-noised variants K from 0 to 1 (ablation) | PureProof ARA · Qwen3-VL and LLaVA-1.5 vs DiffPure | attack efficacy | Improved |
| increase K beyond 2 (ablation) | PureProof ARA · Qwen3-VL and LLaVA-1.5 vs DiffPure | attack efficacy | No clear effect |
| increase Gaussian noise from 8/255 to 16/255 | PureProof · MiniGPT-4 Gaussian noise | Ensemble CLIP score: 0.6342 -> 0.6315 | No clear effect |
| increase Gaussian noise from 8/255 to 16/255 | CoA · MiniGPT-4 Gaussian noise | Ensemble CLIP score: 0.6095 -> 0.4969 | Hurt |
| combine PureProof with CoA objective | PureProof · no defense, GPT-5 | Ensemble CLIP score: 0.6245 -> 0.6551 | Improved |
| use full-chain backpropagation (DA-cos/DH-cos) instead of PureProof | diffusion-aware attacks · DiffPure, Qwen3-VL | ASR (Fool) %: 38.3 -> 77.5 in favor of PureProof | Hurt |
Closest papers by meaning
- Towards Understanding the Robustness of Diffusion-Based Purification: A Stochastic Perspective (ICLR 2025 · similarity 0.90)
- Attention! Your Vision Language Model Could Be Maliciously Manipulated (NeurIPS 2025 · similarity 0.88)
- DiffBreak: Is Diffusion-Based Purification Robust? (NeurIPS 2025 · similarity 0.88)
- Pandora's Box: Towards Building Universal Attackers against Real-World Large Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- Exploring Visual Vulnerabilities via Multi-Loss Adversarial Search for Jailbreaking Vision-Language Models (CVPR 2025 · similarity 0.87)
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (CVPR 2025 · similarity 0.87)
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.
