Research map / Adversarial robustness and security / Black-box and transfer attacks
Multi-Paradigm Collaborative Adversarial Attack Against Multi-Modal Large Language Models
Yuanbo Li, Tianyang Xu, Cong Hu, Tao Zhou, Xiao-Jun Wu, Josef Kittler
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 multi-paradigm collaborative optimization (MPCAttack) instead of the best single-paradigm attack | MPCAttack · ImageNet, targeted, open-source MLLMs | ASR (%): 48.60 -> 63.33 | Improved |
| Use MPCAttack instead of FOA-Attack on closed-source models | MPCAttack · ImageNet, untargeted, closed-source MLLMs | ASR (%): 82.63 -> 90.55 | Improved |
| Use MPCAttack instead of FOA-Attack on Flickr30K | MPCAttack · Flickr30K, targeted, open-source MLLMs | ASR (%): 15.41 -> 29.54 | Improved |
| MPCAttack on Qwen2.5-VL-7B untargeted Flickr30K versus X-Transfer | MPCAttack · Flickr30K, untargeted, Qwen2.5-VL-7B | ASR (%): 42.84 -> 32.14 | Hurt |
| Add the cross-modal alignment paradigm (CLIP) (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Improved |
| Add the visual self-supervised paradigm (DINOv2) (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Improved |
| Add the multimodal understanding paradigm (InternVL3) (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Improved |
| Add the MPCO contrastive matching strategy (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Improved |
| Replace CLIP with SigLIP2 as cross-modal surrogate (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Hurt |
| Replace InternVL3-1B with InternVL3-2B (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Improved |
| Replace DINOv2 with DINOv3 (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | No clear effect |
| Use image-only features (lambda = 1) instead of intermediate image-text fusion weight (ablation) | MPCAttack · ImageNet, open-source MLLMs | ASR (%) | Hurt |
Closest papers by meaning
- Mutual-Modality Adversarial Attack with Semantic Perturbation (AAAI 2024 · similarity 0.88)
- Pandora's Box: Towards Building Universal Attackers against Real-World Large Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- HQA-VLAttack: Towards High Quality Adversarial Attack on Vision-Language Pre-Trained Models (NeurIPS 2025 · similarity 0.87)
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (CVPR 2025 · similarity 0.87)
- A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1 (NeurIPS 2025 · similarity 0.87)
- Anyattack: Towards Large-scale Self-supervised Adversarial Attacks on Vision-language Models (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.
