VSF-Med: A Vulnerability Scoring Framework for Medical Vision-Language Models
Published in arXiv preprint, 2025
VSF-Med provides a systematic approach to evaluate vulnerabilities in medical VLMs through text-prompt attack templates, imperceptible visual perturbations, and an eight-dimensional risk assessment rubric. The framework evaluated 68,478 attack scenarios across five state-of-the-art models.
Recommended citation: Sadanandan, B., & Behzadan, V. (2025). VSF-Med: A Vulnerability Scoring Framework for Medical Vision-Language Models. arXiv preprint arXiv:2507.00052.
Download Paper