When Background Matters: Breaking Medical Vision Language Models by Transferable Attack

arXiv:2604.17318v1 Announce Type: new Abstract: Vision-Language Models (VLMs) are increasingly used in clinical diagnostics, yet their robustness to adversarial attacks remains largely unexplored, posing serious risks. Existing medical attacks focus on secondary objectives such as model stealing or adversarial fine-tuning, while transferable attacks from natural images introduce visible distortions that clinicians can easily detect. To address this, we propose MedFocusLeak, a highly transferable black-box multimodal attack that induces incorrect yet clinically plausible diagnoses while keeping perturbations imperceptible. The method injects coordinated perturbations into non-diagnostic background regions and employs an attention distraction mechanism to shift the model's focus away from pathological areas. Extensive evaluations across six medical imaging modalities show that MedFocusLeak achieves state-of-the-art performance, generating misleading yet realistic diagnostic outputs across diverse VLMs. We further introduce a unified evaluation framework with novel metrics that jointly capture attack success and image fidelity, revealing a critical weakness in the reasoning capabilities of modern clinical VLMs.

Related

• Adversarial Evasion Attacks on Computer Vision using SHAP Values
• BioVLM: Routing Prompts, Not Parameters, for Cross-Modality Generalization in Biomedical VLMs
• Democratising Pathology Co-Pilots: An Open Pipeline and Dataset for Whole-Slide Vision-Language Modelling
• Phantasia: Context-Adaptive Backdoors in Vision Language Models
• Beyond the Failures: Rethinking Foundation Models in Pathology
• HaloProbe: Bayesian Detection and Mitigation of Object Hallucinations in Vision-Language Models

Source: arXiv cs.CV | 2026-04-21