Article

Image-embedded prompt injection vulnerability of vision-language models in dental radiology: a cross-vendor attack–defense evaluation

2026-07-01

Abstract excerpt

<title>Abstract</title> <p>Image-embedded prompt injection — adversarial text rendered into the pixel data of medical images — is an emerging threat to vision-language models (VLMs) used in clinical decision support. We systematically evaluated this vulnerability across four production-tier VLMs (GPT-4o, Gemini 2.5 Flash, Claude Sonnet 4.5, MedGemma 4B) on 270 dental panoramic radiographs from the DenTeX dataset...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 03601834-cf94-5beb-bec7-98b2de6cf07c
DOI: 10.21203/rs.3.rs-9932271/v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Image-embedded prompt injection vulnerability of vision-language models in dental radiology: a cross-vendor attack–defense evaluationDOI 10.21203/rs.3.rs-9932271/v1Read

OCR-Mediated Modality Dominance in Vision-Language Models: Implications for Radiology AI TrustworthinessSemantic proximity Read

Prompt injection attacks on vision-language models for surgical decision supportSemantic proximity Read

A Clinically-Informed Framework for Evaluating Vision-Language Models in Radiology Report Generation: Taxonomy of Errors and Risk-Aware MetricSemantic proximity Read

Drug or Pokémon? Large language model performance in identification of fabricated medicationsSemantic proximity Read

Vision and Language Models for Classifying Maxillary Sinus Disease on Cone-Beam Computed Tomography: A Transparent Multimodal BenchmarkSemantic proximity Read

Invisible Text Injection: The Trojan Horse of AI-Assisted Medical Peer ReviewSemantic proximity Read

Benchmarking Multimodal Large Language Models for Forensic Science and Medicine: A Comprehensive Dataset and Evaluation FrameworkSemantic proximity Read

Red-Teaming Medical AI: Systematic Adversarial Evaluation of LLM Safety Guardrails in Clinical ContextsSemantic proximity Read

A comparative study of deepfake creation and detection in Lung Adenocarcinoma ImagesSemantic proximity Read

Safety Evaluation of Google’s Gemini Nano Banana Image Model under Adversarial and Realistic Prompt ConditionsSemantic proximity Read

How AI is used in FDA-authorized medical devices: a taxonomy across 1,016 authorizationsSemantic proximity Read

Natural Images Allow Universal Adversarial Attacks on Medical Image Classification Using Deep Neural Networks With Transfer LearningSemantic proximity Read