paperJuly 2026Unreviewed

Devil in the Lens: Analyzing and Defending Physical Prompt Injection Against Vision-Language Models on Wearable Devices

Yaxin Li, Hao Wang, Yanda Shao, Shuhao Zhang, Yan Long

Abstract

Vision-Language Models (VLMs) are rapidly deployed on human-facing wearable devices such as smart glasses to enable multimodal perception and AI-assisted decision-making. While prior research has demonstrated the risks of visual prompt injection into digital image inputs of VLMs, the unique security challenges posed by the increasing integration between physical environments and wearable intelligence, such as those embodied in VLM-enabled AI glasses, remain underexplored. Toward understanding an

Framework mappings

OWASP Top 10 for LLM Applications

LLM01Prompt Injection

MITRE ATLAS

AML.T0051LLM Prompt Injection

Suggested from the entry's categories.

Cite

@misc{li2026devil,
  title = {{Devil in the Lens: Analyzing and Defending Physical Prompt Injection Against Vision-Language Models on Wearable Devices}},
  author = {Yaxin Li and Hao Wang and Yanda Shao and Shuhao Zhang and Yan Long},
  year = {2026},
  month = jul,
  eprint = {2607.10269},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2607.10269}
}