Preliminary Exploration of Pre-Trained Vision-Language Models for Cyber-Threat Modelling in Internet of Medical Things (IoMT)

The field of image processing is undergoing a significant transformation, driven by the advancements in vision-language models (VLMs) based on groundbreaking transformer architectures. With the expansion of Internet of Medical Things (IoMT) devices, the need for robust and efficient threat detection methods has become increasingly critical. The rapid expansion of IoMT networks demands solutions that can automatically identify network-based threats with high precision and low computational cost. This paper presents a novel approach using a pre-trained vision-language model (VLM) that enhances threat detection and develops mitigation strategies for securing IoMT architectures. A custom-designed graphical user interface (GUI) has been created, allowing automated threat detection and the recommendation of mitigation measures. The GUI processes IoMT model images, identifying the components, associated cyber-threats, and propose appropriate mitigation steps. We present results from a real-world case study demonstrating the efficacy of real-time threat detection in reducing the potential impact of cyberattacks on IoMT networks.