Leveraging open-source LLMs for IoT honeypot simulation and enhancing vulnerability assessment and threat intelligence

This paper investigates the integration of open-source large language models (LLMs) for simulating vulnerable Internet of Things (IoT) environments, with the aim of enhancing cybersecurity posture through advanced threat modelling and automated response testing, while also comparing their effectiveness to cloud-based commercial models. We examine the potential of LLM-driven simulations to expose hidden vulnerabilities, evaluate incident response strategies, and support the development of more resilient IoT ecosystems. Particular emphasis is placed on the use of open-source LLMs, which offer distinct advantages in privacy-preserving threat analysis due to their ability to operate entirely on local infrastructure without transmitting sensitive data to external cloud services. As cyber threats targeting IoT devices continue to evolve in complexity and scale, this research highlights the practical applications of LLMs in proactive defence, threat intelligence generation, and vulnerability assessment.