Leveraging Large Language Models for Cybersecurity Risk Assessment of Autonomous Forestry Machines

Abstract

Large language models are a type of deep-learning model trained on massive data to generate responses to user prompts similar to natural language. Large language models can be specialized into several different domains such as medical, and e-commerce. This thesis investigated how cybersecurity experts can benefit from large language models in the risk assessment process in the forestry domain. This research was carried out in collaboration with the Research Institutes of Sweden and Chalmers University of Technology. This thesis is a part of the EU-funded AGRARSENSE project. We conducted a design science study including 15 interviews, 12 demos, and a survey. We used local Llama 2 7B and developed an RAG application by supplying the model with data relevant to cybersecurity and the AGRARSENSE project. We created a generated risk assessment document using the tool. The tool is the main artifact of this study. Also, we analyzed several articles. The study demonstrated that large language models can be used in multiple ways in a risk assessment process such as an evaluation tool, assisting chatbot, or generating risk assessments. The findings showed that trust remains an issue for large language models. Even though cybersecurity is one of a software system’s most critical work areas, experts are willing to use such LLMs. Experts are willing to use evaluation and assisting features more than the generation feature.