GM-TI: GAI Model for Threat Intelligence

This paper focuses on the development of a domain-specific Generative AI (GAI) model specialized in threat intelligence, while providing the literature research and experiments using different available models. Our research and evaluation result led to the usage of Mistral 7B as the baseline model and incorporating Retrieval-Augmented Generation (RAG). Datasets from frameworks like MITRE ATT&CK, CVE repositories, and open threat intelligence platforms are integrated, allowing the model to address the challenges of this rapidly evolving threat landscape. This solution helps to speed up and simplify the process of threat detection while helping analysts to identify their corresponding mitigation processes conveniently and quickly. The result shows the capability of the model in cybersecurity applications, with significant improvements in accuracy and contextual relevance compared to traditional models.