Open-Source Intelligence Decision-Making and Analysis Method Based on Large Language Models (Llms)

Open-source intelligence (OSINT) decision making and analysis constitute one of the core challenges in intelligence analysis. To address the need for comprehensive and rapid OSINT collection and the engineering implementation of its analysis, we propose an OSINT acquisition method based on the self-reflective capabilities of Large Language Models (LLMs). Initially, the LLM performs background information retrieval and analysis on specific events, extracting relevant contextual data. Subsequently, it automatically generates search terms related to the involved parties based on the event background extracted, facilitating the retrieval and analysis of these parties. The retrieved information about the related parties is then summarized, and the LLM evaluates and suggests modifications to enhance the summaries. This feedback loop iterates, allowing the LLM to refine the related party content until it meets the predefined standards. Upon confirmation by the LLM, the analysis is finalized, and the results are returned. Validation of the reported results demonstrates that this method effectively enhances the completeness of information retrieval and summarization while filtering erroneous content, thus achieving efficient OSINT acquisition and analysis.