Large Language Models in Online Open-Source Intelligence

Applications and Defensive Strategies

Abstract

This study examines the emergence of large language models (LLMs) in internet-based open source intelligence (OSINT) and the associated challenges. It presents the source environment and application domains of internet-based OSINT, as well as the role of LLMs in this area. The research focuses on two applications (LLM OSINT PoC and InfoHound) that specifically integrate large language models into frameworks designed for internet-based open source intelligence. The author analyzes and compares the architectural characteristics and prompts of these projects, while also assessing their efficiencies and threats. The study concludes with proposals for defensive strategies against the capabilities of such frameworks, aiming to mitigate the effectiveness of LLM-supported OSINT operations.