Principal Security Architect Motorola Solutions, Illinois
Intelligence documents such as threat advisories and reports contain a wealth of information regarding emerging trends, attacker techniques and exploits against public safety infrastructure. However, synthesizing available information into prioritized, actionable bits is challenging. During this talk, we explore how large language models like GPT can be leveraged to generate vector embeddings from unstructured threat documents. This talk provides relevant demonstrations to show how threat embeddings thus created capture semantics in threats, vulnerabilities and attacker techniques. Real world scenarios will showcase how threat embeddings augment security operations by prioritizing the highest risks. Finally, we discuss relevant advantages and limitations.
Learning Objectives:
Comprehend how AI learns threat intelligence patterns and uses them to answer questions. In other words, how LLMs generate vector embeddings from unstructured intel documents and capture threat semantics.
Learn by watching a demo on how to create a AI threat intelligence agent that answers public safety specific questions.
Understand the advantages and limitations of current LLMs in aiding intelligence-driven security operations.
