Abstract

In the context of building enterprise security posture, leveraging Advanced Persistent Threat (APT) intelligence operations and Breach & Attack Simulation (BAS) exercises is critical for proactive cybersecurity defense. However, these efforts face bottlenecks such as inefficient manual processing, fragmented intelligence, and high costs in script development.

In this talk, we will introduce an integrated technical framework that leverages large language models, knowledge graphs, and AI agents to address these challenges. First, we will explain how we automate the construction of an attack knowledge base using APT intelligence. Next, we will demonstrate how AI agents convert APT intelligence into BAS exercise scripts, enabling realistic drills in internal environments—highlighting several innovations. Finally, we will walk through the workflow in a demonstration. We look forward to sharing this with you.