Aurora is a novel system that automates cyberattack emulation by combining large language models with classical planning to generate realistic attack scenarios for security testing.

• Creates comprehensive, up-to-date cyberattack datasets to support threat-informed defense
• Automatically generates attack plans, sets up emulation environments, and executes attacks
• Addresses the critical shortage of modern cyberattack benchmarks in security research
• Enables more effective evaluation and benchmarking of defense systems

This research significantly advances security testing capabilities by providing automated tools to simulate sophisticated APT campaigns, helping organizations prepare for and defend against evolving cyber threats.

From Sands to Mansions: Towards Automated Cyberattack Emulation with Classical Planning and Large Language Models