is an automated penetration testing framework that leverages Deep Reinforcement Learning (DRL) to determine and execute optimal attack paths within a logical network. Developed by researchers at the Japan Advanced Institute of Science and Technology (JAIST) , it aims to bridge the gap between AI-driven decision-making and practical cybersecurity auditing. Key Capabilities
One of the most powerful features of is its dual-mode operation, which allows for both safe study and active testing: autopentest-drl
These agents communicate via a shared attention mechanism (a variant of the Transformer architecture), learning emergent strategies like “have the scanner trigger an IDS alert on a decoy while the pivot agent quietly moves through a different subnet.” is an automated penetration testing framework that leverages
, a logic-based security analyzer, to generate an attack graph for comparison. Real Attack Mode Real Attack Mode A Survey for Deep Reinforcement
A Survey for Deep Reinforcement Learning Based Network Intrusion Detection
: The official source code and documentation for the project, maintained by the CROND laboratory at JAIST.