The Challenge

Conventional systems for preventing attacks rely upon attack signatures, indicators of compromise, or detection of anomalous activity. These methods are challenging to tailor to different network environments and can flood cyber defenders with false alarms.

Cyber expertise is a difficult skillset to acquire. Defenders need to focus on the most important information. This creates a need for defensive systems that are aware of organizational objectives and can collaborate, learn, and reason to manage cyber threats dynamically and autonomously.

Argonne’s Solution

The Descartes program builds autonomous agents for cyber defense. Leveraging decades of cognitive science and the latest trends in artificial intelligence, cybersecurity risk management, and high-performance computing, these agents protect critical systems even if human operators have limited or no ability to interact with them in real-time. 

Environments where Descartes agents have significant potential include electrical distribution grids and generation facilities, civilian or military autonomous vehicles, automated laboratories, and undersea or deep space vehicles.

Benefits

• Dynamically manages cyber threats and makes human defenders more effective
• Protects mission of system it is defending 
• Configured with knowledge of mission objectives
• Monitors cyber environment to learn baseline and objectives
• Different from intrusion detection; proactively balancing cyber defenses with mission assurance
• Uniquely applies cognitive science psychology/AI concepts to cyber defense

History  

Descartes is an extension of an Argonne project for NATO Allied Command Transformation and the AICA International Working Group. Supported by positive feedback from NATO ACT, Argonne’s work was the basis for an experimental campaign at NATO Cyber Coalition 2022 in Tallinn, Estonia.