Deep Reinforcement Learning
AI defenders for critical government cyber systems lack a deep learning system that can continuously improve defense capabilities. This problem is heightened by a lack of advanced AI attackers that can test a diverse range of cyber defenses.
Knexus
Solution
Train a robust AI attacker and AI defender to continuously improve cyber defenses.
Created a multi-agent system to pit both “Red” AI attackers and “Blue” AI defenders against one another to continually improve both AI cyber capabilities.
Utilize a decision transformer as an innovative solution due to its ability to learn with sparce rewards and large action spaces enabling an unprecedented amount and type of attack scenarios.
The AI attacking model presents unique attacking vectors not traditionally seen by cyber defenses.
After seeing these unique attack vectors, the robustness of the AI defense model significantly enhances cyber defenses.
Customer Outcome
Successfully trained a “Red” AI attacker which consistently won games against a weak defender after training the AI attacker against a strong defender.
The performance of the AI attacker showed an average increase of 25% compared to the in-domain attacking model.
This proved that a multi-agent approach will train both more effective attackers and defenders.
