Evolved Prevention Strategies for 6G Networks Through Stochastic Games and Reinforcement Learning

In this letter, a method to find security strategies for intrusion prevention in a Software Defined Perimeter (SDP) enabled 6G network is demonstrated. These strategies are developed through reinforcement learning by modeling the interaction between an attacker and the network as a stochastic game. This derives an optimal policy which is tested in a multi-agent simulation. Utilizing the optimal strategies, the attacker has a 50% chance of gaining access to the network’s critical information in an SDP-disabled 6G network. Testing the same optimal strategies with an SDP-enabled 6G network, the probability of the attacker gaining access to the same information drops to 1%.