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

Abstract

Several zero trust security frameworks were introduced to protect the modern network infrastructures. Among them, the SDP which offers a security perimeter for a network by dividing its control and data plane. Under the SDP framework, a controller and accepting host module must verify and authenticate a legitimate initiating host/clients before gaining access to the application infrastructure. This eventually blackens the entire infrastructure making it undetectable to unauthorized users. Although this framework is proven to be very immune against attacks, there is no clear strategy to prevent the intrusion. Precisely, the framework is relying on detect rather than prevent. Herein, a method to find security strategies for intrusion prevention in SDP network is presented based on modeling the interaction between an attacker and an SDP Controller as a stochastic game. These security strategies evolve through reinforcement learning and can emerge from self-play.