Secure Migration in NGN: An Optimal Stopping Problem Approach with Partial Observability

To ensure uninterrupted access to critical VM applications, a robust VM migration plan is necessary, especially during natural disasters or various cyber incidents such as state-led actions, espionage, and attacks. Existing literature often neglects VM migration security. To address this, we formulate a model that treats secure VM migration as an optimal stopping problem with partial observability in the context of Next Generation Core Networks (NGCN). We envision a scenario wherein a malevolent actor launches attacks during migration, and a defender oversees the process to ensure security. The migration is guaranteed to occur only via a secure path fortified with security checkpoints. We postulate that the VMs host diverse 5G Core (5GC) entities, a critical component of NGCN. This scenario is modeled via a Partial Observable Markov Decision Process (POMDP), with the Generalized Proximity Policy Optimization (GePPO) algorithm employed to address the POMDP. Our results indicate that the defender's policy for secure VM migration converges quicker than the benchmark Proximity Policy Optimization (PPO) algorithm, highlighting the effectiveness of our approach in NGCN.