Deep Reinforcement Learning for Dynamic Network Slice Security Using Moving Target Defense
Andreas Andreou, Constandinos X. Mavromoustakis, Houbing Herbert Song, Evangelos Markakis, Athina Bourdena, George N. Mastorakis · 2025
Network slicing has emerged as a transformative enabler for meeting the diverse requirements of 5G and beyond networks, including 6G. However, network slices’ dynamic and virtualized nature introduces significant security challenges, particularly against evolving cyber threats. We propose a Deep Reinforcement Learning (DRL)–based Moving Target Defense (MTD) strategy tailored for secure network slicing to address these challenges. Our approach utilizes a Q-Learning framework to manage MTD actions dynamically, optimizing security while maintaining service quality. Extensive simulations demonstrate the effectiveness of our framework in minimizing attack success rates and ensuring operational stability, significantly outperforming baseline methods such as random decision-making.