This academic paper discusses the potential of autonomous cyber defense in industrial control systems, with a focus on maritime operational technology (OT). It introduces a simulation environment, IPMSRL, for a generic Integrated Platform Management System (IPMS). The paper explores the use of Multi-Agent Reinforcement Learning (MARL) for autonomous cyber defense decision-making on generic maritime-based IPMS OT. It highlights the challenges of cyber defense in OT due to the brittle nature of OT infrastructure, the use of legacy systems, and the limitations of traditional IT-centric cyber defense solutions. The paper presents experiments where MAPPO outperforms IPPO in terms of training performance and optimal policy achievement.
Publication date: 18 Jan 2024
Project Page: http://ceur-ws.org
Paper: https://arxiv.org/pdf/2401.10149
