Deep Learning-based Enhanced Security in Cyber- Physical Systems: A Multi-Attack Perspective

Today{\prime}s key industries, including energy, healthcare, and transportation, rely heavily on Cyber-Physical Systems (CPS) as their foundation. All the same, these systems are seriously threatened by the wide range and growing complexity of cyberattacks. Traditional security measures typically fail when identifying intricate, multi-phase, and unique threats. This research presents a deep learning-based approach to improve CPS{\prime}s resistance against cyber threats. This work aims to detect multiple network attackers with distinct attack natures and provide a novel hybrid architecture. To analyze the performance of the proposed architecture, a publicly available dataset OPC UA is used, which contains Denial of Service (DoS), Man-in-the-middle (MITM), and spoofing attack data. The performance is compared with existing detection frameworks. The findings show a notable increase in detection accuracy over current approaches, underscoring the promise of deep learning strategies in protecting CPS against cyberattacks. This research contributes to cyber-physical security and provides valuable information for implementing intelligent attack detection systems in CPS.