Enhancing Cyber-Physical Systems Resilience: Adaptive Self-Healing Security Using Long Short-Term Memory Networks
Franciskus Antonius Alijoyo, Chamandeep Kaur, Afsana Anjum, Veera Ankalu Vuyyuru, B. K. Bala · 2024
Cyber-Physical Systems (CPS) form the backbone of critical infrastructures, integrating computational and physical processes to enhance efficiency and automation. However, the increasing interconnectivity exposes these systems to diverse cyber threats, necessitating proactive security measures. This research seeks to advance the security of Cyber-Physical Systems (CPS) through the implementation of a self-healing mechanism driven by neural networks. CPS, pivotal in critical infrastructures, have become increasingly susceptible to a myriad of cyber threats owing to their intricate interconnectivity. The paramount significance of this research is rooted in the creation of a dynamic and intelligent defense system capable of autonomously identifying, responding to, and recuperating from cyber-physical attacks. The traditional CPS security landscape has grappled with static and rule-based approaches, struggling to keep pace with the dynamic nature of contemporary cyber threats. Moreover, the recovery processes in place have been predominantly manual and time-consuming. This research addresses these longstanding issues by introducing LSTM into the CPS security framework. This incorporation represents a paradigm shift, ushering in an era of adaptive resilience. The novelty of the research lies in the seamless integration of neural networks, enabling the system to learn from past incidents and adapt to emerging threats. The proposed self-healing mechanism emphasizes real-time threat detection, allowing for swift responses and the automation of the recovery phase, ultimately reducing downtime associated with security incidents. the integration of self-healing mechanisms using Long Short-Term Memory (LSTM) networks proves to be a promising approach for advancing cybersecurity in Cyber-Physical Systems (CPS), with the proposed model achieving an impressive accuracy of 99%. The research not only tackles existing vulnerabilities but also pioneers a transformative approach to CPS security, leveraging the capabilities of neural networks to create a more robust and adaptive defense mechanism against evolving cyber threats.