Enhanced Adaptive Butterfly Optimizer based Feature Selection for Protecting the Data in Industry based WSN

Industry 4.0 relies heavily on networked systems, facilitating seamless communication between machines, sensors, devices, and humans in real-time. This interconnectedness, coupled with data transparency, fosters comprehensive insights and analytics, facilitating informed decision-making. However, the increasing reliance on wireless sensor networks (WSNs) within Industry 4.0 exposes them to heightened cyber threats. As these networks continuously gather data and refine processes, they become prime targets for cybercriminals. Robust cybersecurity measures are imperative to safeguard sensitive data, particularly with the proliferation of connected devices. To effectively mitigate cybersecurity risks in Industry 4.0 WSNs, this paper proposes a predictive methodology. This approach employs a multi-criteria framework to enhance WSN cybersecurity, leveraging deep-learning algorithms to prioritize threats based on severity and contextual factors. Notably, the study introduces an Enhanced Adaptive Butterfly Optimization Algorithm (EABOA) for feature selection, addressing shortcomings of existing methods such as sluggish convergence, susceptibility to local optima, and lack of population diversity. Furthermore, the paper presents a novel adaptive fragrance model to improve algorithmic performance. Incorporating Lévy flight to facilitate escaping local optima, this model exhibits both high-frequency short-step hopping and low-frequency long-step walking, enhancing convergence speed and accuracy. Comparative analysis demonstrates superior performance of the proposed framework over benchmark representations. This system offers a smart prioritization approach crucial for efficiently identifying and mitigating high-risk breaches. Experimental evaluation attests to its efficacy, achieving an accuracy of 94.76%, precision of 86%, recall of 80%, and an F-score of 80.06%.