Dynamic Parameter Optimization for Industrial Internet Security Models Using Neural Networks
International journal of intelligent engineering and systems · 2025
The Industrial Internet of Things, or IIoT, transforms industries by allowing complex systems to interact with each other, but opens up industries to new and more advanced threats, including DDoS attacks, ransomware, and malware.These current security models for IIoT inherently use static optimization approaches that do not match realtime changes and other attack patterns.To overcome these limitations, this study calls for developing the Dynamic Parameter Optimization Framework that will incorporate the use of both the Neural Network and Black Winged Kite Algorithm (BKA) to optimize a given process in real time.The framework incorporates several models: XGBoost, a spatial CNN particularly designed using PSO with Default, and a BKA for tuning the CNN for dynamic parameters.Thus, the proposed method, which balances the exploration and exploitation capabilities of BKA optimizes such variables like learning rates and weights well.Moreover, Proposed mechanisms make consideration of changes in threats, as well as a large-scale IIoT environment, possible.The evaluations of TON_IoT and UNSW-NB15 datasets presented the framework's efficiency, where the accuracy achieved 96.8% and F1-score of 95.3%, outperforming the Particle Swarm Optimization (PSO) and XGBoost methods.The framework also provided a 20% improvement in convergence time and adaptation capability by updating the parameters in real-time.This study provides a foundation to develop a sound, elastically adaptive, and autonomously executable security framework for IIoT that calls for minimal human intervention to tackle cyber threats.Future work plans to develop this framework to work with multimodal data and to perform well in various industrial environments.