Interpretable and Adaptive Security Mechanism for Next-Gen Industrial Networks

Abstract

In the era of Industry 5.0, the evolution of smart and interconnected industrial environments calls for highly adaptive and interpretable security mechanisms. With the rapid proliferation of IoT devices, automation systems, and cloud-integrated platforms, industrial networks have become increasingly vulnerable to complex and evolving cyber threats. Ensuring the resilience of such environments requires the deployment of intelligent intrusion detection techniques capable of operating in real-time with high accuracy and interpretability. To address this necessity, an advanced security mechanism was developed using the CIC-DDoS 2019 dataset, which provides a comprehensive set of simulated Distributed Denial of Service (DDoS) attacks across multiple protocols in industrial network settings. The detection system was enhanced using prominent feature selection techniques such as SelectKBest, SelectPercentile, and Mutual Information. These techniques helped identify the most relevant attributes from the dataset, effectively reducing dimensionality while improving model generalization and training efficiency. The core of the detection system is a robust ensemble classification strategy, leveraging a voting-based classifier that integrates the strengths of bagging through Random Forest and boosting through Boosted Decision Trees. This hybrid ensemble architecture enables precise classification by combining the diversity and stability of multiple models. Among all tested configurations, the voting classifier demonstrated superior performance, achieving a high accuracy of 95.8%, thereby confirming the strength of ensemble learning in dynamic and heterogeneous network environments. The proposed detection system is scalable and adaptive, designed to function in real-time, and supports decision transparency—crucial for building trust in automated industrial defense systems. This mechanism offers a dependable layer of security against network intrusions and enhances the resilience of hyper-connected infrastructures, fulfilling the cybersecurity demands of next-generation industrial systems under the Industry 5.0 vision.