Machine learning: enhanced dynamic clustering for privacy preservation and malicious node detection in industrial internet of things

• Published in: Discover Computing Vol. 28; no. 1; pp. 166 - 27
• Main Authors: Hasan, Nabeela, Saleem, Saima, Khan, Mudassir, Alabdultif, Abdulatif, Nezami, Mohammad Mazhar, Alam, Mansaf
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.12.2025; Springer Nature B.V; Springer
• Subjects: Automation; Clustering; Computer Science; Cryptography; Cybersecurity; Data integrity; Data management; Data Mining and Knowledge Discovery; Data Structures and Information Theory; Dynamic clustering; Industrial applications; Industrial Internet of Things; Industrial internet of things (IIoT); Information Storage and Retrieval; Internet of Things; Machine learning; Malicious node detection; Natural Language Processing (NLP); Pattern Recognition; Privacy; Privacy preservation; Privacy preservation; Malicious node detection; Industrial internet of things (IIoT); Dynamic clustering; Cryptography; Machine learning
• ISSN: 2948-2992; 1386-4564; 2948-2992; 1573-7659
• DOI: 10.1007/s10791-025-09689-w

The Industrial Internet of Things (IIoT)continues to redefine industrial automation through connected smart devices, yet it remains highly vulnerable to privacy breaches and malicious intrusions. This research introduces ML-DCPP, a Machine Learning-based Dynamic Clustering and Privacy Preservation framework tailored to safeguard IIoT ecosystems. By integrating adaptive clustering via the LEACH protocol, secure key distribution through Public Key Generators (PKGs), and private data aggregation using the Location Privacy Tree (LPT) and Chinese Remainder Theorem (CRT), the framework addresses both communication efficiency and data confidentiality. To enhance security at the edge, differential privacy with Laplace noise is incorporated, and a Random Forest model is employed to detect malicious behavior with 99.89% accuracy, validated by a highly significant t-test (–7.87, p < 0.0001). Experimental results reveal notable system improvements, including a reduction in latency to 0.45 s, an increase in throughput to 280 tps, and an optimized cluster convergence time of 4.91 s. ML-DCPP not only reinforces privacy and security measures but also demonstrates scalability and low overhead, making it a practical and forward-looking solution for securing industrial IoT infrastructures.