Scholarlink

<p>The Industrial Internet of Things (IIoT) is widely adopted in critical sectors such as manufacturing, healthcare, and energy, but its openness exposes networks to severe cyber threats. Existing machine learning (ML)-based intrusion detection systems (IDS) provide strong detection capabilities but depend on centralised data and remain vulnerable to adversarial manipulation, while blockchain ensures data integrity but lacks adaptive intelligence. To address these shortcomings, we propose a Blockchain-Integrated Federated Transfer Learning (FTL) Framework for real-time IIoT intrusion detection. A custom MQTT-driven IIoT dataset (126,354 instances, 43 features) is developed with both real and simulated attacks (DoS, DDoS, brute force, port scan, malware). The framework employs Ethereum smart contracts for decentralised authentication, tamper-proof access control, and immutable anomaly logging, while FTL enables privacy-preserving, cross-domain model training. Evaluation of classical ML classifiers and anomaly detection methods shows LightGBM achieves 99.68% accuracy, 99.7% precision, 99.6% recall, and 99.65% F1-score. Security and ablation analyses confirm resilience against poisoning, tampering, and replay attacks, with low latency (&#xa0;120&#xa0;ms). The proposed hybrid approach outperforms ML-only and blockchain-only baselines, offering a scalable, secure, and real-time IDS for IIoT infrastructures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Blockchain-Integrated Federated Transfer Learning Framework for Real-Time and Privacy-Preserving Intrusion Detection in Industrial IoT

  • 期刊论文

  • 发表日期：

    2026年7月22日

查看全文
  • Pradeep Kumar,
  • Rajdeep Das,
  • Indrajit Banerjee

摘要

The Industrial Internet of Things (IIoT) is widely adopted in critical sectors such as manufacturing, healthcare, and energy, but its openness exposes networks to severe cyber threats. Existing machine learning (ML)-based intrusion detection systems (IDS) provide strong detection capabilities but depend on centralised data and remain vulnerable to adversarial manipulation, while blockchain ensures data integrity but lacks adaptive intelligence. To address these shortcomings, we propose a Blockchain-Integrated Federated Transfer Learning (FTL) Framework for real-time IIoT intrusion detection. A custom MQTT-driven IIoT dataset (126,354 instances, 43 features) is developed with both real and simulated attacks (DoS, DDoS, brute force, port scan, malware). The framework employs Ethereum smart contracts for decentralised authentication, tamper-proof access control, and immutable anomaly logging, while FTL enables privacy-preserving, cross-domain model training. Evaluation of classical ML classifiers and anomaly detection methods shows LightGBM achieves 99.68% accuracy, 99.7% precision, 99.6% recall, and 99.65% F1-score. Security and ablation analyses confirm resilience against poisoning, tampering, and replay attacks, with low latency ( 120 ms). The proposed hybrid approach outperforms ML-only and blockchain-only baselines, offering a scalable, secure, and real-time IDS for IIoT infrastructures.