An Adaptive Edge-Blockchain Framework for Real-Time IoT Threat Detection Using Deep Learning and Big Data Analytics a Review

Massive and diverse data has been produced by the Internet of Things' (IoT) explosive growth, posing serious problems for threat detection, privacy protection, scalability, and network edge resource management. Static machine learning models and traditional centralized intrusion detection systems frequently find it difficult to adjust to changing attack patterns and new threats. To overcome these constraints, recent research has looked into explainable AI, blockchain, edge computing, federated learning, and optimization methods. With an emphasis on blockchain-enabled federated learning, edge-based intrusion detection, optimization-based feature selection, and explainable AI techniques, this paper offers a thorough overview of secure and explainable IoT threat detection methodologies. The review looks at how blockchain and smart contracts improve trust, integrity, and security in distributed model aggregation, while federated learning facilitates privacy-preserving collaborative model training. Additionally examined is the function of optimization methods, such as particle swarm optimization, in lowering computing overhead and feature dimensionality. Additionally, SHAP and associated explainable AI techniques are examined to enhance threat detection models' interpretability and transparency. When comparing current methods, benchmark datasets such as CICIDS2017 and TON_IoT are taken into account. In order to build real-time and safe IoT threat analytics, important research gaps and future directions pertaining to scalability, zero-day attack detection, computing efficiency, privacy, trust, and explainability are finally outlined.