Hybrid Ai and Cryptographic Approaches for Secure and Explainable IoT-Based Cyber-Physical Systems

Cyber-Physical Systems (CPS) integrated with Internet of Things (IoT) devices are subject to growing threats of security issues such as threats of intrusion, malware, data tampering and privacy breaches. While the traditional cryptographic methods successfully guarantee the confidentiality and integrity, they do not possess adaptive threat detection capabilities. Conversely, intrusion detection systems based on AI offer good detection of anomalies, but tend to be black boxes and lack trust and interpretation. This paper proposes a novel hybrid framework which synergies between deep learning based intrusion detection approaches with the blockchain enabled cryptographic security and explainable AI (XAI) techniques. Our architecture integrates a hierarchical edge-fog-cloud computing model with federated learning for privacy-preserving distributed intelligence. We implement a CNN-BiLSTM hybrid neural network for multiclass intrusion detection, coupled with Elliptic Curve Cryptography (ECC) for lightweight encryption and blockchain for immutable audit trails. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) provide model interpretability. Experimental evaluation on CICIDS2017, BoT-IoT, and Edge-IIoTset datasets demonstrates 99.47% detection accuracy, 98.3% precision, 99.1% recall, and F1-score of 98.7%. The cryptographic overhead remains minimal at 6.2 % with encryption latency of 78 μs. Our framework achieves 99.2% packet delivery ratio while maintaining energy efficiency of 0.089 mJ per transaction. The proposed system provides a comprehensive solution for secure, explainable, and efficient IoTCPS deployments in critical infrastructure.