DEVELOPMENT OF INTRUSION DETECTION SYSTEM IN IOT USING ARTIFICIAL INTELLIGENCE AND BLOCKCHAIN TECHNOLOGY

Abstract

The increasing adoption of Internet of Things (IoT) technologies has introduced significant security and privacy challenges, necessitating the development of robust Intrusion Detection Systems (IDS). This thesis presents a comprehensive study on enhancing IDS mechanisms for IoT environments by integrating Artificial Intelligence (AI) and blockchain-based security frameworks. The research objectives include conducting a comprehensive literature review of existing IDS approaches, developing AI-driven models for anomaly detection, designing a blockchain-based framework to enhance security and privacy, and performing a comparative analysis with state-of-the-art techniques. To address the challenges of anomaly detection in IoT networks, this research proposes multiple AI-driven IDS models. The first model, Transfer Learning-Enabled Hybrid Model (TL-BILSTM IoT), leverages transfer learning with a hybrid CNN-BiLSTM architecture to detect botnet attacks. The second model, Deep Learning-Enabled Intrusion Detection System for Industrial IoT, combines CNN and Gated Recurrent Units (GRU) for improved detection in IIoT environments. The third model, Alpha-Net, integrates CNN and GRU to ensure dependable and trustworthy intrusion detection with rigorous statistical validation. Additionally, an Explainable AI-based IDS, Cyber-Sentinet, is introduced to enhance interpretability using Shapley Additive Explanations (SHAP), fostering transparency in decision-making. To strengthen security and privacy in IDS, this research develops a blockchain-based framework incorporating Elliptic Curve Cryptography (ECC), Digital Signature Algorithm (DSA), and SHA-512 for enhanced data integrity and authentication. The framework employs a hybrid SADE algorithm for cryptographic key optimization, the Practical Byzantine Fault Tolerance (PBFT) consensus mechanism for secure transactions, and the Interplanetary File System (IPFS) for scalable off-chain storage. A Genetic Algorithm is applied to optimize IDS performance, while an XGBoost-based model is designed to detect intrusions in heterogeneous IoT environments. Furthermore, this thesis explores the application of blockchain in IoT healthcare by proposing a decentralized system for secure medical certificate management. The system integrates Fully Homomorphic Encryption (FHE) and Non-Interactive Zero-Knowledge Proofs (NIZKPs) to vi ensure privacy-preserving computations and verification, facilitating trustworthy data sharing among healthcare entities. A rigorous comparative analysis is conducted, evaluating the proposed IDS models against state-of-the-art techniques using benchmark datasets such as N_BaIoT and Edge-IIoT 2022. Performance metrics, including accuracy, recall, precision, F1-score, and computational efficiency, demonstrate the superiority of the proposed models in detecting intrusions while ensuring security and scalability. This thesis contributes significantly to advancing IDS for IoT by integrating AI, blockchain, and Explainable AI methodologies. The findings provide a strong foundation for future research in securing IoT ecosystems, emphasizing the importance of interpretable AI and decentralized security mechanisms. The conclusions highlight key insights and outline directions for further enhancing IDS frameworks, including the exploration of federated learning and quantum-resistant cryptographic techniques.