<scp>DLB</scp> ‐ <scp>IoT IDS</scp> : Deep Learning and Blockchain Based <scp>IoT</scp> Intrusion Detection and Prevention System

ABSTRACT The Internet of Things (IoT) ecosystem is growing rapidly, which has made ensuring the security and integrity of connected devices and transmitted data a significant challenge. In contrast, numerous research efforts have integrated deep learning with blockchain to address intrusion detection related issues. But still, there are some issues in terms of prevention capability, accuracy, and system robustness. In order to detect and prevent cyberattacks in IoT environments, this paper offers a new intrusion detection and prevention system (DLB‐IoT‐IDS) that is based on blockchain technology. The system introduces a multilayered security architecture that leverages Elliptic Curve Cryptography (ECC) for lightweight key generation, smart contracts for secure key registration, and a proof of authority (PoA) mechanism for validating legitimate nodes and transactions on the blockchain. A novel Elite Chaotic based Golden Jackal Optimization (Ec‐GJO) algorithm is employed for optimal feature selection, improving detection precision. For classification, a uniquely designed Probabilistic Dwarf mongoose self‐attention based dense assisted Bi‐LSTM (PDSAttdBi‐LSTM) is introduced, which enhances detection accuracy and reduces false positive rates. The proposed model can obtain an accuracy of 99.72% and 99.80% for the UNSW‐NB 15 dataset and the ToN‐IoT dataset. The proposed model is better for intrusion detection than other compared models. This approach offers a comprehensive and scalable solution for securing IoT networks through both prevention and detection mechanisms.