A Deep Learning–Enhanced Blockchain Architecture for Intrusion Detection and Classification

An Intrusion Detection System (IDS) is a significant cybersecurity process that comprises network traffic monitoring for malicious activity and taking appropriate protective actions. However, inadequate training data or inappropriately selected thresholds often restrict the performance of these systems, leading to low detection rates. Blockchain (BC) technology can offer a secure, decentralized, and immutable ledger to monitor suspicious activities over time and classify intrusions globally. Integrating advanced Deep Learning (DL) and BC improves detection and overall security. The decentralized nature of BC eliminates single points of failure. DL can efficiently detect patterns and anomalies, mitigating false alerts, and when integrated with public BC, it ensures secure, transparent, and tamper-proof storage of intrusion data. This study presents a BC-assisted Coati Optimization Algorithm (COA) with DL for Intrusion Detection and Classification (BCOADL-IDC) method, using the BC architecture to ensure data integrity and immutability. Initially, min-max normalization is applied to scale the input data. Then, an Attention-based Bidirectional Recurrent Neural Network (ABiRNN) model is utilized for intrusion detection. COA is used to fine-tune the crucial parameters of the ABiRNN model to improve detection performance. Finally, the integration of BC helps to ensure the integrity of the detection results, prevent tampering, and provide a transparent and secure record of network actions. The comparative study of the BCOADL-IDC approach showed a higher accuracy of 98.79% over existing models on the ToN_IoT dataset.