Opti-IntrusionNet: Hybrid Deep Learning Entrenched End-to-End Intrusion Detection System Framework Design for Blockchain Assisted Egde-IoMT Environment

Abstract

 Malicious cyber threats and data privacy were the major issue in Internet of Medical Things (IoMT) environment due adoption of cutting-edge
technologies such Artificial Intelligence (AI) and Blockchain. There were lot of research works undertook by many of the existing works by designing Intrusion
Detection System (IDS) using AI and blockchain technologies respectively. However, the essence of emancipating the security and privacy clutches in IoMT
environment were not yet resolved. To address this issue, we have designed an End-to-End IDS framework for IoMT environment using hybrid deep learning
method named Optimized Intrusion detection Network (Opti-IntrusionNet). Initially, we perform authentication by the Trusted Authority (TA) to resolve the
unwanted malicious traffic at the initial stage using Hash based Block Cipher-256 (HBC-256) algorithm. Upon authenticated, the secrecy and privacy of the
information is ensured by performing optimal channel selection method using Binary Grasshopper Optimization Algorithm (BGOA) based on adequate channel
related metrics. Followed by, we perform malicious traffic detection at the distributed edge servers using hybrid of two deep learning algorithms named Optimized
AlexNet and Gated Recurrent Unit (OA-GRU). The optimized AlexNet is utilized for feature extraction whereas the GRU is utilized for intrusion classification into
normal and malicious traffic respectively. At last, all the IoMT environment information are stored in the cloud assisted blockchain using Delegated Proof of
Accessibility (DPOA) consensus algorithm. The implementation of the proposed work is carried out using python 3.11.4 with valid python and machine learning
libraries respectively. The performance of the proposed work is analyzed from several performance metrics such accuracy, precision, recall, F-score, and ROC-AUC
that shows that the proposed Opti-IntrusionNet outperforms than the existing works.