Bi-Level Intrusion Detection in IoT Networks Using Ensemble Method and A-GRU-RNN Classifier

Abstract

For minimizing security vulnerabilities and attacks, which affect Internet of Things (IoT) applications’ performance, it is essential to design an efficient authentication protocol owing to the rapid deployment of IoT. This paper proposes a bi-level Intrusion Detection (ID) in IoT using an ensemble and Arctan-based Gated Recurrent Unit-Recurrent Neural Network (A-GRU-RNN) classifier. This work mainly concentrates on both major and minor attacks. Allowing the trusted nodes to join the network is the other motive of this work. Here, the nodes are registered to the network and then initialized. Then, to verify the trust levels of the nodes, test packet transmission is performed. By employing the Deterministic Initialization Method-centric K-Means algorithm (DIM-K-Means) algorithm, the trusted nodes are formed into the cluster. Next, the respective CH is selected by Multiplex-Valued Encoding Sea Lion Optimization (MVE-SLO). Afterwards, by employing Multi-Point Relays-Optimized Link State Routing (MPR-OLSR), routing is taken place. The steps, namely preprocessing, attribute extraction, attribute reduction, and classification are taken to detect the confidentiality of the sensed data. The classification phase significantly determines whether the data is attacked or not. If the data is attacked, then it detects the attack type. Here, the publicly available dataset is used. As per the experimental outcome, the proposed method withstands high-security levels when analogized to the prevailing methodologies.