An Intelligent Deep Learning-Based Intrusion Detection System for IoT Nodes using CNN- BiLSTM and Grasshopper Optimization

Abstract

Recent networks have a remote larger attack surface owing to the quick spread of Internet of Things (IoT) strategies, which for effective and instantaneous Intrusion Detection Systems (IDS). Deep-IDS, a real-time Deep Learning (DL) IDS intended for IoT nodes with limited resources, is presented in this work. The raw network traffic data fed to data pre-processing using data cleansing and missing value to improve the quality of the data. Normalization after pertinent features is recovered and design to capture both longitudinal as well as sequential features of system behaviour. The proposed Bidirectional Long Short-Term Memory (BiLSTM) - Convolutional Neural Network (CNN) mimic sequential dependencies in traffic flows, then train spatial feature illustrations in Deep-IDS hybrid CNN-BiLSTM architecture. The Grasshopper Optimization Algorithm (GOA) for optimal hyper parameter tuning, which further improves detection accuracy and overview capability. According to experimental results, the Deep-IDS are appropriate for real-time deployment in IoT environments then it provides high detection accuracy of 99\% with low false alarm rates. The proposed achieves high detection accuracy of 99\% with low false alarm rates, showing strong reliability and robustness in identifying multiple attack categories