Design and Analysis of Intrusion Detection System Based on Ensemble Deep Neural Network and XAI

Abstract

One of the main aims of designing smart objects is to enhance the comfort and efficiency of living beings. The perspective of the Internet of Things (IoT) is rapidly evolving into a technology that provides smart environments. The practice of IoT in healthcare has automated the method of examining the health of patients in real-time. In any real-world smart environment that focuses on the IoT model, privacy and security are considered critical. Any Security flaws in IoT-based systems generate security threats that impact intelligent environment applications and can lead to the loss or malicious replacement of important information. As a result, Intrusion Detection Systems (IDSs) developed for IoT environments are critical for justifying security threats. Design and development of an adaptable and robust intrusion detection system for unpredictable attacks is a challenging job. Today, deep learning, classical ML classifiers, and existing ensemble libraries are widely suggested to build such smart Intrusion Detection Systems. But it is difficult to find a suitable ensemble configuration for a particular dataset and develop a vigorous solution with good accuracy and trustable solution that also explain model output, and why the model made certain decisions. This research is focused on how to construct a smart fusion intrusion detection system with deep learning and ensemble-based learning strategy for intrusion detection to enhance the security of IoT devices. This Ensemble architecture is used by integrating 4 different Bidirectional LSTM as a Base classifier with XG BOOST as a Meta classifier, while explainable AI is used for its trustworthiness and interpretability.