An Enhanced Model for Detection of Network Intrusions Using Explainable Artificial Intelligence Through an Enhanced Elman Neural Network

Abstract

The fourth industrial revolution facilitates innovative business models, including customized manufacturing,
continuous surveillance of process conditions and advancement, autonomous choice-making, and online maintenance,
among others. Nevertheless, they are more vulnerable to a wide array of cyber-attacks due to constrained resources
and their varied characteristics. Such threats result in monetary and social harm to firms, along with the loss of
confidential data. This study presents an improved framework for detecting network intrusions using explainable
artificial intelligence (DNI-EAI). The Enhanced Elman Neural Network (EENN) model has been employed for DNI,
using the Improved Fruitfly Optimization (IFFO) method for optimizing variables. This work used SHAP-EAI
technique to enhance the explanation of detection outcomes. The test configuration is established using MATLAB
software, involving Honeypot database as inputs. The investigation indicates that the suggested technique attains
exceptional performance in DNI, with a detection accuracy of 98.5%.