Automated Infrastructure Cybersecurity Management Using Deep Neural Networks: A Network Intrusion Detection Case Study

Abstract

This study presents a Deep Neural Network (DNN)-based framework for Automated Infrastructure Management (AIM), the main aim being to develop and evaluate a DNN-based framework for real-time intrusion detection in automated infrastructure systems. Leveraging the UNSW-NB15 dataset, the research addresses the limitations of traditional rule-based systems by employing advanced techniques such as SMOTE for handling data imbalance, batch normalization for training stability, and feature selection to optimize model performance. The developed DNN model achieved an accuracy of 99.61% and an AUC of 0.9993, demonstrating exceptional capability in classifying normal and attack traffic with high precision and recall. While the results highlight the potential of DNNs to revolutionize infrastructure management through predictive maintenance and real-time threat detection, challenges such as reliance on synthetic data, computational demands, and cross-domain generalizability remain. The study underscores the importance of integrating real-time data, developing lightweight models for edge deployment, and addressing ethical considerations to ensure scalable and trustworthy AIM solutions in diverse infrastructure environments.