Improving network security using deep learning for intrusion detection

Abstract

As cyber threats and network complexity grow, it is crucial to implement effective intrusion detection systems (IDS) to safeguard sensitive data and infrastructure. Traditional methods often struggle to identify sophisticated attacks, necessitating advanced approaches like machine learning (ML) and deep learning (DL). This study explores the application of ML and DL algorithms in IDS. Feature selection techniques, such as correlation and variance analysis, were employed to identify key factors contributing to accurate classification. Tools like WEKA and MATLAB supported data pre-processing and model development. Using the UNSW-NB15 and NSL-KDD datasets, the study highlights the superior performance of random forest (RF) and multi-layer perceptron (MLP) algorithms. RF ensemble decision trees and MLP multi-layered architecture enable accurate attack detection, demonstrating the potential of these advanced techniques for enhanced network security.