What is it about?
The prevention of attacks, policy violations, and unauthorized access to networks is greatly enhanced by intrusion detection systems, commonly referred to as IDSs. Data preprocessing methods and classification models utilized to increase precision and decrease training and testing time for models are crucial to IDS efficacy. An enormous data set including irrelevant or redundant information might drastically reduce performance due to the efficacy of deep learning models. The HDL-IDS suggested in this research is a combination of deep learning (DL) and feature selection techniques.
Featured Image
Photo by Domenico Loia on Unsplash
Why is it important?
The suggested method protects network data from any cyber threat by utilizing a detection model. Deep learning and feature selection are the classification strategies used by the suggested model, which leads to improved accuracy and precision. Because our approach reduces computational time and complexity while lowering the effect of overfitting caused by duplicate feature removal using feature selection methods, it is important to note that current methods are primarily focused on accuracy.
Perspectives
To better detect unknown threats, anomaly-based IDSs learn to distinguish between normal and abnormal network behavior. The suggested model was able to decrease features to attain 100% detection accuracy in 5 seconds when detecting 10% of features using ANOVA, compared to 99.99% at time 67 seconds before feature selection, according to experimental results obtained using the UNSW-NB15 and NSL-KDD datasets.
Muna Hadi Saleh
University of Baghdad
Read the Original
This page is a summary of: HDL-IDS: Integrating feature selection methods with hybrid deep learning for improved intrusion detection systems, January 2025, American Institute of Physics,
DOI: 10.1063/5.0261847.
You can read the full text:
Contributors
The following have contributed to this page
