A robust intrusion detection system based on a shallow learning model and feature extraction techniques.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Cybersecurity threats are increasing, necessitating advanced detection strategies.
• Existing deep learning models can be computationally intensive and require extensive features.
• There is a need for efficient and accurate intrusion detection systems.

Purpose of the Study:

• To propose a novel cybersecurity detection model integrating feature selection and shallow learning.
• To evaluate the model's performance using benchmark datasets (NSL-KDD and UNSW-NB15).
• To demonstrate the model's superiority over traditional deep learning approaches.

Main Methods:

• Feature selection using SHapley Additive exPlanations (SHAP) values.
• Implementation of a shallow learning algorithm, PV-DM.
• Classification using XGBOOST machine learning algorithm.
• Validation on NSL-KDD and UNSW-NB15 datasets.

Main Results:

• Achieved 98.92% accuracy, 98.92% precision, 95.44% recall, and 96.77% F1-score on NSL-KDD using only four features.
• Attained 82.86% accuracy, 84.07% precision, 77.70% recall, and 80.20% F1-score on UNSW-NB15 using six features.
• Demonstrated superior performance compared to traditional deep learning models across all metrics.

Conclusions:

• The proposed model offers an efficient and effective solution for cybersecurity threat detection.
• Feature selection with SHAP values significantly enhances model performance and reduces complexity.
• The integration of PV-DM and XGBOOST provides a robust and accurate detection mechanism.