The Use of Ensemble Models for Multiple Class and Binary Class Classification for Improving Intrusion Detection...

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Intrusion detection systems (IDS) are crucial for identifying abnormal network behaviors.
• Single classifiers struggle to effectively detect diverse network intruders.
• Machine learning is widely applied, but limitations exist in single-model performance.

Purpose of the Study:

• To improve intrusion detection systems (IDS) performance.
• To propose a novel approach combining Correlation-based Feature Selection (CFS) with ensemble classifiers.
• To achieve high accuracy, high packet detection rate (DR), and low false alarm rate (FAR).

Main Methods:

• Implemented dimensionality reduction using the Correlation-based Feature Selection (CFS) approach.
• Developed refined ensemble models with base classifiers: J48, Random Forest, and Reptree.
• Performed binary and multiclass classification on KDD99 and NSLKDD datasets, categorizing attacks into DoS, Probe, U2R, R2L, and Normal.

Main Results:

• The proposed CFS + Ensemble Classifiers model achieved 0% FAR and 99.90% DR on the KDD99 dataset.
• The model demonstrated 0.5% FAR and 98.60% DR on the NSLKDD dataset.
• Effective feature selection was achieved using 6 and 13 selected features.

Conclusions:

• The proposed CFS + Ensemble Classifiers model significantly enhances intrusion detection capabilities.
• The approach offers a robust solution for identifying various network attacks with high precision.
• This method provides a promising direction for developing more effective and efficient IDS.