Application of BukaGini algorithm for enhanced feature interaction analysis in intrusion detection systems.

Area of Science:

• Machine Learning
• Cybersecurity
• Data Science

Background:

• Intrusion Detection Systems (IDS) require robust feature selection for effective performance.
• Existing feature interaction analysis methods in IDS need improvement.
• BukaGini is a stability-aware Gini index feature selection algorithm.

Purpose of the Study:

• To evaluate the effectiveness of BukaGini in enhancing machine learning model performance for IDS.
• To assess BukaGini's feature interaction analysis capabilities within the IDS domain.
• To compare BukaGini's performance against existing methodologies.

Main Methods:

• BukaGini algorithm utilizing the Gini index for feature selection.
• Evaluation across four diverse IDS datasets: NSLKDD, WUSTL EHMS, WSN-DS, and UNSWNB15.
• Performance assessment using metrics: stability score, accuracy, F1-score, recall, precision, and ROC AUC.

Main Results:

• BukaGini achieved accuracy rates up to 99% and stability scores exceeding 99% across all datasets.
• Average dimensionality reduction of 25% was observed, with 10 features selected per dataset.
• Significant advancements in IDS performance metrics were noted.

Conclusions:

• BukaGini is a promising solution for feature interaction analysis in cybersecurity, particularly for IDS.
• The algorithm contributes to robust model performance and enhanced intrusion detection capabilities.
• BukaGini shows potential for real-world application in improving cybersecurity defenses.