<![CDATA[The increasing complexity and volume of network traffic have created challenges for attack detection systems in recognizing diverse attack patterns and handling imbalanced class distributions. This study aims to detect network attacks on the UNSW-NB15 dataset using supervised learning methods, namely Logistic Regression and Random Forest. The research stages include data preprocessing, model development, and evaluation using Stratified 5-Fold Cross Validation with Accuracy, Precision, Recall, F1-Score, and AUC metrics. The results show that Random Forest achieved better performance than Logistic Regression, with an accuracy of 0.9515, precision of 0.9631, and AUC of 0.9924. Cross-validation results also showed that the average accuracy of Random Forest was 0.9504, higher than Logistic Regression at 0.9038. Feature analysis indicates that attributes based on Time to Live, traffic, data volume, and temporal characteristics contribute significantly to the detection process. Therefore, Random Forest demonstrates more optimal and stable performance in detecting network attacks on the UNSW-NB15 dataset, while Logistic Regression remains relevant as a simple and interpretable comparison model.]]>
Copyrights © 2026
