Decision Tree for Boosting Firewall Accuracy: Optimizing Action Classification for Enhanced Cybersecurity
Prasert Teppap, Wirot Ponglangka, Prasert Luekhong · 2024
This study aims to categorize many “actions” types presented in the comprehensive dataset, known as “Internet Firewall Data.” We employed this dataset to train and assess several models, including KNN, NB, RF, DL, and DT, aiming to identify a highly accurate classification model for firewall log data. Through 10-fold cross-validation, we determined that the RF and DT models achieved robust performance, with accuracies of 99.86% and 99.85%, respectively. Moreover, the NB and DT models exhibited the quickest processing times, measured at 0.22 s and 0.35 s, respectively. These findings underscore the substantial impact of reducing the number of “actions” on the model's capacity for discernment and precision. This paper offers insights and recommendations valuable to both cybersecurity professionals and researchers, advocating for the consolidation of firewall action groups to enhance manageability. Simplifying firewall configurations not only bolsters security but also diminishes operational complexities.