Enhancing Cybersecurity: Two-Phase Detection Approach for Intrusion Network for Anomaly Data

This research explores the performance of two-phase classification methods in intrusion detection systems (IDS) with various algorithms, including SVM + Naive Bayes, Random Forest + KNN, and other combinations. This research validates the superiority of combination models over individual ones, highlighting RF+KNN as the most effective combination, followed by RF+SVM and SVM+NB. These findings offer valuable insights for cybersecurity practitioners seeking to enhance security measures by selecting balanced model combinations based on performance metrics and computing resources. The RF, KNN, and RF + KNN methods emerge as optimal choices, boasting an accuracy, F1 score, recall, and precision of 99.998%. This combination proves to be highly reliable in minimizing identification errors and detecting potential threats. While the Naive Bayes method demonstrates swift processing times, its detection efficacy lags significantly behind RF, KNN, and RF + KNN. This research underscores the critical importance of strategic model selection for optimal performance in intrusion detection systems.