Comparative Analysis Of Machine Learning Algorithms For Network Attack Detection Using NetFlow Data

Authors: Pavithra V, Dr. D. Rajinigirinath, T. Saranya

Abstract: Machine learning has become an efficient approach for automating network intrusion detection. This paper compares three supervised algorithms—Random Forest (RF), Support Vector Machine (SVM), and Gaussian Naïve Bayes (GNB)—for classifying NetFlow traffic into normal and attack categories. The preprocessed dataset developed in the previous phase was used for model training and validation. Each algorithm was evaluated using accuracy, precision, recall, and F1-score metrics. Results show that Random Forest achieved 92% accuracy, outperforming SVM (85%) and GNB (78%). This demonstrates that ensemble-based learning provides better generalization for adaptive network attack detection systems.

DOI: http://doi.org/10.5281/zenodo.17471269