Enhancing Cybersecurity Through Fast Machine Learning Algorithms

As cyber attacks grow in complexity, their detection becomes increasingly challenging, underscoring the critical role of advanced machine learning methods in cybersecurity. These techniques are pivotal for identifying and mitigating network anomalies and intrusions, employing diverse machine learning models to uncover harmful activities of network users. In this paper, we explore the efficacy of various fast machine learning models on the UNSW-NB15 dataset for detecting nine attacks: Fuzzers, Analysis, Backdoors, DoS, Exploits, Generic, Reconnaissance, Shellcode, and Worms. Classification models include Gaussian Naïve Bayes, K-Nearest Neighbors, Decision Tree, Random Forests, Extra-Trees, and XGBoost, evaluated based on Accuracy, F1-Score, Precision, Recall, Training Time, and Test Time. Additionally, we conduct both binary and multiclass classifications to thoroughly evaluate the performance of these models.