Improving Accuracy in Network Intrusion Detection via Machine Learning Algorithms: An In-Depth Examination

Intrusion detection is essential for securing computer networks by monitoring and analyzing activity to detect and respond to unauthorized access and malicious behavior. By examining network traffic and system logs, intrusion detection systems (IDS) help protect data integrity and availability, enabling swift responses to potential security incidents. In an ever-evolving cyber threat landscape, IDS plays a critical role in maintaining network security. This research investigates the performance of various machine learning algorithms for classification tasks using a dataset from Kyoto collected on December 31st, 2015. The algorithms explored include Naive Bayes, K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), AdaBoost, and CatBoost. Training and testing the model with different data splits (80/20, 70/30, 60/40) revealed that AdaBoost, LGBM, XGBoost, Decision Tree, Gradient Boosting, and Random Forest consistently delivered the highest accuracy. The study concludes that selecting the appropriate algorithms and data splits is crucial for enhancing IDS effectiveness, providing valuable insights for future research. Explore the code behind this study at https://github.com/Trishlaa07/Kyoto_2015_Dec_day31