|
Original research
OPTIMIZING NETWORK INTRUSION DETECTION FOR ENHANCING DIGITAL FINANCIAL SYSTEM SECURITY AND INVESTMENT ENVIRONMENT STABILITYPages 113-122
Anh Huynh Van,
Duyen Ngo Ky,
Nhi Nguyen Thi Yen,
Vy Nguyen Thi,
Abstract
As internal networks become increasingly essential for modern IT systems, they face heightened risks from cyber threats, including network intrusions that can disrupt operations and compromise sensitive data. This study focuses on enhancing Intrusion Detection Systems (IDS) by utilizing NetFlow datasets to detect anomalies and potential intrusions efficiently. By integrating Principal Component Analysis (PCA) for feature reduction and applying Grid Search for hyperparameter tuning, the proposed model achieves faster execution times and improved detection accuracy. Initial evaluations demonstrate the model’s effectiveness in binary classification scenarios, suggesting its scalability for multi-class classification tasks. These findings contribute to developing a practical and adaptable solution for securing internal networks against evolving threats. This study achieved over a 97% reduction in training time through the Binary Classification experiment and improved training accuracy by over 12% and test accuracy by 21% through the Multiclass Classification experiment.
Keywords: K Nearest Neighbors (KNN) classification, PCA feature, Netflow Datasets, Binary classification, Multiclass classification, Intrusion Detection System.
|
Phan-Anh-Huy Nguyen 
PDF 