IMPROVING THE EFFICACY OF NETWORK SECURITY BASED ON DIMENSIONALITY REDUCTION TECHNIQUES

Abstract

This paper focuses on proposing a network intrusion detection model applying fundamental machine learning techniques to enhance early detection of network intrusions (rapid detection of attack behaviors) for improved efficiency in preventing network attacks. The system must still ensure technical accuracy in providing high-precision alerts. The research employs several dimensionality reduction techniques to detect abnormal network intrusions caused by Distributed Denial of Service (DDoS) attacks. The proposed model aims to reduce computation time for early attack detection. The results show that the proposed system performs best across all three datasets through the combination of the KNN algorithm and the Feature Importance dimensionality reduction technique. After calculating and returning the number of important features in attack detection using the Importance technique, the performance of the KNN algorithm is enhanced. By retaining only important features, as the dimensionality of the data decreases, the computation speed of KNN increases. Therefore, although the accuracy may slightly decrease, the computation time is significantly reduced. This is acceptable for practical purposes.