Application and Performance Analysis of Data Preprocessing for Intrusion Detection System

摘要

In the era of network and big data, network information security has become a major issue. Intrusion Detection System (IDS) is an essential component of network security facilities, which utilizes network traffic data to detect attacks. IDS can adopt data analysis and data mining technologies to detect attacks to network systems. However, the computational overhead of IDS is too large to serve for real-time detection due to the redundancy and irrelevant features in the network traffic dataset. We hence analyze seven classification algorithms for intrusion detection, where we separately perform data preprocessing with two kinds of dimensionality reduction techniques, Principal Component Analysis (PCA) and Singular Value Decomposition (SVD), to improve the performance of IDS. The experimental results on the NSL-KDD dataset indicate that the classification algorithms with dimensionality reduction outstands in detection rate and detection speed. Meanwhile, SVD demonstrate its superiority to PCA in boosting these algorithms.