Machine Learning Based Network Traffic Classification Detection Research
Zhuohong Zhang, Chin Soon Ku · Advances in transdisciplinary engineering · 2025
In the digital era, the explosive growth and increased complexity of network traffic have brought great challenges to traditional security detection methods, especially when dealing with new types of attacks such as APT and DDoS. Machine learning, with its advantages in complex tasks, provides new ideas for network security protection. This paper is based on the CICIDS-2017 dataset, which contains nearly 2.85 million network traffic records, covering normal traffic and multiple attack types. After pre-processing steps such as data cleaning, feature extraction and selection, three machine learning models, namely, random forest, decision tree and K-nearest neighbour, are established and optimized for the random forest model. The experimental results show that the optimised Random Forest model performs well in terms of accuracy, precision, recall and other metrics, and the AUC value is close to perfect, which is able to identify various types of network attacks efficiently. Comparative analysis shows that Random Forest has significant advantages in dealing with complex classification problems. This study provides an efficient and reliable detection solution for the field of network security, which has significant theoretical and practical implications.