A Comparative Study of Machine Learning Models for Network Traffic Classification in the IoT Landscape

Network security is a top priority in today’s digital economy, where intrusions grow in sophistication and size. To safeguard network infrastructure and data, it is critical to discern between legitimate and malicious network traffic. Machine learning classification models have proven useful in this task, with decision trees, random forests, and support vector machines among the most common options. This study analyzes six machine learning techniques for determining whether network data is benign or malicious. The performance evaluation is based on the accuracy, precision, F1-score, AUC ROC score, and computational efficiency of the latest IoT dataset, the TII-SSRC-23 Dataset from Kaggle. The findings shed light on each technique’s strengths and drawbacks, as well as recommendations for choosing the best approach for network traffic classification.