Comparative Analysis of Machine Learning Methods in the Detection of Network Intrusion

Abstract

In the context of increasing network security threats, traditional Intrusion Detection Systems (IDS) face challenges in detecting complex and evolving attacks. This paper presents a comparative study of three machine learning algorithmsLogistic Regression, Naive Bayes, and Multilayer Perceptron (MLP)for network intrusion detection. Using a comprehensive dataset, the performance of these models is evaluated based on metrics such as accuracy, precision, recall, and F1 score. Results show that MLP with two hidden layers significantly outperforms other models, achieving high accuracy and robustness in detecting both normal and anomalous network traffic. The study highlights the limitations of traditional models in handling nonlinear and complex features, while also emphasizing the potential of advanced machine learning techniques to improve detection performance. Future research directions include optimizing model complexity, reducing false positives, and integrating deep learning architectures for enhanced real-time intrusion detection.