Improving Network Traffic Security with Parametric and Non-parametric Anomaly Detection Techniques

Abstract

Introduction: Anomaly detection in network traffic is a critical component in multiple domains like IoT, Cloud Computing, cybersecurity and other field, focusing on the identification of malicious activities to preserve the integrity of network systems.

Objectives: This research investigates the performance of both parametric and non-parametric machine learning algorithms in detecting anomalies within network traffic datasets. Parametric models such as Logistic Regression and Support Vector Machines (SVM) were evaluated alongside non-parametric methods, including Random Forest and K-Nearest Neighbors (KNN).

Methods: The dataset underwent an extensive preprocessing pipeline to address issues such as missing data, feature normalization, and categorical encoding to improve model accuracy.

Results: Among the different algorithms assessed, Random Forest demonstrated the highest efficacy, achieving an accuracy rate of 98.68%. This notable performance highlights the advantages of ensemble techniques in capturing complex, non-linear patterns inherent in network traffic. The results underscore the significant contribution of machine learning, particularly non-parametric methods, in enhancing anomaly detection systems within cybersecurity.

Conclusions: Furthermore, this study provides valuable insights into algorithm selection for network traffic analysis, facilitating the development of more robust and efficient intrusion detection systems.