Application and Ethical Aspects of Machine Learning Techniques in Networking: A Review

Abstract

The rapid growth of Internet systems in complexity and scale, combined with advances in Machine Learning (ML), has driven the use of ML for data-driven design, optimization, and analysis of network systems. Researchers and practitioners apply ML to address various challenges, including network attack detection and mitigation, efficient resource management, and Quality of Service (QoS) enhancement. This review classifies ML applications in network systems into six key areas: 1) Domain Name System, where ML aids in enhancing security and efficiency; 2) Application Identification, which improves traffic management and user experience; 3) QoS enhancement, where ML models optimize performance metrics; 4) Cloud Services, where ML facilitates scalable resource allocation; 5) Network Security, employing ML for threat detection and prevention; and 6) Traffic Prediction, using ML to anticipate demand and reduce congestion. This survey examines ML techniques and datasets for each area, highlighting significant contributions in addressing key challenges. We also delve into networking-specific knowledge essential for critical ML phases, such as problem formulation, feature engineering, feature selection, and deployment practices. To conclude, we summarize the prevalent practices in network systems and identify research gaps, outlining future directions for ML’s integration into network system development.

Keywords: Machine Learning (ML), Network Security, Domain Name System (DNS), Network System, Internet of Things (IoT), Anomaly-Detection, Decision Trees, Quality of Experience (QoE), Cloud computing, Support Vector Machine (SVM)