Iterative-tuning support vector machine for network traffic classification

摘要: Accurate and timely traffic classification is a key to providing Quality of Service (QoS), application-level visibility, security monitoring for network operations management. A class techniques have emerged that apply machine learning technology predict the application flow based on statistical properties flow-features. In this paper, we propose novel iterative-tuning scheme increase training speed algorithm using Support Vector Machine (SVM) learning. Meanwhile derive equations obtain SVM parameters by conducting theoretical analysis SVM. Traffic carried out flow-level information extracted from NetFlow data. Performance evaluation demonstrates proposed exhibits two ten times faster than eight other previously found in literature, while maintaining comparable accuracy as those techniques. presence millions flows Terabytes data network, speeds essential making viable option real-world deployment modules. addition, operators cloud service providers can address range issues including semi-real-time engineering.