基于机器学习的即时通信流量分类技术

目录

Chapter 1 Introduction

1.1 Research Background

1.1.1Port based traffic classification

1.1.2Deep packet inspection

1.1.3Machine learning based traffic identification

1.2Survey of related work

1.2.1Network traffic classification feature description and extraction

1.2.2Supervised learning traffic classification

1.2.3Un-Supervised learning traffic classification

1.2.4Early stage traffic classification

1.3Practical Background of Network Traffic Classification

1.3.1Bayes net machine learning classifier

1.3.2Na(i)ve Bayes machine learning classifier

1.3.3Support vector machine learning classifier

1.3.4C4.5 decision tree machine learning classifier

1.4Internet trace trafficdata set

1.5IM traffic classification result analysis

1.5.1Performance measurement

1.5.2Results and analysis

1.5.3Contributions for IM traffic classification research

1.7Structure of Thesis

Chapter 2Effective Feature Selection for IM Application Traffic Classification

2.1 Introduction

2.2Feature Selection Metrics

2.2.1Mutual Information Based Metric

2.3Proposed Method

2.3.1Weighted Mutual Information (WMI) Metric

2.3.2ACC Metric

2.3.3WMI_ACC Algorithm

2.3.4Statistical Test

2.4Evaluation Methodology

2.4.1Data Sets

2.4.2HIT Trace I Dataset

2.4.3NIMS Dataset

2.4.4Performance Measures

2.5Experimental Results and Analysis

2.5.1Wilcoxon Pairwise Statistical Test Result

2.5.2Selected Features of Our Propose Algorithm

2.5.3 Comparison

2.6Analysis and Discussion

2.7 Summary

Chapter 3Feature Selection forImbalance IM Applications Traffic Classification

3.1 Introduction

3.2Related Work

3.3 Methodology

3.3.1Feature Selection Metrics

3.3.2AUC Metric

3.3.3Feature Selection Algorithms

3.3.4WMI_AUC Algorithm

3.3.5RFS Algorithm

3.4Evaluation Methodology

3.4.1Data Sets

3.4.2Evaluation Criteria for Performance Measurements

3.5Experimental Results and Analysis

3.5.1Analysis Results of HIT Trace 1 Dataset

3.5.2Analysis Results of NIMS Dataset

3.6Analysis andComparison

3.7 Summary

Chapter 4 Robust Feature Selection Approach for IM Applications Traffic Classification

4.1 Introduction

4.2 Methodology

4.2.1FSA and FEA Proposed Methods

4.2.2Feature Selection Based Metrics

4.2.3Symmetrical Uncertainty Based Metric

4.3The Feature Selection Approach (FSA)

4.3.1The Proposed Algorithm

4.3.2The Feature Evaluation Approach (FEA)

4.4Evaluation Methodology

4.4.1 Datasets

4.4.2Performance Measurement

4.5Experimental Results and Analysis

4.6Analysis andComparison

4.7 Summary

Chapter 5 Effective Feature Packet for IM Application At Early Stage Traffic Classification

5.1 Introduction

5.2Data Sets

5.2.1HIT Lab Trace Dataset

5.2.2HIT Dorm Trace Dataset

5.3Proposed Model

5.4 Methodology

5.4.1Machine Learning Classifiers

5.4.2Statistical Test

5.5Evaluation Criteria for Performance Measurement

5.6Results and analysis

5.7Mutual Information Results of the HIT Trace I Data Set

5.7.1Mutual Information Results of the HIT Trace II Data Set

5.7.2Analysis Results of HIT Lab Trace Data Set of the Text Messages Traffic Data Set

5.8 Summary

Conclusions

参考文献

基于机器学习的即时通信流量分类技术

Published Papers

声明

致谢

Resume