A Survey of Machine Learning Based Packet Classification

In this paper, I will categories and analysis different approaches to classify different Internet traffics using Machine Learning (ML) technic. The traffic classification can be used as an important tool to detect intrusion detection. And it also can be used by network operator to control the network. However it opens a topic related to protection of personal information. After realizing the advantage and disadvantage of packet classification, I will briefly introduce three classification methods and related researches. The classification methods are port-based, payload-based and statistical-based classification. And ML is a well-known technic used in statistical-based classification. After a brief introduction of ML, I will focus on analysis different researches related to traffic classification based on ML. There exists a paper related to traffic categorization using ML [26]. But the researches mentioned in their paper are not up-todate. My work extended their research to introducing different approaches to classify encrypted traffic such as Skype, GTalk, and SSH.