Detecting a Proficient Flow Label Propagation from Network Traffic Classification

Abstract

The system model in a traffic classification method introduces the threefunctions that are flow label propagation, nearest cluster based classifier and compound classification. For some special cases, the flow label propagation is critical to require that the labeled flows and unlabelled flows must be captured on the same network in a short period of time. In flow label propagation the peer whose initial flow we have labeled may be off-line, so no other peer will connect to it and flow label propagation is useless for this case. Same for servers, say, and FTP server is popular on a network, and then you will correctly propagate flow information to all other hosts connecting to that server. However, if the server changes its IP address (without changing its DNS name), then will fail to propagate its label. For these type of case propose a alter flow label propagation and also detect the unknown flow in the network