Publication: P2P traffic classification for residential network
Issued Date
2016-02-08
Resource Type
Other identifier(s)
2-s2.0-84964370971
Rights
Mahidol University
Rights Holder(s)
SCOPUS
Bibliographic Citation
ICSEC 2015 - 19th International Computer Science and Engineering Conference: Hybrid Cloud Computing: A New Approach for Big Data Era. (2016)
Suggested Citation
Channary Thay, Vasaka Visoottiviseth, Sophon Mongkolluksamee P2P traffic classification for residential network. ICSEC 2015 - 19th International Computer Science and Engineering Conference: Hybrid Cloud Computing: A New Approach for Big Data Era. (2016). doi:10.1109/ICSEC.2015.7401433 Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/43527
Research Projects
Organizational Units
Authors
Journal Issue
Thesis
Title
P2P traffic classification for residential network
Other Contributor(s)
Abstract
© 2015 IEEE. Excessive bandwidth consuming by peer-to-peer (P2P) applications is one of serious problems in residential networks such as in dorms, apartments and even Small and Medium-sized Enterprises (SMEs) networks which have a limited bandwidth. P2P file sharing and P2P streaming applications usually are the cause of this problem. To share the bandwidth fairly among users, the traffic of these applications needs to be classified and filtered out. However, traditional port-based and payload-based classification will fail when the applications use dynamic ports, port disguise and payload encryption. In this paper, we present the classification technique that based on characteristics of number of peer connection and number of traffic in both incoming and outgoing direction for 5-minute duration to classify the P2P traffic. We make use of decision tree J48 to model and classify the traffic. Experimental results over three well-known P2P applications (BitTorrent, Skype and SopCast) confirm that this technique can detect the existence of P2P traffic from the background traffic with 100% accuracy and can classify three types of P2P applications with 90% accuracy.