[Nrg-l] Guest speaker, Thu Nov 12th @ 1pm - Grad lounge
jmlon at cs.bu.edu
Mon Nov 9 15:04:47 EST 2009
*Title:* Random Traffic Classification, or How to Achieve 99% of
Accuracy Knowing Nothing.
Internet traffic classification is becoming one of the most difficult
task to solve in the present Internet. The proliferation of proprietary
applications, the adoption of Peer-to-Peer mechanism and the leverage on
encryption techniques make traditional tools based on Deep Packet
Inspection (DPI) ineffective. However, traffic classification is an
important tool that allows both network providers, researchers and users
to understand applications' requirements and users' habits. Resource
provisioning, QoS differentiation, traffic engineering, and security
mechanisms rely on traffic classification to correctly identify traffic
and allow the network to properly work.
To overcome the limitations of state of the art traffic
classification algorithms, we develop a novel technique that leverages
the characteristics of protocol format, but ignores the protocol
semantic and synchronization requirements. By applying simple
statistical tests, the proposed classifier is able to assess the amount
of randomness each protocol has, and therefore to correctly identify the
application that generated that traffic.
Results show that the proposed methodology reaches almost perfect
classification, even if considering complex applications such as Skype
or P2P traffic.
Marco Mellia holds Ph.D. in Electronic and Telecommunication Engineering
from Politecnico di Torino, where he is currently an Assistant
Professor. In 1999 he was a Visiting PhD Student at the Computer Science
Department of the Carnegie Mellon University. From February to March
2002 he visited the Sprint Advanced Technology Laboratories Burlingame,
California, working at the IP Monitoring Project (IPMON). He has
co-authored over 100 papers published in international journals and
presented in leading international conferences, all of them in the area
of telecommunication networks.
His research interests cover Internet traffic characterization and
classification and Peer-to-Peer TV application design.
More information about the Nrg-l