Abstract:

Since the year of 2006, peer-to-peer (P2P) streaming media service has been developing rapidly, the user scale and income scale achieve synchronous growth. However, while people enjoying the benefits of the distributed resources, a great deal of network bandwidth is consumed at the same time. Research on P2P streaming traffic characteristics and identification is essential to Internet service providers (ISPs) in terms of network planning and resource allocation. In this paper, we introduce the current common P2P traffic detection technology, and analyze the payload length distribution and payload length pattern in one flow of four popular P2P streaming media applications. Combining with the deep flow inspection and machine learning algorithm, a nearly real-time identification approach for P2P streaming media is proposed. The experiments proved that this approach can achieve a high accuracy with low false positives.

Key words:

deep flow inspection, machine learning, payload length distribution, traffic identification