Research And Implementation Of Feature Extraction Of P2P Streaming Media

In recent years, with P2P technology expanding the scope and P2P streaming media technology continues to expand, P2P technology takes the issue of network resources more and more serious, caused by network service providers and network operators are highly valued. In the current network of limited bandwidth, P2P related applications take up most of the bandwidth, has affected the normal application of the users. Therefore, studying how to identify known and unknown P2P traffic, and its feature extraction and classification, has been a serious problem.Currently P2P traffic detection technology can be divided into two categories: Deep Packet Inspection (DPI) and detection technology based on flow characteristics (Transport Layer Identification), but the detection on P2P streaming media in related research is still in the stage, and a common approach to identify the traffic is signature matching. Based on this situation, the main content of this article has done the following.Proposed a statistical -based algorithm for automatic feature extraction of which achieved good results on known and unknown P2P streaming media applications. Focuses on automatic extraction of signature algorithm, and compared with existing algorithms advantages and disadvantages of the algorithm. In the statistics-based automatic feature extraction algorithm, the fragment contained location information of the data well preserved the packet location information of the load characteristics, through the efficient HASH algorithm to achieve a real-time extraction of large flow environment. And proposed the corresponding merging strategy for the signature and the phase-out strategy to ensure that the signature of the availability and stability.Designed and implemented a prototype system, which consists of three modules: data processing module, automatic feature extraction module, and re-verification module. On this basis, proposed several performance optimization strategies for statistic-based feature extraction algorithm, and optimized with the actual test proved the correctness and efficiency.In real network environment to test the system, the experimental results show that the system in gigabit network environment can achieve a real-time feature extraction of P2P streaming media traffic effectively. Keywords: P2P streaming media, Statistical characteristics, Automatic feature extraction...