Statistical SIP traffic modeling and analysis system

Recently, the Session Initiation Protocol (SIP) has been adopted as the signaling protocol for multimedia services. In this paper, we present a SIP traffic modeling and analysis system for Internet telephony services. The proposed modeling process mainly uses statistical analysis schemes, such as the Mahalanobis distance and a weighted sliding windows; this process can build traffic profiles after excluding outlier traffic and can automatically update the generated profiles in a database. By using the proposed analysis process, we can observe not only traffic change but also detect any abnormality of each type of SIP REQUEST message. We describe our approach and present experimental and implementation results in this paper.