A Comparison of Classification Techniques for Detection of VoIP Traffic

A new breed of applications for Voice over IP (VoIP) communications has emerged in recent years, providing a viable service for voice communications on the Internet. Mostly provided as over-the-top services, both standards-based Session Initiation Protocol (SIP) VoIP services and proprietary alternatives such as Skype, Google Talk and Yahoo! Voice have earned considerable popularity, mostly because of their low cost and ease of use. However, for various reasons, such as usage policy enforcement, quality assurance, security or specific business interests, there is a need to detect and monitor the presence of VoIP traffic on data networks. The main objective of this paper is to present a survey of techniques for detection and classification of VoIP protocol traffic, with an emphasis on two major categories (profiling of network traffic patterns and modeling of communication flows for anomaly detection) also discussing the several algorithms and techniques within these two groups. Legacy techniques are also briefly described, albeit for the sake of completeness.