Protocol oblivious classification of multimedia traffic

Abstract

SummaryVoice and video over IP are becoming increasingly popular and represent the largest source of profits as consumer interest in online voice and video services increases, and as broadband deployments proliferate. In order to tap the potential profits that VoIP and IPTV offer, carrier networks have to efficiently and accurately manage and track the delivery of IP services. The traditional approach of using port numbers to classify traffic is infeasible due to the usage of dynamic port number. In this paper, we focus on a statistical pattern classification technique to identify multimedia traffic. Based on the intuitions that voice and video data streams show strong regularities in the packet inter-arrival times and the associated packet sizes when combined together in one single stochastic process, we propose a system, called VOVClassifier , for voice and video traffic classification. VOVClassifier is an automated self-learning system that classifies traffic data by extracting features from frequency domain using Power Spectral Density analysis and grouping features using Subspace Decomposition. We applied VOVClassifier to real packet traces collected from different network scenarios. Results demonstrate the effectiveness and robustness of our approach that is capable of achieving a detection rate of up to 100% for voice and 96.5% for video while keeping the false positive rate close to 0%.

Exploiting Adaptive Packet-Sampling Measurements for Multimedia Traffic Classification

“…The evolution of the Internet has induced various multimedia applications (such as Skype, MSN, IPTV, etc.) [2], [3], and the mobility of users will likely translate into the mobility of services. The explosion of the multimedia services implies that we should reconsider the profound connotation of the data traffic.…”

“…Among the different applications of traffic measurement, traffic classification has recently attracted considerable attention. Traditional traffic classification based on well-known transport layer port numbers becomes unreliable, due to the fact that emerging Internet applications tend to mask their identifications by using random ports [3]. There exist three other classification methods in recent years: (1) DPI (deep packet inspection) which searches for the known signatures in the packet payload; (2) host behavior based methods to dig the hidden connection patterns between hosts; (3) machine learning based methods using statistical features of traffic such as packet size, flow duration [16].…”

“…We characterize the distortion introduced by sampling by means of two statistical metrics, namely the sampling loss and the sampling loss coefficient between the sampled and unsampled values. As for the impact of sampling on traffic classification, we evaluate the classification performance through two classifiers: support vector machine (SVM), a widely known supervised classification algorithm, and VOVClassifier, an automated self-learning system that classifies traffic data by extracting features from frequency domain [3].…”

“…Therefore, some features are more important and representative than others for classification purpose. Fan et al [3] demonstrates that multimedia traffic show strong regularities in these two features (PS and IAT), using which the classifier can achieve a high accuracy of the multimedia traffic classification. This implies that these two features are the very important information contents of a flow, and there exist correlation between PA and IAT.…”

“…Since sampled traffic has rarely been used for multimedia traffic classification, it is still unclear which algorithm might be the most apt to deal with such data. Motivated by work such as Fan et al [3], which proposes a VOVClassifier extracting features from frequency domain through two features: PS and IAT, our choice falls on two classification algorithms: VOVClassifier and SVM.…”