Multifractal characterization for classification of network traffic

In this paper, a novel multifractal approach to the classification of self-affine network traffic is presented. The fundamental advantages of using multifractal measures include their boundedness and a very high compression ratio of a signature of the traffic, thereby leading to faster implementations, and the ability to add new traffic classes without redesigning the traffic classifier. The variance fractal dimension trajectory is used to provide a multifractal "signature" for each type of traffic over its duration, and the modelling of its statistical histograms provides further compression and generalization. Principal component analysis is used to reduce the dimensionality of the data, and the K-means clustering algorithm is used to assign classes to the data. A probabilistic neural network (PNN) with a locally optimal spread parameter is trained with these signatures, and a plot of the PNN percentage correct classification accuracy as the number of assigned classes increases reveals that there are most likely three classes in the traffic recording. Finally, an optimized PNN is trained with 50 % of the multifractal signatures sampled at regular intervals from the trajectory, and achieves a representative classification accuracy of 94.8 % when classifying previously unobserved self-affine network traffic.