A novel aggregated statistical feature based accurate classification for internet traffic

Traffic Classification plays a vital role and is the premise for the modern era of network security and management. This technology categorizes network traffic into several traffic classes based on some fusion of parameters. A number of restrictions have been revealed by the older methods like port based, payload based, and heuristics based classification. Due to inadequate classifier performance in each aspect, the overall classification accuracy is affected while small training samples are used. Hence statistical feature based approach incorporating supervised machine learning techniques are used here to analyze the network applications. This paper proposes a novel approach which combines Hidden Naive Bayes (HNB) and KStar (K*) lazy classifier for accurate classification. Correlation based feature selection (CFS) and Entropy based Minimum Description Length (ENT-MDL) discretization method is also used as a pre-processing task. The proposed system is analyzed and compared with other Bayesian models and lazy classifiers and the experimental results shows better outcomes compared with the state of the art methods.