Network traffic characterization based on Time Series Analysis and Computational Intelligence.
Adriana Ferrari Dos Santos · Journal of Computational Interdisciplinary Sciences · 2011
This paper presents an approach for computer network traffic characterization by using Time Series Analysis and Computational Intelligence techniques.HTTP network traffic datasets grouped into different periods of day were analyzed under Kurtosis, DFA and SOM-based clustering algorithms.The results obtained from the calculation of DFA and Kurtosis for each value of the attributes of the network session mapped a range of values of kurtosis and DFA regarded as the standard network.Any sessions of traffic whose attribute values when calculated with Kurtosis and DFA result values within the range mapped means the session is "normal" for that day and period.Were also obtained satisfactory results in the characterization of network traffic pattern through the application of clustering technique with rates of diversion and similarity of 10% and 70% respectively.The results have shown that, according to the observed datasets at certain time of day, the clusters may vary within a range of values, thus representing the traffic pattern behavior of the monitored network in specific period of the day and day of the week.