Network traffic analysis using clustering ants
T. Ekola, Mikko Laurikkala, Teemu Lehto, Hannu J. Koivisto · 2004
This paper exploits a newish self-organizing clustering method to analyze network data. The method is based on cooperative behaviour of ants. Collective intelligence of an ant colony rests on interactions of individual ants and the environment. Hence no central control is needed and the whole process is performed unsupervised by simple agents. The data used in the analysis was recorded from a border gateway of a campus network. The theory of clustering ants is introduced before the clustering process takes place. The target of the research is outlined and preprocessing of the data is described in detail. Several clustering procedures are performed and the results are visualized. Daily variation in network traffic is discovered and the method is found to be working. A way to detect long range changes is presented and some changes are observed in the data. Future work and lacks of the clustering method are discussed.