Neural network diagnosis of anomalous network activity in telecommunication systems

Abstract

This paper describes the technology of artificial neural network application to solve the problem of anomalous network activity diagnosis. We offer methods for network activity data collection and training set formation. We select network packets parameters whose values together with network activity characteristics constitute the sample for artificial neural network training. We offer artificial neural network structure, train this network, estimate it's value and classification ability. We show the possibility of the effective use of artificial neural network model composed of intelligent system of anomalous network activity diagnosis.