USE OF THE STATISTICAL ANALYSIS METHODS TO DETECT VOIP NETWORK TRAFFIC ANOMALIES

Abstract

The purpose of the study is to automate detection of signs of technical malfunction and deterioration in the quality of services in the VoIP network. Deterioration of quality leads to a decrease in the volume of services provided, and consequently a profit, a decrea se in customer loyalty and loss of market share. There are three levels of quality degradation in VoIP networks: lack of access at the data network layer, inability to establish a voice connection and lack of access at the level of application services, and degradation of service quality. The analysis of absolute and relative statistical quality indicators of IP-telephony in accordance with the requirements of the international ITU-T standards has been performed, and main indicators and formulas for their calculation are listed. The values of statistical quality indicators are periodically calculated for external channels, subscriber groups and tariff directions. The primary data is CDR (call data records). To detect anomalous changes in the values of the quality indicators it is proposed to use the exponential smoothing method – the Holt-Winters additive model. The deviation of the current value of the quality indicators from the forecasted one is calculated. The confidence range is calculated using the Brutlag method. If the deviation goes beyond the confidence range, the change in the value of the indicator is considered anomalous and the value of the anomaly coefficient of the indicator is set. The period is characterized by a vector of anomaly coefficients of all quality indicators. To classify the period as anomalous, the value of the module of the anomaly coefficients vector of the given period is used. The features of the method application are also considered, in particular the choice of the seasonal period and the calculation of anomaly coefficients during periods of minimum load. The proposed method allows to diagnose an abnormal change in the values of VoIP service quality indicators in automated mode. The prototype of an automated system for the quality of VoIP service monitoring was developed on the base of the method described.