Method for detection of quasiperiodic intensity spikes in multiservice networks with mobile subscribers

DOI: 10.31673/2412-9070.2026.017409

Abstract

The article proposes a new approach to analyzing multiservice network traffic based on the detection of a previously unused information feature – the presence of quasi-periodic intensity spikes in self-similar traffic. Unlike existing methods based on spectral, entropy, or statistical characterris tics, the proposed method uses binarization of traffic time series by an adaptive threshold (k-sigma) followed by analysis of the autocorrelation function of the indicator series of bursts. This allows you to effectively detect recurring, albeit non-periodic, traffic patterns that are characteristic of user be havior in Triple Play and Quadruple Play environments. The key advantage of the method is its robustness to destabilizing factors with a priori incomplete statistical characteristics – in particular, random emissions, activity fluctuations, subscriber mobility, and background noise. This robustness is explained by the fact that the analysis is carried out on a binary data series in which the influence of absolute anomalies is eliminated.The software implementation of the method in Python confirmed its high sensitivity to quasi-periodic structures (up to 100% in test scenarios) and low false positive rate (< 5%) compared to FFT and entropy approaches, which demonstrate significant sensitivity to non-stationarities. The research results expand the capabilities of monitoring, diagnosing, and predicting load in modern mixed-type networks, in particular in mobile broadband access environments. The proposed feature can be integrated into intelligent network traffic analysis systems, IDS/IPS, and also used to improve the efficiency of QoS mechanisms and adaptive resource management.

Keywords: self-similar traffic; quasi-periodic bursts; information signature; robustness; multi-service networks; Quadruple Play; autocorrelation analysis.