| |
This paper addresses the problem of intelligent traffic analysis in multivendor wired networks to detect performance anomalies effectively. Due to the high degree of telemetry data fragmentation caused by the use of heterogeneous network equipment from various vendors, traditional monitoring approaches prove inadequate for modern corporate infrastructures.
The study substantiates the need to develop adaptive analytical systems capable of operating under conditions of partial observability and high traffic dynamics. A taxonomy of network traffic analysis methods is proposed, categorized into statistical models, machine learning algorithms, and hybrid approaches. Evaluation criteria for algorithm performance are outlined, including accuracy, detection delay, computational overhead, and noise robustness. A metric normalization mechanism is developed to unify telemetry formats across vendors, along with an adaptive algorithm selection system based on current resource utilization.
Experimental validation on real and synthetic datasets demonstrates the effectiveness of the proposed approach, showing reduced false positives and improved detection accuracy. The study concludes with outlining future research directions, such as the implementation of quantum-inspired algorithms, federated learning for distributed environments, and the use of digital twins for predictive network diagnostics.
Keywords:intelligent traffic analysis, multivendor networks, anomaly detection, machine learning, statistical models, telemetry, hybrid algorithms, network monitoring, federated learning
|
