Volume anomaly detection in data networks : an optimal detection algorithm vs. the PCA approach

The crucial future role of Internet in society makes of network monitoring a critical issue for network operators in future network scenarios. The Future Internet will have to cope with new and different anomalies, motivating the development of accurate detection algorithms. This paper presents a no...

Descripción completa

Detalles Bibliográficos
Autores: Casas, Pedro, Fillatre, Lionel, Vaton, Sandrine, Nikiforov, Igor
Tipo de recurso: capítulo de libro
Estado:Versión publicada
Fecha de publicación:2009
País:Uruguay
Institución:Universidad de la República
Repositorio:COLIBRI
Idioma:inglés
OAI Identifier:oai:colibri.udelar.edu.uy:20.500.12008/38649
Acceso en línea:https://hdl.handle.net/20.500.12008/38649
https://doi-org.proxy.timbo.org.uy/10.1007/978-3-642-04576-9_7
Access Level:acceso abierto
Palabra clave:Network monitoring and traffic analysis
Network traffic modeling
Optimal volume anomaly detection
Telecomunicaciones
Descripción
Sumario:The crucial future role of Internet in society makes of network monitoring a critical issue for network operators in future network scenarios. The Future Internet will have to cope with new and different anomalies, motivating the development of accurate detection algorithms. This paper presents a novel approach to detect unexpected and large traffic variations in data networks. We introduce an optimal volume anomaly detection algorithm in which the anomaly-free traffic is treated as a nuisance parameter. The algorithm relies on an original parsimonious model for traffic demands which allows detecting anomalies from link traffic measurements, reducing the overhead of data collection. The performance of the method is compared to that obtained with the Principal Components Analysis (PCA) approach. We choose this method as benchmark given its relevance in the anomaly detection literature. Our proposal is validated using data from an operational network, showing how the method outperforms the PCA approach