Novel statistical network model: The hyperbolic distribution

Document Type

Article

Publication Date

12-1-2004

Abstract

A novel statistical model, the one-dimensional hyperbolic distribution, is proposed for network traffic activity. This includes the modelling of network traffic performance and descriptive parameters, organised as probability distribution functions of activity monitored during an observation window. This model promises accurate representation in a compact manner, of particular interest in statistical anomaly intrusion detection systems. The hyperbolic distribution has been tested in modelling several performance parameters in data traces of IP traffic collected from various actual network environments, including the campus LANs of the New Jersey Institute of Technology and the WAN of the Bergen County Cooperative Library Network System (BCCLNS), as well as simulated client-server LANs. By fitting the model parameters to the experimental data, it is shown that this analytic model performs significantly better than the models currently in common use, such as the Weibull, Pareto, normal and lognormal distributions. © IEE, 2004.

Identifier

11844293432 (Scopus)

Publication Title

IEE Proceedings Communications

External Full Text Location

https://doi.org/10.1049/ip-com:20040522

ISSN

13502425

First Page

539

Last Page

548

Issue

6

Volume

151

This document is currently not available here.

Share

COinS