Back to search

Institutional Repository of Vilnius University

Neuroninių tinklų metodų taikymo interneto srauto prognozei tyrimas /

Abstract

dc:description

Research of Prediction of Internet Traffic Using Methods of Neural Networks Aim of the work: investigation of adequacy of application of methods of nonlinear dynamics and artificial neural network for internet traffic prediction. Main goals: 1) research on application of methods of nonlinear dynamics for prediction of internet traffic flow, using artificial neural network to determine the reconstruction parameters, 2) research on application of the method of time-delay neural network for prediction of internet traffic flow, 3) application of the internet traffic prediction to detect a DDoS attack. It is very common that internet traffic prediction is required for the computer network to function effectively: traffic prediction is an irreplaceable tool for network optimization, prevention of malfunctions and attacks. In this work, internet traffic flow is interpreted as a chaotic system and is predicted using method of nonlinear dynamics, comparing prediction results that were achieved using different methods of obtaining the reconstruction parameters (delay τ and embedding dimension m) required for the prediction. The methods are mutual information, false nearest neighbors, time-window and artificial neural network. Also, internet traffic flow prediction is achieved using the method of time-delay neural network. In all cases, the methods of prediction are tested using a well-known chaotic time series Mackey Glass. MATLAB code programmed for the purpose of this work allows to determine whether the time series is chaotic by evaluating the Lyapunov coefficient, enables obtaining reconstruction parameters (time delay and embedding dimension) and predicting the series using method of nonlinear dynamics and time-delay neural network. In this work, chaotic time series (Mackey Glass) data and internet traffic flow published by MIT (Massachusetts Institute of Technology) are analyzed. The data of the DDoS attack were published by CAIDA (Center for Applied Internet Data Analysis). Main results and conclusions: 1) the analysis of the results has shown that the method of nonlinear dynamics for internet traffic prediction is not applicable, since there is no single method of determining the reconstruction parameters suitable for all chaotic systems; 2) time-delay neural network method of prediction allows satisfactory long term prediction accuracy of chaotic time series and internet traffic flow; 3) DDoS attack detection, using internet traffic flow prediction by using time-delay neural network was successful.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2017

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Pavlova, Anželika,
Contributors dc:contributor
  • Kuliešius, Feliksas

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Pavlova, Anželika,. Neuroninių tinklų metodų taikymo interneto srauto prognozei tyrimas /. Institutional Repository of Vilnius University, 2017. https://repository.vu.lt/VU:ELABAETD23192591&prefLang=en_US