Analyse et modélisation du trafic internet

Abstract : This thesis is a contribution to the field of Internet traffic analysis at the flow level. For traffic engineering purposes like supervision and security for example, it is important to be able to characterize flows, especially the large ones. Due to the very high bit rate and the huge num- ber of flows in IP traffic, it is very difficult to extract on-line statistics on flows. In this thesis we focused on two kinds of on-line algorithms for Internet traffic analysis. In the first part, we developed a new algorithm based on Bloom filters for large flows iden- tification. The advantage of this algorithm is it can adapt to traffic variations. An interesting application to this algorithm is the on-line detection of denial of service attacks. For this purpose, we proposed an adapted algorithm taking into account attacks specificities. On-line experiments show that this new method is able to identify almost all sources of anomalous traffic in a very short delay. In addition, we analyzed the performances of the algorithm for on-line identification of large flows. We analytically estimated the error generated by the algorithm on the number of elephants. The algorithms presented in the first part are performed on the exhaustive traffic which is not usually possible because in some cases we have only access to the sampled traffic. The second part of the thesis is dedicated to sampling analysis and the study of algorithms inferring the original traffic characteristics. First, using a result on Poisson approximations, we proved that the two sampling methods : deterministic and probabilistic give equivalent results in terms of sampled traffic composition at the flow level. Then we developed a new method inferring, from the sampled traffic, via asymptotic procedures, flows number and size distribution in a small time window. This enables us to suppose à priori that this distribution is a Pareto. This hypothesis was validated against different traffic traces.
Complete list of metadatas

Cited literature [66 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00463733
Contributor : Yousra Chabchoub <>
Submitted on : Monday, March 15, 2010 - 1:16:54 AM
Last modification on : Thursday, February 7, 2019 - 1:32:26 AM
Long-term archiving on : Thursday, June 30, 2011 - 12:11:03 PM

Identifiers

  • HAL Id : tel-00463733, version 1

Citation

Yousra Chabchoub. Analyse et modélisation du trafic internet. Informatique [cs]. Université Pierre et Marie Curie - Paris VI, 2009. Français. ⟨tel-00463733⟩

Share

Metrics

Record views

588

Files downloads

616