| dc.description.abstract | Monitoring networks for anomalies is a typical duty of network operators. The conventional
monitoring tools available today tend to almost ignore the topological characteristics
of the whole network. This thesis takes a different approach from the conventional
monitoring tools, by employing the principle of Eigenvector Centrality. Traditionally,
this principle is used to analyse vulnerability and social aspects of networks.
The proposed model reveals that topological characteristics of a network can be used
to improve the conventional unreliability predictors, and to give a better indicator
of its potential weaknesses. An effective expected adjacency matrix, k, is introduced in
this work to be used with centrality calculations, and it reflects the factors which affect
the reliability of a network, for e.g. link downtimes, link metrics, packet loss,
etc. Using these calculations, all network backbone routers are assigned values which
correspond to the importance of those routers in comparison to the rest of the network
nodes. Furthermore, to observe how vulnerable each node could be, nodes are
ranked according to the importance values, where the nodes with high ranking values
are more vulnerable. This model is able to analyse temporal stability of the network,
observing and comparing the rate of change in node ranking values and connectivity
caused by the network link failures. The results show that the proposed model is
dynamic, and changes according to the dynamics of the topology of the network, i.e.
upgrading, link failures, etc. | en_US |
