Unit:
Κατεύθυνση / ειδίκευση Δικτύωση Υπολογιστών (ΔΙΚ)Library of the School of Science
Supervisors info:
Ιωάννης Σταυρακάκης Καθηγητής, Μερκούριος Καραλιόπουλος Επιστημονικός Συνεργάτης
Original Title:
Point centrality indices and ISP network vulnerability
Translated title:
Μετρικές κεντρικότητας και ευαισθησία ISP δικτύων σε κακόβουλες επιθέσεις
Summary:
Social Network Analysis (SNA) primarily studies the networks with
graph-theoretic tools modeling individual actors within the network. One of the
SNA fundamentals is the notion of point centrality which was designed to
reflect the relevant social significance among actors and has lately come to
the attention of computer scientists; the latter seek to identify highly
central network nodes that can be exploited in the design of efficient
protocols. The thesis studies the concept of centrality along three lines:
Firstly, we systematically review the different definitions of centrality
indices that have been proposed throughout the last 40 years or so. We also
iterate on the different interpretations the metric admits and try to relate
them to particular networking applications/properties. Secondly, we explore the
way a set of seven well-known centrality indices rank-correlate over real-world
network topologies. Lastly, we turn our attention to networking applications of
centrality indices studying the vulnerability of router-level ISP topologies.
Namely, we seek to compare the previously studied indices on the basis of their
impact on the network performance i.e., the connectivity and the maximum flow
the network accommodates when highly central nodes are removed. Our results
provide countermeasure guidelines against centrality-driven malicious node
attacks.
Keywords:
Centrality metrics, Correlation, Router-level topologies, Node resilience, Graph theory
Number of index pages:
8-11
Number of references:
109
