A Simulation Study of Connectivity Metrics for Wireless Ad Hoc Networks

TitleA Simulation Study of Connectivity Metrics for Wireless Ad Hoc Networks
Publication TypeThesis
Year of Publication2006
AuthorsNelson, S. C.
AdvisorPerrone, F. L.
Academic DepartmentComputer Science
DegreeBachelor of Science
UniversityBucknell University
Thesis TypeUndergraduate Honors Thesis
AbstractWireless ad hoc networks are a new technology, and the focus of much research. We have found, however, little literature investigating the quantification of network health, in a simulation environment. In this thesis, we propose a formulation and solution to the problem of summarizing the connectivity information of the network into a single real number, called a metric value. Using a divide-and-conquer based approach, we first summarize the useful connectivity information, for our purposes, into a weighted, undirected graph. The weight of an edge represents the ratio of discrete message units, called packets, successfully delivered over the link represented by that edge in a given time period. Using this graph as the foundation, the second step consists of applying a function, called a metric algorithm, to the graph to obtain a metric value. After cataloging a set of metric relevant algorithms, we use hypothesis testing to determine which metric algorithms best detect changes in network connectivity, due to malicious attacks on the network. Since all of our studies are done in a simulation environment, we have the ability to adjust the severity and time of an attack, allowing us to study how well each metric algorithm detects attacks under specific conditions. The results of this testing allow us to compare the metric algorithms and determine different sets of these algorithms best suited for detecting attacks within a specific range of simulation parameters.
Full Text