A Virtual Infrastructure for Wireless Sensor Network Protocols

Disssertation defense by: Ashraf Wadaa
Advisor: Prof. Stephan Olariu

Abstract: Wireless sensor networks are emerging as a new technology for
instrumenting the physical world.  Researchers can use miniaturization
techniques used in state-of-the-art  semiconductor manufacturing to
build increasingly smaller devices that contain low-power radios,  and
mechanical structures that sense fields and forces in the physical
world. Adding  computing capabilities enables the device to process, and
communicate local sensory information. Networking a population of these
devices enables collaboration among  them to derive global state
information from local sensory information.

A major challenge in wireless sensor networks is network initialization,
the self-organized transition  from an unstructured state, to a state
where both communications and organization structures have been
established. This work focuses on this challenge.

The network addressed consists of a massive number of commodity sensor
nodes and one or more base stations. Each sensor node is a device that
integrates sensing, computing, and wireless communications capabilities.


Individual sensor nodes are inherently  constrained in terms of their
computing, communications, and energy resources. Also,  individual nodes
may not necessarily have unique identifiers at fabrication time. The
network deployment environment is typically remote and invariably
hostile. Initially, nodes are unaware of their physical location or the
network topology. Anonymity of nodes calls for network protocols to
perform addressing and identification in the network at a coarser
granularity than the individual node level. Network longevity subject to
scarcity of node energy resources calls for a mode of operation where
nodes maintain short duty cycles. Thus, network protocols must assume
that nodes have only intermittent connectivity to the network.
Intermittent connectivity also precipitates a highly evolving network
topology. This calls for network protocols to support associations for
collaboration among nodes that are dynamic, transient, and task driven.
In addition, the sheer number of nodes in the network calls for network
protocols that are highly scalable, as well as resource efficient.
Having no physical infrastructure to leverage poses serious challenges
to the development of network protocols that meet the above
requirements.

The first contribution of this work is a  new paradigm for protocol
independent initialization in  wireless sensor networks consisting of
massively large number of anonymous nodes. The paradigm  encompasses
both organization and communications structures. The paradigm is based
on: (1) a  virtual network infrastructure that is dynamic,
reconfigurable, and general purpose in its scope, and (2) network
training;  a novel approach in wireless sensor networks introduced in
this work that is built around a set of scalable and extremely energy
efficient protocols  for initializing the network, i.e. creating the
elements of the virtual infrastructure.

The second contribution of this work are scalable and energy efficient
protocols for task-based work organization and management, leveraging
the virtual infrastructure. The main features of these protocols are:
(1) addressing both network level and sensor node level programming in a
unified framework, and (2) providing an integrated solution for task
formulation, node to task assignment, routing, in-network and en-route
data processing, and MAC, and (3) Can be customized to accommodate
various application dependent task semantics, for example, real time
versus non real time operation, and pull versus push data flow models.

The third contribution of this work are scalable and energy efficient
protocols for group key management, leveraging the virtual
infrastructure.

The main features of these protocols are:(1) Supporting group key
initialization without any communications in the trained sensor network,
thus eliminating all security threats pertinent to communications during
the critical group key initialization process, and (2) Supporting
both backward, and forward secrecy.