Sensor Network Security - Networked Embedded Systems - page 211

among each other by some high bandwidth link (e.g., IEEE .). The base stations furthermore
communicate with sensor nodes over a low bandwidth link. As not all sensor nodes can communicate
directlywiththebasestation,multi-hopcommunicationisusedinthesensornetworktorelayqueries
or commands sent by the base station to all sensors, as well as to send back the answers from sensor
nodes to the base station. If multiple sensors contribute to one query, partial results may be aggre-
gated as they flow toward the base station. To communicate results or report events to an application
residing outside of the sensor network, one or more base stations may be connected to a classical
infrastructure network.
As the above description already points out, there are significant differences between wireless sen-
sor and so-called ad hoc networks, to which they are often compared. Both types of networks can be
differentiated more specifically by considering the following characteristics [KWa]:
• Sensor networks show distinctive application-specific characteristics, e.g., depending on
its application, a sensor network might be very sparse or dense.
• Interaction of the network with its environment may cause rather bursty traffic patterns.
Consider, e.g., a sensor network deployed for detecting/predicting earthquakes or fire
detection.Mostofthetime,therewillbelittletraic,butifanincidenthappensthetraic
load will increase heavily.
• Scale of sensor networks is expected to vary between tens and thousands of sensors.
• Energy is even more scarce than in ad hoc networks as sensors will be either battery-
powered or powered by environmental phenomena (e.g., vibration).
• Self-configurability will be an important feature of sensor networks. While this require-
ment also exists for ad hoc networks, its importance is even higher in sensor networks, as
for example human interaction during coniguration might be prohibitive, the geographic
position of sensor nodes has to be learned, etc.
• Regarding dependability and quality of service (QoS), classical QoS notions like through-
put, jitter, etc. are of little interest in sensor networks, as the main requirement in such
networks is the plain delivery of requested information, and most envisaged applications
only pose low bandwidth requirements.
• As sensor networks follow a data-centric model, sensor identities are of little interest, and
new addressing schemes, e.g., based on semantics or geography, are more appealing.
• Required simplicity of sensor nodes in terms of operating system, networking software,
memory footprint, etc. is much more constraining than in ad hoc networks.
Thus far, we have mainly described sensor networks according to their intrinsic characteristics, and
regardingtheirsecurity,wehaveonlystatedthatprincipallythesamesecurityobjectivesneedtobe
met as in other types of networks. his leads to the question, what makes security in sensor networks
a genuine area of network security research?
To give a short answer, there are three main reasons for this. First, sensor nodes are deployed under
particularly harsh conditions from a security point of view, as there will often be a high number of
nodes distributed in a (potentially hostile) geographical area, so that it has to be assumed that at least
some nodes may get captured and compromised by an attacker. Second, the severe resource con-
straints of sensor nodes in terms of computation time, memory, and energy consumption demand
for very optimized implementation of security services, and also lead to a very unfair power balance
between potential attacker (e.g., equipped with a notebook) and defender (cheap sensor node). hird,
the specific property of sensor networks to aggregate (partial) answers to a request as the information
flows from the sensors toward the base station calls for new approaches for ensuring the authen-
ticity of sensor query results, as established end-to-end security approaches are not appropriate
for this.