21.08.2013 Views

Protocols for Secure Communication in Wireless Sensor Networks

Protocols for Secure Communication in Wireless Sensor Networks

Protocols for Secure Communication in Wireless Sensor Networks

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

3.5. Approximat<strong>in</strong>g End-to-End Security 93<br />

work. This measure is def<strong>in</strong>ed only if there is at least one live path <strong>in</strong> the network.<br />

It is consistent with end-to-end security techniques: <strong>for</strong> these, it always<br />

yields 1, <strong>in</strong>dependent of the number of compromised nodes <strong>in</strong> the network,<br />

s<strong>in</strong>ce every live path is also functional.<br />

Byzant<strong>in</strong>e Agreement<br />

In many applications of distributed systems, at some po<strong>in</strong>t a consensus problem<br />

has to be solved. For example, the hosts have to agree whether or not to<br />

per<strong>for</strong>m a specific action, such as committ<strong>in</strong>g a database transaction. In sensor<br />

networks, nodes may have to agree on the value of aggregated sensor data be<strong>for</strong>e<br />

report<strong>in</strong>g it. Or a distributed <strong>in</strong>trusion detection system is concerned with<br />

the expulsion of a sensor node that is suspected to falsify sensor read<strong>in</strong>gs.<br />

The problem of reach<strong>in</strong>g consensus <strong>in</strong> the presence of malicious faults is<br />

called the Byzant<strong>in</strong>e agreement problem. It is well-known that solutions to this<br />

problem exist only under specific conditions on the synchronization of hosts,<br />

the characteristics of the communication network, and the authentication of<br />

messages. We will not go <strong>in</strong>to detailed descriptions of appropriate conditions.<br />

We are <strong>in</strong>terested <strong>in</strong> evaluat<strong>in</strong>g the ability of a network to reach consensus<br />

when it is subject to an attack. This evaluation provides a metrics <strong>for</strong> the level<br />

of security that is delivered by the network.<br />

Our model is a synchronous system with po<strong>in</strong>t-to-po<strong>in</strong>t connections. In a<br />

fully connected network, this would allow <strong>for</strong> Byzant<strong>in</strong>e agreement <strong>in</strong> case<br />

there are n > 3t nodes <strong>in</strong> the network. S<strong>in</strong>ce we are deal<strong>in</strong>g with a sparsely<br />

connected, multi-hop network, message authentication is used to simulate full<br />

connectivity, i.e. provide resilient po<strong>in</strong>t-to-po<strong>in</strong>t connections. Digital signatures<br />

would allow tolerat<strong>in</strong>g arbitrary values of t, but we disregard this possibility<br />

here and concentrate on end-to-end security properties.<br />

The synchrony assumption is a strong assumption to be made <strong>in</strong> a sensor<br />

network. This assumption demands that messages are reliably transmitted between<br />

nodes with<strong>in</strong> a “round” of operation. This requires a reliable message<br />

transport service that retransmits lost messages. Transmission failures have to<br />

be detected. Such a service should be possible to implement <strong>in</strong> a sensor network,<br />

though it may be unusable <strong>in</strong> practice.<br />

<strong>Protocols</strong> <strong>for</strong> distributed consensus are very complex. In order to tolerate t<br />

faulty nodes, they require at least t + 1 rounds of message exchanges between<br />

all node pairs. It is clearly unacceptable <strong>in</strong> a large-scale sensor network to <strong>in</strong>clude<br />

all nodes <strong>in</strong> such a protocol. Thus, we do not consider Byzant<strong>in</strong>e agreement<br />

among all nodes <strong>in</strong> a sensor network to be of practical value. However,

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!