Harald Vogt Protocols for secure communication in wireless sensor networks. 
PhD thesis No. 18174, ETH Zurich, Zurich, Switzerland, 2009 

Abstract 

Wireless sensor networks are comprised of large numbers of 
resource-constrained and wirelessly communicating computing devices. Advances 
in computing and communication technology have made it possible to integrate 
sensing capabilities, wireless communication interfaces, and microprocessors 
into tiny devices that allow to embed compuational power in arbitrary 
environments. The applications of wireless sensor networks range from 
surveillance and environmental monitoring to healthcare and the provisioning 
of context information for computing applications. Many of these applications 
have a direct impact on the welfare of human beings or are of high economic 
significance. Security breaches might lead to grave consequences, so it is 
important to protect wireless sensor networks against such threats. 

The specific characteristics of wireless sensor networks make them vulnerable 
to attacks on their communication channels and their hardware. Cryptographic 
mechanisms can be employed to protect against some of the possible attacks: 
eavesdropping on messages is countered by encryption, and the injection of 
messages by the attacker is prevented by authentication. Unfortunately, direct 
physical access to the sensor nodes allows an attacker to manipulate them 
almost arbitrarily. In particular, nodes could be compromised and then made to 
execute malicious code injected by the attacker. Tamper resistance mechanisms 
applied to the nodes. hardware, concealment, surveillance and other techniques 
may be used to mitigate such attacks. However, they cannot be completely 
prevented and therefore, any communication security scheme being used must be 
sufficiently resilient to tolerate a certain amount of compromised nodes. 
Consequently an important objective is to limit the impact of a set of 
compromised nodes on the legitimate operation of the network to a minimum. 

This objective can optimally achieved by cryptographic mechanisms that 
establish a direct security relationship between communicating end-points. 
This limits the influence that a single compromised node has to its own 
resources. Thereby, it cannot tamper with messages that originate at other 
nodes. However, such mechanisms are overly resource demanding for many sensor 
nodes in terms of computational or communication complexity, especially due to 
the often ad hoc and transient nature of communication relationships. Thus, 
novel mechanisms are required that provide a sufficient level of security 
while respecting the constraints in wireless sensor networks. 

Our thesis is that key pre-distribution is an appropriate technique for 
secret key agreement in wireless sensor networks, and that based on locally 
shared keys, multi-hop communication can be adequately protected using an 
interleaved message authentication scheme. 

We argue that combined key pre-distribution schemes provide a feasible 
mechanism for key agreement in wireless sensor networks. They require only 
simple operations on sensor nodes and their memory requirements can be adapted 
to the required security level and the available resources. Based on keys 
shared between nodes within a k-hop neighbourhood (with small k), a message 
authentication scheme is devised that allows for the secure transmission of 
messages over long distances. In particular, our contributions are: 

- A key establishment scheme for pairwise key agreement that can be 
efficiently implemented on resource-constrained wireless sensor nodes and 
provides resilience against node capture attacks. 

- A message authentication scheme that relies on locally shared keys and 
symmetric cryptographic operations only, and provides a level of security 
approximating that of end-to-end security mechanisms. The foundation of the 
scheme.s security is the creation of multiple disjoint authentication paths. 

- An evaluation of this authentication scheme showing that it provides at 
least the same security level as a general communication scheme that relies on 
multiple disjoint physical paths. 

The proposed security mechanisms protect the integrity of messages that are 
exchanged within a wireless sensor network. The achievable level of security 
is, given an attacker with moderate strength that is only able to capture a 
fraction of all nodes, comparable to that provided by end-to-end security 
mechanisms at a significantly lower cost in terms of computational resources.