An Authentication Service for Sensor and Ad Hoc Networks

Remote sensing applications are becoming an increasingly important area for research and development due to the critical need for applications that will perform environmental monitoring, provide security assurance, assist in healthcare services and facilitate factory automation. In remote sensing scenarios, one or more applications are connected to a sensor network through a communication network. The sensors in the sensor network make measurements, such as local temperature or barometric pressure, and communicate this data with the appropriate application via the network. Providing security mechanisms for sensor networks is of critical importance since sensors will ultimately be used to assist in our daily lives. The authentication of the data source as well as the data are critical concerns since adversaries might attempt to capture sensors and tamper with sensor data. Traditional authentication frameworks based on public key cryptography are not suitable for sensor networks since the sensor network will ultimately consist of small, low-powered devices that are mobile. The limited computational and storage resources available to sensors necessitates alternatives to authentication based on public key certificates.

Recently, a set of security protocols for sensor networks, known as SPINS, has been proposed [163]. SPINS addresses authentication on limited resource sensor networks by introducing two security protocols that rely on the presence of a more powerful basestation and an initial shared secret between the basestation and each participating sensor node: SNEP and μTESLA. SNEP is a simple protocol that provides data confidentiality, two-party data authentication, and evidence of data freshness using only symmetric keys and counters. μTESLA is a modified version of the TESLA protocol, which performs bootstrapping without using a public key infrastructure (PKI) and discloses one key each epoch independently of the packet rate to provide broadcast authentication. Another work that focused on authentication for ad hoc networks was presented in [164]. In this chapter, a distributed light-weight model for authentication was presented that involves network nodes requesting trust references from neighboring nodes in order to establish the trust relationships needed for network authentication. Each entity maintains a list of trusted entities, and using these lists trusted communication paths between two arbitrary entities can be derived. One drawback of this method, however, is its scalability. For large networks, the size of the trust tables can become prohibitive. Another work on authentication for ad hoc networks that addressed the issue of scalability was presented in [165], which introduced the use of cluster heads to reduce the amount of control packets needed. In this work, the network is divided into cluster regions, and cluster heads are elected from the regular network nodes within each cluster. Authentication is provided by using a public key infrastructure that, unfortunately, is not suitable for small sensor devices.

These methods focus on ad hoc networks employing a flat topology. However, ad hoc networks have been recently shown to have capacity limitations, and one approach to address this drawback is to employ a hierarchical ad hoc network. In this chapter we will further explore the advantages of hierarchical ad hoc networks, particularly focusing on the advantages of the hierarchical ad hoc sensor network for performing authentication when compared with flat ad hoc networks. Authentication in hierarchical ad hoc networks has been essentially untouched, and we are aware of only one work in this direction [166], which focused on a military environment. The security of their work is based largely on the assumption that the access points, which corresponded to unmanned aerial vehicles, are unable to be compromised. This is an assumption that does not hold in non-military applications, and therefore we consider a three-tier hierarchical ad hoc network that is suitable for more general remote sensing applications running on the Internet. We develop an authentication framework for our three-tier hierarchical sensor network that addresses the hardware resources of the three-tier network, and employs cryptographic primitives that are appropriate for each type of node.