Reputation-based Framework for High Integrity Sensor Networks

In Accepted for the ACM Transactions on Sensor Networks , 35 pages , ACM , November 2007.

NESL Technical Report #: TR-UCLA-NESL-200711-02

Sensor network technology promises a vast increase in automatic data collection capabilities through efficient deployment of tiny sensing devices. The technology will allow users to mea- sure phenomena of interest at unprecedented spatial and temporal densities. However, as with almost every data-driven technology, the many benefits come with a significant challenge in data reliability. If wireless sensor networks are really going to provide data for the scientific community, citizen-driven activism, or organizations which test that companies are upholding environmental laws, then an important question arises: How can a user trust the accuracy of information provided by the sensor network? Data integrity is vulnerable to both node and system failures. In data collection systems, faults are indicators that sensor nodes are not providing useful information. In data fusion systems the consequences are more dire; the final outcome is easily affected by corrupted sensor measurements, and the problems are no longer visibly obvious. In this paper, we investigate a generalized and unified approach for providing information about the data accuracy in sensor networks. Our approach is to allow the sensor nodes to develop a community of trust. We propose a framework where each sensor node maintains reputation metrics which both represent past behavior of other nodes and are used as an inherent aspect in predicting their future behavior. We employ a Bayesian formulation, specifically a beta reputation system, for the algorithm steps of reputation representation, updates, integration and trust evolution. This framework is available as a middleware service on motes and has been ported to two sensor network operating systems, TinyOS and SOS. We evaluate the efficacy of this framework using multiple contexts: (1) a lab-scale testbed of Mica2 motes, (2) Avrora simulations, and (3) real data sets collected from sensor network deployments in James Reserve. Note: this is an advanced copy in the interest of early dissemination.

• Saurabh Ganeriwal
• Laura Balzano
• Mani B Srivastava

• PDF file of paper

Journal Paper