A Framework for Secure Knowledge Management in Pervasive Computing

Abstract

A feature common to many pervasive computing scenarios is that devices acquire information about their environment from peers through short-range ad-hoc wireless connections and use it to maintain a model of their current context. A fundamental issue in such situations, is that knowledge obtained from peer devices may vary in reliability with devices providing incorrect data either inadvertently out of ignorance or other limitations or intentionally in pursuit of malicious or self-serving goals. We describe a heuristic based on a Bayesian approach to infer which of the received answers is most likely to be correct. The suggested answers and the reputation values of the sources themselves are used to determine the most likely answer. We have implemented the techniques and evaluated them in a prototype system using the Glomosim network simulator, and show that our scheme improves data accuracy in low trust environments.