Performance Improvements in Distributed Estimation and Fusion Induced by a Trusted Core

We consider distributed state estimation and fusion in autonomic ad-hoc sensor networks when some nodes are untrusted, corrupted or malicious. We provide rigorous graph models to capture both trust and network functionality relations. We propose a novel hierarchical scheme, inspired from socio-cognitive dynamics in economics and sociology, that utilizes a trusted core. We provide a component-based architecture to show the interplay between the trust and estimation schemes. As an explanatory example we consider the problem of distributed Kalman filtering. We show that trust-based schemes, implemented by a trusted core, result in significant improvement in the state estimation procedure. We also find that the interplay between estimation and trust updates quickly isolates malicious nodes and helps the observation limited nodes (corrupted sensor nodes).