Abstract
With the proliferation of sensor-embedded mobile computing devices, mobile sensing is becoming a popular paradigm to collect information from participating mobile users. Unlike the well-calibrated and well-tested sensor networks, mobile sensing relies on participants with unknown reliability. Data collected from mobile users may be untrustworthy. There are various solutions proposed in the literature for assessing the trustworthiness of the sensing data that describe an individual event or observation. In addition to single-event based trust models, we propose the concept of Provenance Logic, to reason about the logical relations between multiple events by jointly recognizing and linking events from successive sensing observations. We propose an approach that combines logical reasoning and statistical learning techniques. To the best of our knowledge, our work is the first attempt for trust evaluation based on the logical relation among multiple events in the mobile sensing context. We motivate and illustrate our approach with a use case of traffic monitoring mobile sensing. Performance validation has shown that improved trust assessment can be achieved efficiently and effectively on top of single-event based analysis.