Provenance in sensornet republishing

Abstract

Sensornets are being deployed and increasingly brought on-line to share data as it is collected. Sensornet republishing is the process of transforming on-line sensor data and sharing the filtered, aggregated, or improved data with others. We explore the need for data provenance in this system to allow users to understand how processed results are derived and detect and correct anomalies. We describe our sensornet provenance system, exploring design alternatives and quantifying storage trade-offs in the context of a city-sized temperature monitoring application. In that application, our link approach outperforms other alternatives on saving storage requirement and our incremental compression scheme save the storage further up to 83%.