Fault perturbations in building sensor network data streams

The Critter temperature sensor device has been developed to understand the full nature of pervasive sensor networks. Off the shelf integrated sensor devices incorporate some amount of adaptation to make the devices more reliable. The Critter provides raw instantaneous readings including outlier data that may be considered anomalies or perturbations. We have deployed Critter data sensors pervasively through one academic building for almost 18 months. This paper explores the causes of temperature data perturbations, defined as two temperature data readings taken within seconds of each other that differ by several degrees. Temperature sensor data perturbations are actually the effects of user activity within buildings. By capturing the raw data without automatic processing, we are able to show a correlation between time of the work day and the frequency of data perturbation.