Anonymizing building data for data analytics in cross-organizational settings

Abstract

Low-cost sensors are being installed in smart buildings to gather large amounts of sensor data on building operation and occupant comfort. These sensor data enables the development of data-driven applications and the analysis of building use. Many of such applications are cross-organizational because data are being shared between a building owner and a contractor that works with data at different spatial granularities, e.g., an open plan office or a heating ventilation and air conditioning (HVAC) zone. This is a challenge as 1) sharing the sensor data in its original form can reveal performance indexes amongst occupants and can violate occupant s privacy by revealing behavioral patterns; 2) methods proposed by previous work fails to anonymize the limited number of individual sensor streams available at smaller spatial granularities, e.g., at the zone-level. In this paper, we propose a meta-method, Time-slicer for anonymizing datasets with a limited number of individual sensor streams and for variable length to enable zone-level applications on anonymized data. The evaluation of the Time-Slicer shows that the method provides privacy protection with only a few individual data streams as it can replace the need for individual sensors with past data.