Automated Sensor Verification Using Outlier Detection in the Internet of Things

Cyber-physical systems uses sensors, actuators, and computation to solve problems that cross the physical world and computational world. Traditionally, the systems have been designed to be specific to one application domain and to be managed by a single entity. However, to achieve scalability, it is necessary to reuse sensors and actuators across multiple application domains. This broad use of sensors and actuators, as envisioned for the Internet of Things, raises many new problems related to interoperability, resource sharing, security and trust, and economics. In this paper, we address the problem of trust - specifically how to verify sensor information that is gathered from multiple sensors that are managed by different entities using outlier detection. We present a technique for automatically deriving a model of the physical phenomenon that is measured by the sensors. This model is then used to compare sensor readings and to identify outliers through spatial and temporal interpolation. We evaluate our system and demonstrate its effectiveness in the context of weather sensing. Nevertheless, the technique is applicable to any application domain where the underlying phenomenon is continuous.