CONCLUSIONS
We identified opportunities to improve the Internet of Things, proposing the creation of a new architecture with Quality of Data Targets, Security and Cognitive Layers, mathematical-model based Data Proxies, and an Application Agent to optimizing sampling costs or minimizing error subject to constraints. Building upon the human model of applying context and cognition to data management, our architecture abstracts physical from digital systems to improve security and efficiency. It applies context information to supervise systems and to protect them against malicious commands, fuses data to provide difficult to obtain measurements, and uses estimation to minimize sampling cost. Together with clear ownership policies and data sharing visualizations [10], [11], this architecture’s use of abstraction and creation of “black boxed” aggregate data addresses privacy concerns. Using the practical application of Usage Based Insurance, we demonstrated that Proxy models which are well calibrated to an underlying physical process may allow us to reduce the energy necessary to represent that process in the cloud. We demonstrated that querying information does not require one-to-one sampling of the sensors instrumenting that process, and showed that it is possible to substantially minimize costs without significantly increasing measurement error. This level of abstraction and sensor fusion improves security by eliminating applications’ direct access to physical systems and preventing the long-term storage of sensitive data. Further, this same technique may be used to minimize data transmitted, conserving costly bandwidth.
