Detecting Smart Plug Configuration Changes in Smart Homes

Abstract

Over recent years several smart home systems have emerged that use wireless sensors - called smart plugs - for measuring and controlling electrical consumers. These sensors provide the basis for applications like home automation and energy monitoring. However, the application logic requires correct metadata about the association of sensors with electrical devices. Today's solutions lack technical means for ensuring correctness of the associations but rely on the users diligence. In this paper we present a solution that uses machine-learning to automate tasks of metadata management for smart homes. The solution is tailored to the limited sensing capabilities of typical smart plugs. We also present experimental results based on real-world data from a pilot with several smart home installations. The experiments give insights into the applicability of different machine learning algorithms, suitable feature sets, and the overall performance of the solution.