A Home Intrusion Detection System using Recycled Edge Devices and Machine Learning Algorithm

This paper proposes a home intrusion detection system that makes the best use of a retired smartphone and an existing Wi-Fi access point. On-board sensors in the smartphone mounted on an entrance door records signals upon unwanted door opening. The access point is reconfigured to serve as a home server and thus it can process sensor data to detect unauthorized access to home by an intruder. Recycling devices enables a home owner to build own security system with no cost as well as helps our society deal with millions of retired devices and waste of computing resources in already-deployed IT devices. In order to improve detection accuracy, this paper proposes a detection method that employs a machine learning algorithm and an analysis technique of time series data. To minimize energy consumption on a battery-powered smartphone, the proposed system utilizes as few sensors as possible and offloads all the computation to the home edge server. We develop a prototype and run experiments to evaluate accuracy performance of the proposed system. Results show that it can detect intrusion with probability of 95% to 100%.