| Summary: | Much research has been conducted on anomaly detection by wireless sensor networks (WSNs). The existing WSNs require specialized knowledge and skills to install the sensors in environments such as houses and buildings. Therefore, we have developed a flexible WSN based on small sensor devices that can be easily installed. The users only need to place these sensors at the locations where they want to sense and to provide some information to the server through a web page. Then, these small sensor devices automatically create wireless networks, start communicating with the central server for logging continuous data, and show anomalies by using inference based on a basic Bayesian Network. However, a serious problem is that a large amount of noise data prevents correct inferences. Therefore, in this paper, we propose a method for reducing noise data based on location sampling of real human movements. Our experimental results show that our method is effective in increasing the inference accuracy for detecting anomaly data.
|
|---|
